ERROR: LoadError: Enzyme compilation failed.
Current scope: 
; ModuleID = 'start'
source_filename = "start"
target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64-ni:10:11:12:13"
target triple = "nvptx64-nvidia-cuda"

%printf_args.0 = type { i64 }

@0 = private unnamed_addr constant [36 x i8] c"ERROR: Out-of-bounds array access.\0A\00", align 1
@exception4 = private unnamed_addr constant [10 x i8] c"exception\00", align 1
@exception6 = private unnamed_addr constant [10 x i8] c"exception\00", align 1
@exception7 = private unnamed_addr constant [10 x i8] c"exception\00", align 1
@exception9 = private unnamed_addr constant [13 x i8] c"bounds error\00", align 1
@.str = private unnamed_addr constant [11 x i8] c"__CUDA_FTZ\00", align 1
@.str.2 = private unnamed_addr constant [17 x i8] c"__CUDA_PREC_SQRT\00", align 1
@.str.1 = private unnamed_addr constant [12 x i8] c"__CUDA_ARCH\00", align 1
@__cudart_i2opi_f = internal unnamed_addr addrspace(1) constant [6 x i32] [i32 1011060801, i32 -614296167, i32 -181084736, i32 -64530479, i32 1313084713, i32 -1560706194], align 4
@__cudart_i2opi_d = internal unnamed_addr addrspace(1) constant [18 x i64] [i64 7780917995555872008, i64 4397547296490951402, i64 8441921394348257659, i64 5712322887342352941, i64 7869616827067468215, i64 -1211730484530615009, i64 2303758334597371919, i64 -7168499653074671557, i64 4148332274289687028, i64 -1613291254968254911, i64 -1692731182770600828, i64 -135693905287338178, i64 452944820249399836, i64 -5249950069107600672, i64 -121206125134887583, i64 -2638381946312093631, i64 -277156292786332224, i64 -6703182060581546711], align 8
@__cudart_sin_cos_coeffs = internal unnamed_addr addrspace(1) constant [16 x double] [double 0x3DE5DB65F9785EBA, double 0xBE5AE5F12CB0D246, double 0x3EC71DE369ACE392, double 0xBF2A01A019DB62A1, double 0x3F81111111110818, double 0xBFC5555555555554, double 0.000000e+00, double 0.000000e+00, double 0xBDA8FF8320FD8164, double 0x3E21EEA7C1EF8528, double 0xBE927E4F8E06E6D9, double 0x3EFA01A019DDBCE9, double 0xBF56C16C16C15D47, double 0x3FA5555555555551, double -5.000000e-01, double 1.000000e+00], align 8
@1 = private unnamed_addr constant [108 x i8] c"ERROR: a %s was thrown during kernel execution.\0A       Run Julia on debug level 2 for device stack traces.\0A\00", align 1
@2 = private unnamed_addr constant [110 x i8] c"WARNING: could not signal exception status to the host, execution will continue.\0A         Please file a bug.\0A\00", align 1

; Function Attrs: nofree readnone
declare {}*** @julia.get_pgcstack() local_unnamed_addr #0

declare i32 @vprintf(i8*, i8*) local_unnamed_addr #1

; Function Attrs: argmemonly nofree nounwind willreturn
declare void @llvm.memcpy.p0i8.p11i8.i64(i8* noalias nocapture writeonly, i8 addrspace(11)* noalias nocapture readonly, i64, i1 immarg) #2

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.ctaid.x() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.ctaid.y() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.ctaid.z() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.ntid.x() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.ntid.y() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.ntid.z() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.tid.x() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.tid.y() #3

; Function Attrs: nounwind readnone speculatable
declare i32 @llvm.nvvm.read.ptx.sreg.tid.z() #3

; Function Attrs: argmemonly nounwind readonly
declare float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* nocapture, i32) #4

; Function Attrs: noinline noreturn
define internal fastcc void @julia__throw_boundserror_25134() unnamed_addr #5 !dbg !48 {
top:
  %0 = call {}*** @julia.get_pgcstack()
  %1 = call i32 @vprintf(i8* noundef getelementptr inbounds ([36 x i8], [36 x i8]* @0, i64 0, i64 0), i8* noundef null), !dbg !50
  call fastcc void @gpu_report_exception(i64 noundef ptrtoint ([10 x i8]* @exception4 to i64)), !dbg !66
  call fastcc void @gpu_signal_exception(), !dbg !66
  call void asm sideeffect "exit;", ""() #139, !dbg !66
  unreachable, !dbg !66
}

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.ctlz.i32(i32, i1 immarg) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i64 @llvm.ctlz.i64(i64, i1 immarg) #6

declare i32 @__nvvm_reflect(i8*) local_unnamed_addr #7

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.floor.ftz.f(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.floor.f(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.floor.d(double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fabs.ftz.f(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fabs.f(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.fabs.d(double) #6

; Function Attrs: nofree nosync nounwind readnone willreturn
declare double @llvm.nvvm.rcp.approx.ftz.d(double) #8

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fmin.ftz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fmin.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fmax.ftz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fmax.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.rsqrt.approx.ftz.f(float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.rsqrt.approx.f(float) #8

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.fmin.d(double, double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.fmax.d(double, double) #6

; Function Attrs: nofree nosync nounwind readnone willreturn
declare double @llvm.nvvm.rsqrt.approx.d(double) #8

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.trunc.d(double) #6

; Function Attrs: noinline nounwind
define dso_local float @__nv_exp2f(float %x) local_unnamed_addr #9 {
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  br i1 %.not, label %4, label %2

2:                                                ; preds = %0
  %3 = call float @llvm.nvvm.ex2.approx.ftz.f(float %x) #139
  br label %__exp2f.exit

4:                                                ; preds = %0
  %5 = call float @llvm.nvvm.ex2.approx.f(float %x) #139
  br label %__exp2f.exit

__exp2f.exit:                                     ; preds = %4, %2
  %.0 = phi float [ %3, %2 ], [ %5, %4 ]
  ret float %.0
}

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.ex2.approx.ftz.f(float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.ex2.approx.f(float) #8

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.trunc.f(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.saturate.ftz.f(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.saturate.f(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fma.rn.ftz.f(float, float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fma.rn.f(float, float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fma.rz.ftz.f(float, float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fma.rz.f(float, float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fma.rm.ftz.f(float, float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.fma.rm.f(float, float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.fma.rn.d(double, double, double) #6

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.div.approx.ftz.f(float, float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.div.approx.f(float, float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.sqrt.rn.ftz.f(float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.sqrt.rn.f(float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare double @llvm.nvvm.sqrt.rn.d(double) #8

; Function Attrs: noinline nounwind
define dso_local float @__nv_sqrtf(float %x) local_unnamed_addr #10 {
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  br i1 %.not, label %8, label %2

2:                                                ; preds = %0
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %3, 0
  br i1 %.not2, label %6, label %4

4:                                                ; preds = %2
  %5 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %x) #139
  br label %__nvvm_sqrt_f.exit

6:                                                ; preds = %2
  %7 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %x) #139
  br label %__nvvm_sqrt_f.exit

8:                                                ; preds = %0
  %9 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %9, 0
  br i1 %.not1, label %12, label %10

10:                                               ; preds = %8
  %11 = call float @llvm.sqrt.f32(float %x)
  br label %__nvvm_sqrt_f.exit

12:                                               ; preds = %8
  %13 = call float @llvm.nvvm.sqrt.approx.f(float %x) #139
  br label %__nvvm_sqrt_f.exit

__nvvm_sqrt_f.exit:                               ; preds = %12, %10, %6, %4
  %.0 = phi float [ %5, %4 ], [ %7, %6 ], [ %11, %10 ], [ %13, %12 ]
  ret float %.0
}

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.sqrt.approx.ftz.f(float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.sqrt.approx.f(float) #8

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_sqrt(double %x) local_unnamed_addr #11 {
  %1 = call double @llvm.sqrt.f64(double %x) #140
  ret double %1
}

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.add.rn.d(double, double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.add.rz.d(double, double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.mul.rn.d(double, double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.add.rn.ftz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.add.rn.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.add.rz.ftz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.add.rz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.mul.rn.ftz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.mul.rn.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.mul.rz.ftz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.mul.rz.f(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.nvvm.d2i.rn(double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.nvvm.f2i.rn.ftz(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.nvvm.f2i.rn(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.nvvm.f2i.rz(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.lohi.i2d(i32, i32) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.nvvm.d2i.lo(double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.nvvm.d2i.hi(double) #6

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.lg2.approx.ftz.f(float) #8

; Function Attrs: nofree nosync nounwind readnone willreturn
declare float @llvm.nvvm.lg2.approx.f(float) #8

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.nvvm.round.f(float) #6

; Function Attrs: noinline nounwind
define dso_local i32 @__nv_isinff(float %a) local_unnamed_addr #12 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = fcmp oeq float %.0, 0x7FF0000000000000
  %4 = zext i1 %3 to i32
  ret i32 %4
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_sinf(float %a) local_unnamed_addr #13 {
__nv_float2int_rn.exit.i.i:
  %result.i.i.i = alloca [7 x i32], align 4
  %0 = fmul float %a, 0x3FE45F3060000000
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call i32 @llvm.nvvm.f2i.rn.ftz(float %0) #139
  %3 = call i32 @llvm.nvvm.f2i.rn(float %0) #139
  %.01 = select i1 %.not, i32 %3, i32 %2
  %4 = sitofp i32 %.01 to float
  %5 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %5, 0
  %6 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBFF921FB40000000, float %a) #139
  %7 = call float @llvm.fma.f32(float %4, float noundef 0xBFF921FB40000000, float %a)
  %.02 = select i1 %.not1, float %7, float %6
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBE74442D00000000, float %.02) #139
  %10 = call float @llvm.fma.f32(float %4, float noundef 0xBE74442D00000000, float %.02)
  %.03 = select i1 %.not2, float %10, float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBCF84698A0000000, float %.03) #139
  %13 = call float @llvm.fma.f32(float %4, float noundef 0xBCF84698A0000000, float %.03)
  %.04 = select i1 %.not3, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %16 = call float @llvm.fabs.f32(float %a)
  %.06 = select i1 %.not4, float %16, float %15
  %17 = fcmp ult float %.06, 1.056150e+05
  br i1 %17, label %__internal_trig_reduction_kernel.exit.i, label %__nv_isinff.exit.i.i

__nv_isinff.exit.i.i:                             ; preds = %__nv_float2int_rn.exit.i.i
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %18, 0
  %.07 = select i1 %.not17, float %16, float %15
  %19 = fcmp oeq float %.07, 0x7FF0000000000000
  br i1 %19, label %__nv_fmul_rn.exit.i.i, label %23

__nv_fmul_rn.exit.i.i:                            ; preds = %__nv_isinff.exit.i.i
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.mul.rn.ftz.f(float %a, float noundef 0.000000e+00) #139
  %22 = fmul float %a, 0.000000e+00
  %.08 = select i1 %.not23, float %22, float %21
  br label %__internal_trig_reduction_kernel.exit.i

23:                                               ; preds = %__nv_isinff.exit.i.i
  %24 = bitcast float %a to i32
  %25 = lshr i32 %24, 23
  %26 = and i32 %25, 255
  %27 = add nsw i32 %26, -128
  %28 = shl i32 %24, 8
  %29 = or i32 %28, -2147483648
  %30 = lshr i32 %27, 5
  br label %31

31:                                               ; preds = %23, %31
  %iq.i.i.i.025 = phi i32 [ 0, %23 ], [ %39, %31 ]
  %hi.i.i.i.024 = phi i32 [ 0, %23 ], [ %37, %31 ]
  %32 = zext i32 %iq.i.i.i.025 to i64
  %33 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %32
  %34 = load i32, i32 addrspace(1)* %33, align 4
  %35 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %34, i32 %29, i32 %hi.i.i.i.024) #141, !srcloc !67
  %36 = extractvalue { i32, i32 } %35, 0
  %37 = extractvalue { i32, i32 } %35, 1
  %38 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %32
  store i32 %36, i32* %38, align 4, !noalias !68
  %39 = add nuw nsw i32 %iq.i.i.i.025, 1
  %exitcond.not = icmp eq i32 %39, 6
  br i1 %exitcond.not, label %40, label %31, !llvm.loop !71

40:                                               ; preds = %31
  %41 = and i32 %24, -2147483648
  %42 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 6
  store i32 %37, i32* %42, align 4, !noalias !68
  %43 = and i32 %25, 31
  %44 = sub nsw i32 6, %30
  %45 = sext i32 %44 to i64
  %46 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %45
  %47 = load i32, i32* %46, align 4
  %48 = sub nsw i32 5, %30
  %49 = sext i32 %48 to i64
  %50 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %49
  %51 = load i32, i32* %50, align 4
  %.not18 = icmp eq i32 %43, 0
  br i1 %.not18, label %__internal_trig_reduction_slowpath.exit.i.i, label %52

52:                                               ; preds = %40
  %53 = sub nsw i32 4, %30
  %54 = sub nuw nsw i32 32, %43
  %55 = shl i32 %47, %43
  %56 = lshr i32 %51, %54
  %57 = add i32 %56, %55
  %58 = shl i32 %51, %43
  %59 = sext i32 %53 to i64
  %60 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %59
  %61 = load i32, i32* %60, align 4
  %62 = lshr i32 %61, %54
  %63 = add i32 %62, %58
  br label %__internal_trig_reduction_slowpath.exit.i.i

__internal_trig_reduction_slowpath.exit.i.i:      ; preds = %52, %40
  %hi.i.i.i.1 = phi i32 [ %57, %52 ], [ %47, %40 ]
  %lo.i.i.i.0 = phi i32 [ %63, %52 ], [ %51, %40 ]
  %64 = lshr i32 %hi.i.i.i.1, 30
  %65 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.1, i32 %lo.i.i.i.0, i32 noundef 2)
  %66 = shl i32 %lo.i.i.i.0, 2
  %67 = lshr i32 %65, 31
  %68 = add nuw nsw i32 %67, %64
  %.not19 = icmp eq i32 %41, 0
  %69 = sub nsw i32 0, %68
  %spec.select = select i1 %.not19, i32 %68, i32 %69
  %.not20 = icmp sgt i32 %65, -1
  %70 = xor i32 %41, -2147483648
  %s.i.i.i.0 = select i1 %.not20, i32 %41, i32 %70
  %not..not20 = xor i1 %.not20, true
  %71 = sext i1 %not..not20 to i32
  %hi.i.i.i.2 = xor i32 %65, %71
  %lo.i.i.i.1 = xor i32 %66, %71
  %72 = zext i32 %hi.i.i.i.2 to i64
  %73 = shl nuw i64 %72, 32
  %74 = zext i32 %lo.i.i.i.1 to i64
  %75 = or i64 %73, %74
  %76 = sitofp i64 %75 to double
  %77 = fmul double %76, 0x3BF921FB54442D19
  %78 = fptrunc double %77 to float
  %.not22 = icmp eq i32 %s.i.i.i.0, 0
  %79 = fneg float %78
  %r.i.i.i.0 = select i1 %.not22, float %78, float %79
  br label %__internal_trig_reduction_kernel.exit.i

__internal_trig_reduction_kernel.exit.i:          ; preds = %__nv_fmul_rn.exit.i.i, %__internal_trig_reduction_slowpath.exit.i.i, %__nv_float2int_rn.exit.i.i
  %i.i.1 = phi i32 [ %.01, %__nv_float2int_rn.exit.i.i ], [ 0, %__nv_fmul_rn.exit.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i ]
  %t.i.i.1 = phi float [ %.04, %__nv_float2int_rn.exit.i.i ], [ %.08, %__nv_fmul_rn.exit.i.i ], [ %r.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i ]
  %80 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %80, 0
  %81 = call float @llvm.nvvm.mul.rn.ftz.f(float %t.i.i.1, float %t.i.i.1) #139
  %82 = fmul float %t.i.i.1, %t.i.i.1
  %.011 = select i1 %.not5, float %82, float %81
  %83 = and i32 %i.i.1, 1
  %.not6 = icmp eq i32 %83, 0
  %84 = select i1 %.not6, float %t.i.i.1, float 1.000000e+00
  %85 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %85, 0
  %86 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %84, float noundef 0.000000e+00) #139
  %87 = call float @llvm.fma.f32(float %.011, float %84, float noundef 0.000000e+00)
  %.012 = select i1 %.not7, float %87, float %86
  br i1 %.not6, label %__internal_fmad.exit2.i.i, label %__internal_fmad.exit1.i.i

__internal_fmad.exit1.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i
  %88 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %88, 0
  %89 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.011, float noundef 0xBF56C0FDA0000000) #139
  %90 = call float @llvm.fma.f32(float %.011, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.013 = select i1 %.not16, float %90, float %89
  br label %__internal_fmad.exit2.i.i

__internal_fmad.exit2.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i, %__internal_fmad.exit1.i.i
  %91 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i ]
  %92 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i ]
  %93 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i ], [ %.013, %__internal_fmad.exit1.i.i ]
  %94 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %94, 0
  %95 = call float @llvm.nvvm.fma.rn.ftz.f(float %93, float %.011, float noundef %92) #139
  %96 = call float @llvm.fma.f32(float %93, float %.011, float noundef %92)
  %.010 = select i1 %.not10, float %96, float %95
  %97 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %97, 0
  %98 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %.011, float noundef %91) #139
  %99 = call float @llvm.fma.f32(float %.010, float %.011, float noundef %91)
  %.09 = select i1 %.not12, float %99, float %98
  %100 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %100, 0
  %101 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %.012, float %84) #139
  %102 = call float @llvm.fma.f32(float %.09, float %.012, float %84)
  %.05 = select i1 %.not13, float %102, float %101
  %103 = and i32 %i.i.1, 2
  %.not14 = icmp eq i32 %103, 0
  br i1 %.not14, label %__internal_accurate_sinf.exit, label %__internal_fmad.exit5.i.i

__internal_fmad.exit5.i.i:                        ; preds = %__internal_fmad.exit2.i.i
  %104 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %104, 0
  %105 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %106 = call float @llvm.fma.f32(float %.05, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0 = select i1 %.not15, float %106, float %105
  br label %__internal_accurate_sinf.exit

__internal_accurate_sinf.exit:                    ; preds = %__internal_fmad.exit5.i.i, %__internal_fmad.exit2.i.i
  %z.i.i.0 = phi float [ %.0, %__internal_fmad.exit5.i.i ], [ %.05, %__internal_fmad.exit2.i.i ]
  ret float %z.i.i.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_cosf(float %a) local_unnamed_addr #14 {
__nv_float2int_rn.exit.i.i:
  %result.i.i.i = alloca [7 x i32], align 4
  %0 = fmul float %a, 0x3FE45F3060000000
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call i32 @llvm.nvvm.f2i.rn.ftz(float %0) #139
  %3 = call i32 @llvm.nvvm.f2i.rn(float %0) #139
  %.01 = select i1 %.not, i32 %3, i32 %2
  %4 = sitofp i32 %.01 to float
  %5 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %5, 0
  %6 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBFF921FB40000000, float %a) #139
  %7 = call float @llvm.fma.f32(float %4, float noundef 0xBFF921FB40000000, float %a)
  %.02 = select i1 %.not1, float %7, float %6
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBE74442D00000000, float %.02) #139
  %10 = call float @llvm.fma.f32(float %4, float noundef 0xBE74442D00000000, float %.02)
  %.03 = select i1 %.not2, float %10, float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBCF84698A0000000, float %.03) #139
  %13 = call float @llvm.fma.f32(float %4, float noundef 0xBCF84698A0000000, float %.03)
  %.04 = select i1 %.not3, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %16 = call float @llvm.fabs.f32(float %a)
  %.06 = select i1 %.not4, float %16, float %15
  %17 = fcmp ult float %.06, 1.056150e+05
  br i1 %17, label %__internal_trig_reduction_kernel.exit.i, label %__nv_isinff.exit.i.i

__nv_isinff.exit.i.i:                             ; preds = %__nv_float2int_rn.exit.i.i
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %18, 0
  %.07 = select i1 %.not17, float %16, float %15
  %19 = fcmp oeq float %.07, 0x7FF0000000000000
  br i1 %19, label %__nv_fmul_rn.exit.i.i, label %23

__nv_fmul_rn.exit.i.i:                            ; preds = %__nv_isinff.exit.i.i
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.mul.rn.ftz.f(float %a, float noundef 0.000000e+00) #139
  %22 = fmul float %a, 0.000000e+00
  %.08 = select i1 %.not23, float %22, float %21
  br label %__internal_trig_reduction_kernel.exit.i

23:                                               ; preds = %__nv_isinff.exit.i.i
  %24 = bitcast float %a to i32
  %25 = lshr i32 %24, 23
  %26 = and i32 %25, 255
  %27 = add nsw i32 %26, -128
  %28 = shl i32 %24, 8
  %29 = or i32 %28, -2147483648
  %30 = lshr i32 %27, 5
  br label %31

31:                                               ; preds = %23, %31
  %iq.i.i.i.025 = phi i32 [ 0, %23 ], [ %39, %31 ]
  %hi.i.i.i.024 = phi i32 [ 0, %23 ], [ %37, %31 ]
  %32 = zext i32 %iq.i.i.i.025 to i64
  %33 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %32
  %34 = load i32, i32 addrspace(1)* %33, align 4
  %35 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %34, i32 %29, i32 %hi.i.i.i.024) #141, !srcloc !67
  %36 = extractvalue { i32, i32 } %35, 0
  %37 = extractvalue { i32, i32 } %35, 1
  %38 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %32
  store i32 %36, i32* %38, align 4, !noalias !68
  %39 = add nuw nsw i32 %iq.i.i.i.025, 1
  %exitcond.not = icmp eq i32 %39, 6
  br i1 %exitcond.not, label %40, label %31, !llvm.loop !71

40:                                               ; preds = %31
  %41 = and i32 %24, -2147483648
  %42 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 6
  store i32 %37, i32* %42, align 4, !noalias !68
  %43 = and i32 %25, 31
  %44 = sub nsw i32 6, %30
  %45 = sext i32 %44 to i64
  %46 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %45
  %47 = load i32, i32* %46, align 4
  %48 = sub nsw i32 5, %30
  %49 = sext i32 %48 to i64
  %50 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %49
  %51 = load i32, i32* %50, align 4
  %.not18 = icmp eq i32 %43, 0
  br i1 %.not18, label %__internal_trig_reduction_slowpath.exit.i.i, label %52

52:                                               ; preds = %40
  %53 = sub nsw i32 4, %30
  %54 = sub nuw nsw i32 32, %43
  %55 = shl i32 %47, %43
  %56 = lshr i32 %51, %54
  %57 = add i32 %56, %55
  %58 = shl i32 %51, %43
  %59 = sext i32 %53 to i64
  %60 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %59
  %61 = load i32, i32* %60, align 4
  %62 = lshr i32 %61, %54
  %63 = add i32 %62, %58
  br label %__internal_trig_reduction_slowpath.exit.i.i

__internal_trig_reduction_slowpath.exit.i.i:      ; preds = %52, %40
  %hi.i.i.i.1 = phi i32 [ %57, %52 ], [ %47, %40 ]
  %lo.i.i.i.0 = phi i32 [ %63, %52 ], [ %51, %40 ]
  %64 = lshr i32 %hi.i.i.i.1, 30
  %65 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.1, i32 %lo.i.i.i.0, i32 noundef 2)
  %66 = shl i32 %lo.i.i.i.0, 2
  %67 = lshr i32 %65, 31
  %68 = add nuw nsw i32 %67, %64
  %.not19 = icmp eq i32 %41, 0
  %69 = sub nsw i32 0, %68
  %spec.select = select i1 %.not19, i32 %68, i32 %69
  %.not20 = icmp sgt i32 %65, -1
  %70 = xor i32 %41, -2147483648
  %s.i.i.i.0 = select i1 %.not20, i32 %41, i32 %70
  %not..not20 = xor i1 %.not20, true
  %71 = sext i1 %not..not20 to i32
  %hi.i.i.i.2 = xor i32 %65, %71
  %lo.i.i.i.1 = xor i32 %66, %71
  %72 = zext i32 %hi.i.i.i.2 to i64
  %73 = shl nuw i64 %72, 32
  %74 = zext i32 %lo.i.i.i.1 to i64
  %75 = or i64 %73, %74
  %76 = sitofp i64 %75 to double
  %77 = fmul double %76, 0x3BF921FB54442D19
  %78 = fptrunc double %77 to float
  %.not22 = icmp eq i32 %s.i.i.i.0, 0
  %79 = fneg float %78
  %r.i.i.i.0 = select i1 %.not22, float %78, float %79
  br label %__internal_trig_reduction_kernel.exit.i

__internal_trig_reduction_kernel.exit.i:          ; preds = %__nv_fmul_rn.exit.i.i, %__internal_trig_reduction_slowpath.exit.i.i, %__nv_float2int_rn.exit.i.i
  %i.i.1 = phi i32 [ %.01, %__nv_float2int_rn.exit.i.i ], [ 0, %__nv_fmul_rn.exit.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i ]
  %t.i.i.1 = phi float [ %.04, %__nv_float2int_rn.exit.i.i ], [ %.08, %__nv_fmul_rn.exit.i.i ], [ %r.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i ]
  %80 = add i32 %i.i.1, 1
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %81, 0
  %82 = call float @llvm.nvvm.mul.rn.ftz.f(float %t.i.i.1, float %t.i.i.1) #139
  %83 = fmul float %t.i.i.1, %t.i.i.1
  %.011 = select i1 %.not5, float %83, float %82
  %84 = and i32 %80, 1
  %.not6 = icmp eq i32 %84, 0
  %85 = select i1 %.not6, float %t.i.i.1, float 1.000000e+00
  %86 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %86, 0
  %87 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %85, float noundef 0.000000e+00) #139
  %88 = call float @llvm.fma.f32(float %.011, float %85, float noundef 0.000000e+00)
  %.012 = select i1 %.not7, float %88, float %87
  br i1 %.not6, label %__internal_fmad.exit2.i.i, label %__internal_fmad.exit1.i.i

__internal_fmad.exit1.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i
  %89 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %89, 0
  %90 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.011, float noundef 0xBF56C0FDA0000000) #139
  %91 = call float @llvm.fma.f32(float %.011, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.013 = select i1 %.not16, float %91, float %90
  br label %__internal_fmad.exit2.i.i

__internal_fmad.exit2.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i, %__internal_fmad.exit1.i.i
  %92 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i ]
  %93 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i ]
  %94 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i ], [ %.013, %__internal_fmad.exit1.i.i ]
  %95 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %95, 0
  %96 = call float @llvm.nvvm.fma.rn.ftz.f(float %94, float %.011, float noundef %93) #139
  %97 = call float @llvm.fma.f32(float %94, float %.011, float noundef %93)
  %.010 = select i1 %.not10, float %97, float %96
  %98 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %98, 0
  %99 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %.011, float noundef %92) #139
  %100 = call float @llvm.fma.f32(float %.010, float %.011, float noundef %92)
  %.09 = select i1 %.not12, float %100, float %99
  %101 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %101, 0
  %102 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %.012, float %85) #139
  %103 = call float @llvm.fma.f32(float %.09, float %.012, float %85)
  %.05 = select i1 %.not13, float %103, float %102
  %104 = and i32 %80, 2
  %.not14 = icmp eq i32 %104, 0
  br i1 %.not14, label %__internal_accurate_cosf.exit, label %__internal_fmad.exit5.i.i

__internal_fmad.exit5.i.i:                        ; preds = %__internal_fmad.exit2.i.i
  %105 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %105, 0
  %106 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %107 = call float @llvm.fma.f32(float %.05, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0 = select i1 %.not15, float %107, float %106
  br label %__internal_accurate_cosf.exit

__internal_accurate_cosf.exit:                    ; preds = %__internal_fmad.exit5.i.i, %__internal_fmad.exit2.i.i
  %z.i.i.0 = phi float [ %.0, %__internal_fmad.exit5.i.i ], [ %.05, %__internal_fmad.exit2.i.i ]
  ret float %z.i.i.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_tanf(float %a) local_unnamed_addr #15 {
__nv_float2int_rn.exit.i:
  %result.i.i = alloca [7 x i32], align 4
  %0 = fmul float %a, 0x3FE45F3060000000
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call i32 @llvm.nvvm.f2i.rn.ftz(float %0) #139
  %3 = call i32 @llvm.nvvm.f2i.rn(float %0) #139
  %.01 = select i1 %.not, i32 %3, i32 %2
  %4 = sitofp i32 %.01 to float
  %5 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %5, 0
  %6 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBFF921FB40000000, float %a) #139
  %7 = call float @llvm.fma.f32(float %4, float noundef 0xBFF921FB40000000, float %a)
  %.03 = select i1 %.not1, float %7, float %6
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBE74442D00000000, float %.03) #139
  %10 = call float @llvm.fma.f32(float %4, float noundef 0xBE74442D00000000, float %.03)
  %.04 = select i1 %.not2, float %10, float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %4, float noundef 0xBCF84698A0000000, float %.04) #139
  %13 = call float @llvm.fma.f32(float %4, float noundef 0xBCF84698A0000000, float %.04)
  %.05 = select i1 %.not3, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %16 = call float @llvm.fabs.f32(float %a)
  %.06 = select i1 %.not4, float %16, float %15
  %17 = fcmp ult float %.06, 1.056150e+05
  br i1 %17, label %__internal_trig_reduction_kernel.exit, label %__nv_isinff.exit.i

__nv_isinff.exit.i:                               ; preds = %__nv_float2int_rn.exit.i
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %18, 0
  %.07 = select i1 %.not14, float %16, float %15
  %19 = fcmp oeq float %.07, 0x7FF0000000000000
  br i1 %19, label %__nv_fmul_rn.exit.i, label %23

__nv_fmul_rn.exit.i:                              ; preds = %__nv_isinff.exit.i
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.mul.rn.ftz.f(float %a, float noundef 0.000000e+00) #139
  %22 = fmul float %a, 0.000000e+00
  %.08 = select i1 %.not20, float %22, float %21
  br label %__internal_trig_reduction_kernel.exit

23:                                               ; preds = %__nv_isinff.exit.i
  %24 = bitcast float %a to i32
  %25 = lshr i32 %24, 23
  %26 = and i32 %25, 255
  %27 = add nsw i32 %26, -128
  %28 = shl i32 %24, 8
  %29 = or i32 %28, -2147483648
  %30 = lshr i32 %27, 5
  br label %31

31:                                               ; preds = %23, %31
  %iq.i.i.022 = phi i32 [ 0, %23 ], [ %39, %31 ]
  %hi.i.i.021 = phi i32 [ 0, %23 ], [ %37, %31 ]
  %32 = zext i32 %iq.i.i.022 to i64
  %33 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %32
  %34 = load i32, i32 addrspace(1)* %33, align 4
  %35 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %34, i32 %29, i32 %hi.i.i.021) #141, !srcloc !67
  %36 = extractvalue { i32, i32 } %35, 0
  %37 = extractvalue { i32, i32 } %35, 1
  %38 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i, i64 0, i64 %32
  store i32 %36, i32* %38, align 4, !noalias !68
  %39 = add nuw nsw i32 %iq.i.i.022, 1
  %exitcond.not = icmp eq i32 %39, 6
  br i1 %exitcond.not, label %40, label %31, !llvm.loop !71

40:                                               ; preds = %31
  %41 = and i32 %24, -2147483648
  %42 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i, i64 0, i64 6
  store i32 %37, i32* %42, align 4, !noalias !68
  %43 = and i32 %25, 31
  %44 = sub nsw i32 6, %30
  %45 = sext i32 %44 to i64
  %46 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i, i64 0, i64 %45
  %47 = load i32, i32* %46, align 4
  %48 = sub nsw i32 5, %30
  %49 = sext i32 %48 to i64
  %50 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i, i64 0, i64 %49
  %51 = load i32, i32* %50, align 4
  %.not15 = icmp eq i32 %43, 0
  br i1 %.not15, label %__internal_trig_reduction_slowpath.exit.i, label %52

52:                                               ; preds = %40
  %53 = sub nsw i32 4, %30
  %54 = sub nuw nsw i32 32, %43
  %55 = shl i32 %47, %43
  %56 = lshr i32 %51, %54
  %57 = add i32 %56, %55
  %58 = shl i32 %51, %43
  %59 = sext i32 %53 to i64
  %60 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i, i64 0, i64 %59
  %61 = load i32, i32* %60, align 4
  %62 = lshr i32 %61, %54
  %63 = add i32 %62, %58
  br label %__internal_trig_reduction_slowpath.exit.i

__internal_trig_reduction_slowpath.exit.i:        ; preds = %52, %40
  %hi.i.i.1 = phi i32 [ %57, %52 ], [ %47, %40 ]
  %lo.i.i.0 = phi i32 [ %63, %52 ], [ %51, %40 ]
  %64 = lshr i32 %hi.i.i.1, 30
  %65 = call i32 @llvm.fshl.i32(i32 %hi.i.i.1, i32 %lo.i.i.0, i32 noundef 2)
  %66 = shl i32 %lo.i.i.0, 2
  %67 = lshr i32 %65, 31
  %68 = add nuw nsw i32 %67, %64
  %.not16 = icmp eq i32 %41, 0
  %69 = sub nsw i32 0, %68
  %spec.select = select i1 %.not16, i32 %68, i32 %69
  %.not17 = icmp sgt i32 %65, -1
  %70 = xor i32 %41, -2147483648
  %s.i.i.0 = select i1 %.not17, i32 %41, i32 %70
  %not..not17 = xor i1 %.not17, true
  %71 = sext i1 %not..not17 to i32
  %hi.i.i.2 = xor i32 %65, %71
  %lo.i.i.1 = xor i32 %66, %71
  %72 = zext i32 %hi.i.i.2 to i64
  %73 = shl nuw i64 %72, 32
  %74 = zext i32 %lo.i.i.1 to i64
  %75 = or i64 %73, %74
  %76 = sitofp i64 %75 to double
  %77 = fmul double %76, 0x3BF921FB54442D19
  %78 = fptrunc double %77 to float
  %.not19 = icmp eq i32 %s.i.i.0, 0
  %79 = fneg float %78
  %r.i.i.0 = select i1 %.not19, float %78, float %79
  br label %__internal_trig_reduction_kernel.exit

__internal_trig_reduction_kernel.exit:            ; preds = %__nv_fmul_rn.exit.i, %__internal_trig_reduction_slowpath.exit.i, %__nv_float2int_rn.exit.i
  %i.1 = phi i32 [ %.01, %__nv_float2int_rn.exit.i ], [ 0, %__nv_fmul_rn.exit.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i ]
  %t.i.1 = phi float [ %.05, %__nv_float2int_rn.exit.i ], [ %.08, %__nv_fmul_rn.exit.i ], [ %r.i.i.0, %__internal_trig_reduction_slowpath.exit.i ]
  %80 = fmul float %t.i.1, %t.i.1
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %81, 0
  %82 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F83200000000000, float %80, float noundef 0x3F6AC00000000000) #139
  %83 = call float @llvm.fma.f32(float %80, float noundef 0x3F83200000000000, float noundef 0x3F6AC00000000000)
  %.012 = select i1 %.not5, float %83, float %82
  %84 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %84, 0
  %85 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %80, float noundef 0x3F98E00000000000) #139
  %86 = call float @llvm.fma.f32(float %.012, float %80, float noundef 0x3F98E00000000000)
  %.013 = select i1 %.not6, float %86, float %85
  %87 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %87, 0
  %88 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %80, float noundef 0x3FAB600000000000) #139
  %89 = call float @llvm.fma.f32(float %.013, float %80, float noundef 0x3FAB600000000000)
  %.014 = select i1 %.not7, float %89, float %88
  %90 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %90, 0
  %91 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %80, float noundef 0x3FC1128700000000) #139
  %92 = call float @llvm.fma.f32(float %.014, float %80, float noundef 0x3FC1128700000000)
  %.011 = select i1 %.not8, float %92, float %91
  %93 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %93, 0
  %94 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %80, float noundef 0x3FD5555100000000) #139
  %95 = call float @llvm.fma.f32(float %.011, float %80, float noundef 0x3FD5555100000000)
  %.010 = select i1 %.not9, float %95, float %94
  %96 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %96, 0
  %97 = call float @llvm.nvvm.mul.rn.ftz.f(float %80, float %t.i.1) #139
  %98 = fmul float %t.i.1, %80
  %.09 = select i1 %.not10, float %98, float %97
  %99 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %99, 0
  %100 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %.09, float %t.i.1) #139
  %101 = call float @llvm.fma.f32(float %.010, float %.09, float %t.i.1)
  %.02 = select i1 %.not11, float %101, float %100
  %102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %102, 0
  %103 = call float @llvm.nvvm.fabs.ftz.f(float %t.i.1) #139
  %104 = call float @llvm.fabs.f32(float %t.i.1)
  %.0 = select i1 %.not12, float %104, float %103
  %105 = fcmp oeq float %.0, 0x3F40168780000000
  %t.i1.0 = select i1 %105, float %t.i.1, float %.02
  %106 = and i32 %i.1, 1
  %.not13 = icmp eq i32 %106, 0
  br i1 %.not13, label %110, label %107

107:                                              ; preds = %__internal_trig_reduction_kernel.exit
  %108 = fneg float %t.i1.0
  %109 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %108) #141, !srcloc !73
  br label %110

110:                                              ; preds = %107, %__internal_trig_reduction_kernel.exit
  %z.0 = phi float [ %109, %107 ], [ %t.i1.0, %__internal_trig_reduction_kernel.exit ]
  ret float %z.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_log2f(float %a) local_unnamed_addr #16 {
__nv_fmaf_rn.exit10.i:
  %0 = fcmp olt float %a, 0x3810000000000000
  %1 = fmul float %a, 0x4160000000000000
  %.02 = select i1 %0, float %1, float %a
  %i.i.0 = select i1 %0, float -2.300000e+01, float 0.000000e+00
  %2 = bitcast float %.02 to i32
  %3 = add i32 %2, -1060439283
  %4 = and i32 %3, -8388608
  %5 = sub i32 %2, %4
  %6 = bitcast i32 %5 to float
  %7 = sitofp i32 %4 to float
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %7, float noundef 0x3E80000000000000, float noundef %i.i.0) #139
  %10 = call float @llvm.fma.f32(float %7, float noundef 0x3E80000000000000, float noundef %i.i.0)
  %.08 = select i1 %.not, float %10, float %9
  %11 = fadd float %6, -1.000000e+00
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FB8D64FE0000000, float %11, float noundef 0xBFC58FE600000000) #139
  %14 = call float @llvm.fma.f32(float %11, float noundef 0x3FB8D64FE0000000, float noundef 0xBFC58FE600000000)
  %.010 = select i1 %.not1, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %11, float noundef 0x3FC5F9E540000000) #139
  %17 = call float @llvm.fma.f32(float %.010, float %11, float noundef 0x3FC5F9E540000000)
  %.011 = select i1 %.not2, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %11, float noundef 0xBFC6E9C860000000) #139
  %20 = call float @llvm.fma.f32(float %.011, float %11, float noundef 0xBFC6E9C860000000)
  %.012 = select i1 %.not3, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %11, float noundef 0x3FCA417E80000000) #139
  %23 = call float @llvm.fma.f32(float %.012, float %11, float noundef 0x3FCA417E80000000)
  %.09 = select i1 %.not4, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %11, float noundef 0xBFCEC79160000000) #139
  %26 = call float @llvm.fma.f32(float %.09, float %11, float noundef 0xBFCEC79160000000)
  %.05 = select i1 %.not5, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %11, float noundef 0x3FD277F320000000) #139
  %29 = call float @llvm.fma.f32(float %.05, float %11, float noundef 0x3FD277F320000000)
  %.01 = select i1 %.not6, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %11, float noundef 0xBFD7154920000000) #139
  %32 = call float @llvm.fma.f32(float %.01, float %11, float noundef 0xBFD7154920000000)
  %.0 = select i1 %.not7, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0, float %11, float noundef 0x3FDEC70940000000) #139
  %35 = call float @llvm.fma.f32(float %.0, float %11, float noundef 0x3FDEC70940000000)
  %.07 = select i1 %.not8, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %11, float noundef 0xBFE7154760000000) #139
  %38 = call float @llvm.fma.f32(float %.07, float %11, float noundef 0xBFE7154760000000)
  %.06 = select i1 %.not9, float %38, float %37
  %39 = fmul float %11, %.06
  %40 = fmul float %11, %39
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %41, 0
  %42 = call float @llvm.nvvm.fma.rn.ftz.f(float %11, float noundef 0x3FF7154760000000, float %40) #139
  %43 = call float @llvm.fma.f32(float %11, float noundef 0x3FF7154760000000, float %40)
  %.04 = select i1 %.not10, float %43, float %42
  %44 = fadd float %.08, %.04
  %45 = icmp ugt i32 %2, 2139095039
  br i1 %45, label %__nv_fmaf_rn.exit.i, label %__internal_accurate_log2f.exit

__nv_fmaf_rn.exit.i:                              ; preds = %__nv_fmaf_rn.exit10.i
  %46 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %46, 0
  %47 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %48 = call float @llvm.fma.f32(float %.02, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.03 = select i1 %.not11, float %48, float %47
  br label %__internal_accurate_log2f.exit

__internal_accurate_log2f.exit:                   ; preds = %__nv_fmaf_rn.exit.i, %__nv_fmaf_rn.exit10.i
  %r.i.0 = phi float [ %.03, %__nv_fmaf_rn.exit.i ], [ %44, %__nv_fmaf_rn.exit10.i ]
  %49 = fcmp oeq float %.02, 0.000000e+00
  %r.i.1 = select i1 %49, float 0xFFF0000000000000, float %r.i.0
  ret float %r.i.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_expf(float %a) local_unnamed_addr #17 {
__nv_fmaf_rn.exit.i.i:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float noundef 0x3F777313A0000000, float noundef 5.000000e-01) #139
  %2 = call float @llvm.fma.f32(float %a, float noundef 0x3F777313A0000000, float noundef 5.000000e-01)
  %.02 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.saturate.ftz.f(float %.02) #139
  %5 = call float @llvm.nvvm.saturate.f(float %.02) #139
  %.03 = select i1 %.not1, float %5, float %4
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fma.rm.ftz.f(float %.03, float noundef 2.520000e+02, float noundef 0x4168000020000000) #139
  %8 = call float @llvm.nvvm.fma.rm.f(float %.03, float noundef 2.520000e+02, float noundef 0x4168000020000000) #139
  %.04 = select i1 %.not2, float %8, float %7
  %9 = fadd float %.04, 0xC168000FE0000000
  %10 = fneg float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float noundef 0x3FF7154760000000, float %10) #139
  %13 = call float @llvm.fma.f32(float %a, float noundef 0x3FF7154760000000, float %10)
  %.0 = select i1 %.not3, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float noundef 0x3E54AE0C00000000, float %.0) #139
  %16 = call float @llvm.fma.f32(float %a, float noundef 0x3E54AE0C00000000, float %.0)
  %.01 = select i1 %.not4, float %16, float %15
  %17 = bitcast float %.04 to i32
  %18 = shl i32 %17, 23
  %19 = bitcast i32 %18 to float
  %20 = call float @llvm.nvvm.ex2.approx.ftz.f(float %.01) #139
  %21 = fmul float %20, %19
  ret float %21
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_exp10f(float %a) local_unnamed_addr #18 {
__nv_fmaf_rn.exit.i.i:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float noundef 0x3F8AFF4C40000000, float noundef 5.000000e-01) #139
  %2 = call float @llvm.fma.f32(float %a, float noundef 0x3F8AFF4C40000000, float noundef 5.000000e-01)
  %.02 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.saturate.ftz.f(float %.02) #139
  %5 = call float @llvm.nvvm.saturate.f(float %.02) #139
  %.03 = select i1 %.not1, float %5, float %4
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fma.rm.ftz.f(float %.03, float noundef 2.520000e+02, float noundef 0x4168000020000000) #139
  %8 = call float @llvm.nvvm.fma.rm.f(float %.03, float noundef 2.520000e+02, float noundef 0x4168000020000000) #139
  %.04 = select i1 %.not2, float %8, float %7
  %9 = fadd float %.04, 0xC168000FE0000000
  %10 = fneg float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float noundef 0x400A934F00000000, float %10) #139
  %13 = call float @llvm.fma.f32(float %a, float noundef 0x400A934F00000000, float %10)
  %.0 = select i1 %.not3, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float noundef 0x3E72F346E0000000, float %.0) #139
  %16 = call float @llvm.fma.f32(float %a, float noundef 0x3E72F346E0000000, float %.0)
  %.01 = select i1 %.not4, float %16, float %15
  %17 = bitcast float %.04 to i32
  %18 = shl i32 %17, 23
  %19 = bitcast i32 %18 to float
  %20 = call float @llvm.nvvm.ex2.approx.ftz.f(float %.01) #139
  %21 = fmul float %20, %19
  ret float %21
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_coshf(float %a) local_unnamed_addr #19 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.mul.rn.ftz.f(float %.01, float noundef 0x3FF7154760000000) #139
  %5 = fmul float %.01, 0x3FF7154760000000
  %.06 = select i1 %.not1, float %5, float %4
  %6 = call float @llvm.trunc.f32(float %.06)
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %7, 0
  %8 = call float @llvm.nvvm.fabs.ftz.f(float %6) #139
  %9 = call float @llvm.fabs.f32(float %6)
  %.04 = select i1 %.not2, float %9, float %8
  %10 = fcmp ogt float %.04, 1.260000e+02
  %11 = bitcast float %6 to i32
  %12 = and i32 %11, -2147483648
  %13 = or i32 %12, 1123811328
  %14 = bitcast i32 %13 to float
  %j.i.0 = select i1 %10, float %14, float %6
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.0, float noundef 0xBFE62E4300000000, float %.01) #139
  %17 = call float @llvm.fma.f32(float %j.i.0, float noundef 0xBFE62E4300000000, float %.01)
  %.03 = select i1 %.not3, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.0, float noundef 0x3E205C6100000000, float %.03) #139
  %20 = call float @llvm.fma.f32(float %j.i.0, float noundef 0x3E205C6100000000, float %.03)
  %.02 = select i1 %.not4, float %20, float %19
  %21 = fmul float %.02, 0x3FF7154760000000
  %22 = fadd float %j.i.0, 0x4168000FA0000000
  %23 = bitcast float %22 to i32
  %24 = shl i32 %23, 23
  %25 = bitcast i32 %24 to float
  %26 = call float @llvm.nvvm.ex2.approx.ftz.f(float %21) #139
  %27 = fmul float %26, %25
  %28 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %28, 0
  br i1 %.not5, label %31, label %29

29:                                               ; preds = %__nv_fabsf.exit
  %30 = call float @llvm.nvvm.div.approx.ftz.f(float noundef 1.250000e-01, float %27) #139
  br label %__nv_fast_fdividef.exit

31:                                               ; preds = %__nv_fabsf.exit
  %32 = call float @llvm.nvvm.div.approx.f(float noundef 1.250000e-01, float %27) #139
  br label %__nv_fast_fdividef.exit

__nv_fast_fdividef.exit:                          ; preds = %31, %29
  %.05 = phi float [ %30, %29 ], [ %32, %31 ]
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.000000e+00, float %27, float %.05) #139
  %35 = call float @llvm.fma.f32(float %27, float noundef 2.000000e+00, float %.05)
  %.0 = select i1 %.not6, float %35, float %34
  %36 = fcmp oge float %.01, 9.000000e+01
  %z.0 = select i1 %36, float 0x7FF0000000000000, float %.0
  ret float %z.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_sinhf(float %a) local_unnamed_addr #20 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp ult float %.01, 1.000000e+00
  br i1 %3, label %__internal_fmad.exit4, label %__nv_fmul_rn.exit.i

__nv_fmul_rn.exit.i:                              ; preds = %__nv_fabsf.exit
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %4, 0
  %5 = call float @llvm.nvvm.mul.rn.ftz.f(float %.01, float noundef 0x3FF7154760000000) #139
  %6 = fmul float %.01, 0x3FF7154760000000
  %.06 = select i1 %.not5, float %6, float %5
  %7 = call float @llvm.trunc.f32(float %.06)
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fabs.ftz.f(float %7) #139
  %10 = call float @llvm.fabs.f32(float %7)
  %.05 = select i1 %.not6, float %10, float %9
  %11 = fcmp ogt float %.05, 1.260000e+02
  %12 = bitcast float %7 to i32
  %13 = and i32 %12, -2147483648
  %14 = or i32 %13, 1123811328
  %15 = bitcast i32 %14 to float
  %j.i.0 = select i1 %11, float %15, float %7
  %16 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %16, 0
  %17 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.0, float noundef 0xBFE62E4300000000, float %.01) #139
  %18 = call float @llvm.fma.f32(float %j.i.0, float noundef 0xBFE62E4300000000, float %.01)
  %.04 = select i1 %.not7, float %18, float %17
  %19 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %19, 0
  %20 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.0, float noundef 0x3E205C6100000000, float %.04) #139
  %21 = call float @llvm.fma.f32(float %j.i.0, float noundef 0x3E205C6100000000, float %.04)
  %.02 = select i1 %.not8, float %21, float %20
  %22 = fmul float %.02, 0x3FF7154760000000
  %23 = fadd float %j.i.0, 0x4168000FA0000000
  %24 = bitcast float %23 to i32
  %25 = shl i32 %24, 23
  %26 = bitcast i32 %25 to float
  %27 = call float @llvm.nvvm.ex2.approx.ftz.f(float %22) #139
  %28 = fmul float %27, %26
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %29, 0
  br i1 %.not9, label %32, label %30

30:                                               ; preds = %__nv_fmul_rn.exit.i
  %31 = call float @llvm.nvvm.div.approx.ftz.f(float noundef 1.250000e-01, float %28) #139
  br label %__nv_fast_fdividef.exit

32:                                               ; preds = %__nv_fmul_rn.exit.i
  %33 = call float @llvm.nvvm.div.approx.f(float noundef 1.250000e-01, float %28) #139
  br label %__nv_fast_fdividef.exit

__nv_fast_fdividef.exit:                          ; preds = %32, %30
  %.07 = phi float [ %31, %30 ], [ %33, %32 ]
  %34 = fneg float %.07
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.000000e+00, float %28, float %34) #139
  %37 = call float @llvm.fma.f32(float %28, float noundef 2.000000e+00, float %34)
  %.08 = select i1 %.not10, float %37, float %36
  %38 = fcmp oge float %.01, 9.000000e+01
  %z.0 = select i1 %38, float 0x7FF0000000000000, float %.08
  %39 = bitcast float %z.0 to i32
  %40 = bitcast float %a to i32
  %41 = and i32 %40, -2147483648
  %42 = or i32 %41, %39
  %43 = bitcast i32 %42 to float
  br label %58

__internal_fmad.exit4:                            ; preds = %__nv_fabsf.exit
  %44 = fmul float %a, %a
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EC7A15B40000000, float %44, float noundef 0x3F29FFE920000000) #139
  %47 = call float @llvm.fma.f32(float %44, float noundef 0x3EC7A15B40000000, float noundef 0x3F29FFE920000000)
  %.09 = select i1 %.not1, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %44, float noundef 0x3F81111340000000) #139
  %50 = call float @llvm.fma.f32(float %.09, float %44, float noundef 0x3F81111340000000)
  %.010 = select i1 %.not2, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %44, float noundef 0x3FC5555560000000) #139
  %53 = call float @llvm.fma.f32(float %.010, float %44, float noundef 0x3FC5555560000000)
  %.03 = select i1 %.not3, float %53, float %52
  %54 = fmul float %44, %.03
  %55 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %55, 0
  %56 = call float @llvm.nvvm.fma.rn.ftz.f(float %54, float %a, float %a) #139
  %57 = call float @llvm.fma.f32(float %54, float %a, float %a)
  %.0 = select i1 %.not4, float %57, float %56
  br label %58

58:                                               ; preds = %__internal_fmad.exit4, %__nv_fast_fdividef.exit
  %z.1 = phi float [ %43, %__nv_fast_fdividef.exit ], [ %.0, %__internal_fmad.exit4 ]
  ret float %z.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_tanhf(float %a) local_unnamed_addr #21 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp ult float %.01, 0x3FE3333340000000
  br i1 %3, label %__internal_fmad.exit3, label %__internal_fmad.exit1

__internal_fmad.exit1:                            ; preds = %__nv_fabsf.exit
  %4 = fmul float %.01, 0x4007154760000000
  %5 = call float @llvm.nvvm.ex2.approx.ftz.f(float %4) #139
  %6 = fadd float %5, 1.000000e+00
  %7 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %6) #141, !srcloc !73
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %7, float noundef -2.000000e+00, float noundef 1.000000e+00) #139
  %10 = call float @llvm.fma.f32(float %7, float noundef -2.000000e+00, float noundef 1.000000e+00)
  %.03 = select i1 %.not6, float %10, float %9
  %11 = fcmp oge float %.01, 0x4022059680000000
  %s.0 = select i1 %11, float 1.000000e+00, float %.03
  %12 = bitcast float %s.0 to i32
  %13 = bitcast float %a to i32
  %14 = and i32 %13, -2147483648
  %15 = or i32 %14, %12
  %16 = bitcast i32 %15 to float
  br label %33

__internal_fmad.exit3:                            ; preds = %__nv_fabsf.exit
  %17 = fmul float %a, %a
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F901E1040000000, float %17, float noundef 0xBFAAC795C0000000) #139
  %20 = call float @llvm.fma.f32(float %17, float noundef 0x3F901E1040000000, float noundef 0xBFAAC795C0000000)
  %.06 = select i1 %.not1, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %17, float noundef 0x3FC10B2820000000) #139
  %23 = call float @llvm.fma.f32(float %.06, float %17, float noundef 0x3FC10B2820000000)
  %.05 = select i1 %.not2, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %17, float noundef 0xBFD5553DA0000000) #139
  %26 = call float @llvm.fma.f32(float %.05, float %17, float noundef 0xBFD5553DA0000000)
  %.0 = select i1 %.not3, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0, float %17, float noundef 0.000000e+00) #139
  %29 = call float @llvm.fma.f32(float %.0, float %17, float noundef 0.000000e+00)
  %.04 = select i1 %.not4, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %a, float %a) #139
  %32 = call float @llvm.fma.f32(float %.04, float %a, float %a)
  %.02 = select i1 %.not5, float %32, float %31
  br label %33

33:                                               ; preds = %__internal_fmad.exit3, %__internal_fmad.exit1
  %s.1 = phi float [ %16, %__internal_fmad.exit1 ], [ %.02, %__internal_fmad.exit3 ]
  ret float %s.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_atan2f(float %a, float %b) local_unnamed_addr #22 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %2 = call float @llvm.fabs.f32(float %b)
  %.02 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %5 = call float @llvm.fabs.f32(float %a)
  %.03 = select i1 %.not3, float %5, float %4
  %6 = fcmp oeq float %.02, 0.000000e+00
  %7 = fcmp oeq float %.03, 0.000000e+00
  %or.cond = select i1 %6, i1 %7, i1 false
  br i1 %or.cond, label %8, label %16

8:                                                ; preds = %__nv_fabsf.exit
  %9 = bitcast float %b to i32
  %10 = icmp slt i32 %9, 0
  %11 = select i1 %10, i32 1078530011, i32 0
  %12 = bitcast float %a to i32
  %13 = and i32 %12, -2147483648
  %14 = or i32 %11, %13
  %15 = bitcast i32 %14 to float
  br label %68

16:                                               ; preds = %__nv_fabsf.exit
  %17 = fcmp oeq float %.02, 0x7FF0000000000000
  %18 = fcmp oeq float %.03, 0x7FF0000000000000
  %or.cond2 = select i1 %17, i1 %18, i1 false
  br i1 %or.cond2, label %19, label %__nv_fmaxf.exit

19:                                               ; preds = %16
  %20 = bitcast float %b to i32
  %21 = icmp slt i32 %20, 0
  %22 = select i1 %21, i32 1075235812, i32 1061752795
  %23 = bitcast float %a to i32
  %24 = and i32 %23, -2147483648
  %25 = or i32 %22, %24
  %26 = bitcast i32 %25 to float
  br label %68

__nv_fmaxf.exit:                                  ; preds = %16
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fmax.ftz.f(float %.03, float %.02) #139
  %29 = call float @llvm.maxnum.f32(float %.03, float %.02)
  %.01 = select i1 %.not4, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fmin.ftz.f(float %.03, float %.02) #139
  %32 = call float @llvm.minnum.f32(float %.03, float %.02)
  %.0 = select i1 %.not5, float %32, float %31
  %33 = fdiv float %.0, %.01
  %34 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %34, 0
  %35 = call float @llvm.nvvm.mul.rn.ftz.f(float %33, float %33) #139
  %36 = fmul float %33, %33
  %.04 = select i1 %.not6, float %36, float %35
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float noundef 0xBFEA58FD40000000, float noundef 0xC016B31060000000) #139
  %39 = call float @llvm.fma.f32(float %.04, float noundef 0xBFEA58FD40000000, float noundef 0xC016B31060000000)
  %.05 = select i1 %.not7, float %39, float %38
  %40 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %40, 0
  %41 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %.04, float noundef 0xC01A4320E0000000) #139
  %42 = call float @llvm.fma.f32(float %.05, float %.04, float noundef 0xC01A4320E0000000)
  %.06 = select i1 %.not8, float %42, float %41
  %43 = fmul float %.04, %.06
  %44 = fmul float %33, %43
  %45 = fadd float %.04, 0x4026ABB800000000
  %46 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %46, 0
  %47 = call float @llvm.nvvm.fma.rn.ftz.f(float %45, float %.04, float noundef 0x403CD7AC00000000) #139
  %48 = call float @llvm.fma.f32(float %45, float %.04, float noundef 0x403CD7AC00000000)
  %.07 = select i1 %.not9, float %48, float %47
  %49 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %49, 0
  %50 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %.04, float noundef 0x4033B25900000000) #139
  %51 = call float @llvm.fma.f32(float %.07, float %.04, float noundef 0x4033B25900000000)
  %.08 = select i1 %.not10, float %51, float %50
  %52 = fdiv float 1.000000e+00, %.08
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fma.rn.ftz.f(float %44, float %52, float %33) #139
  %55 = call float @llvm.fma.f32(float %44, float %52, float %33)
  %.09 = select i1 %.not11, float %55, float %54
  %56 = fcmp ogt float %.03, %.02
  %57 = fsub float 0x3FF921FB60000000, %.09
  %t0.0 = select i1 %56, float %57, float %.09
  %58 = bitcast float %b to i32
  %59 = icmp slt i32 %58, 0
  %60 = fsub float 0x400921FB60000000, %t0.0
  %t0.1 = select i1 %59, float %60, float %t0.0
  %61 = bitcast float %t0.1 to i32
  %62 = bitcast float %a to i32
  %63 = and i32 %62, -2147483648
  %64 = or i32 %63, %61
  %65 = bitcast i32 %64 to float
  %66 = fadd float %.02, %.03
  %67 = fcmp ole float %66, 0x7FF0000000000000
  %t0.2 = select i1 %67, float %65, float %66
  br label %68

68:                                               ; preds = %19, %__nv_fmaxf.exit, %8
  %t0.4 = phi float [ %15, %8 ], [ %26, %19 ], [ %t0.2, %__nv_fmaxf.exit ]
  ret float %t0.4
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_atanf(float %a) local_unnamed_addr #23 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = fcmp ogt float %.0, 1.000000e+00
  br i1 %3, label %4, label %__internal_fmad.exit.i

4:                                                ; preds = %__nv_fabsf.exit
  %5 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0) #141, !srcloc !73
  br label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %4, %__nv_fabsf.exit
  %t1.0 = phi float [ %5, %4 ], [ %.0, %__nv_fabsf.exit ]
  %6 = fmul float %t1.0, %t1.0
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %7, 0
  %8 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F64121540000000, float %6, float noundef 0xBF8D7C29E0000000) #139
  %9 = call float @llvm.fma.f32(float %6, float noundef 0x3F64121540000000, float noundef 0xBF8D7C29E0000000)
  %.01 = select i1 %.not1, float %9, float %8
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %10, 0
  %11 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %6, float noundef 0x3FA4672FC0000000) #139
  %12 = call float @llvm.fma.f32(float %.01, float %6, float noundef 0x3FA4672FC0000000)
  %.02 = select i1 %.not2, float %12, float %11
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %13, 0
  %14 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %6, float noundef 0xBFB2914F40000000) #139
  %15 = call float @llvm.fma.f32(float %.02, float %6, float noundef 0xBFB2914F40000000)
  %.03 = select i1 %.not3, float %15, float %14
  %16 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %16, 0
  %17 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %6, float noundef 0x3FBAED6420000000) #139
  %18 = call float @llvm.fma.f32(float %.03, float %6, float noundef 0x3FBAED6420000000)
  %.04 = select i1 %.not4, float %18, float %17
  %19 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %19, 0
  %20 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %6, float noundef 0xBFC223D100000000) #139
  %21 = call float @llvm.fma.f32(float %.04, float %6, float noundef 0xBFC223D100000000)
  %.05 = select i1 %.not5, float %21, float %20
  %22 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %22, 0
  %23 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %6, float noundef 0x3FC995EC00000000) #139
  %24 = call float @llvm.fma.f32(float %.05, float %6, float noundef 0x3FC995EC00000000)
  %.06 = select i1 %.not6, float %24, float %23
  %25 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %25, 0
  %26 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %6, float noundef 0xBFD55544E0000000) #139
  %27 = call float @llvm.fma.f32(float %.06, float %6, float noundef 0xBFD55544E0000000)
  %.07 = select i1 %.not7, float %27, float %26
  %28 = fmul float %6, %.07
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %29, 0
  %30 = call float @llvm.nvvm.fma.rn.ftz.f(float %28, float %t1.0, float %t1.0) #139
  %31 = call float @llvm.fma.f32(float %28, float %t1.0, float %t1.0)
  %.08 = select i1 %.not8, float %31, float %30
  br i1 %3, label %__internal_fmad.exit, label %36

__internal_fmad.exit:                             ; preds = %__internal_fmad.exit.i
  %32 = fneg float %.08
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FEDDCB020000000, float noundef 0x3FFAEE9D60000000, float %32) #139
  %35 = call float @llvm.fma.f32(float noundef 0x3FEDDCB020000000, float noundef 0x3FFAEE9D60000000, float %32)
  %.09 = select i1 %.not9, float %35, float %34
  br label %36

36:                                               ; preds = %__internal_fmad.exit, %__internal_fmad.exit.i
  %t1.1 = phi float [ %.09, %__internal_fmad.exit ], [ %.08, %__internal_fmad.exit.i ]
  %37 = fcmp ugt float %.0, 0x7FF0000000000000
  br i1 %37, label %44, label %38

38:                                               ; preds = %36
  %39 = bitcast float %t1.1 to i32
  %40 = bitcast float %a to i32
  %41 = and i32 %40, -2147483648
  %42 = or i32 %41, %39
  %43 = bitcast i32 %42 to float
  br label %44

44:                                               ; preds = %38, %36
  %t1.2 = phi float [ %43, %38 ], [ %t1.1, %36 ]
  ret float %t1.2
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_asinf(float %a) local_unnamed_addr #24 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fneg float %.01
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %4, 0
  %5 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 5.000000e-01, float %3, float noundef 5.000000e-01) #139
  %6 = call float @llvm.fma.f32(float %3, float noundef 5.000000e-01, float noundef 5.000000e-01)
  %.02 = select i1 %.not1, float %6, float %5
  %7 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.02) #139
  %8 = fmul float %7, %.02
  %9 = fmul float %7, 5.000000e-01
  %10 = fneg float %8
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %10, float %9, float noundef 5.000000e-01) #139
  %13 = call float @llvm.fma.f32(float %10, float %9, float noundef 5.000000e-01)
  %.05 = select i1 %.not2, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %8, float %.05, float %8) #139
  %16 = call float @llvm.fma.f32(float %8, float %.05, float %8)
  %.06 = select i1 %.not3, float %16, float %15
  %17 = fcmp oeq float %.01, 1.000000e+00
  %18 = select i1 %17, float 0.000000e+00, float %.06
  %19 = fcmp ogt float %.01, 0x3FE1EB8520000000
  %20 = select i1 %19, float %18, float %.01
  %21 = fmul float %20, %20
  %22 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %22, 0
  %23 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FA9BA5EE0000000, float %21, float noundef 0x3F933952E0000000) #139
  %24 = call float @llvm.fma.f32(float %21, float noundef 0x3FA9BA5EE0000000, float noundef 0x3F933952E0000000)
  %.07 = select i1 %.not4, float %24, float %23
  %25 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %25, 0
  %26 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %21, float noundef 0x3FA7F21D00000000) #139
  %27 = call float @llvm.fma.f32(float %.07, float %21, float noundef 0x3FA7F21D00000000)
  %.08 = select i1 %.not5, float %27, float %26
  %28 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %28, 0
  %29 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %21, float noundef 0x3FB32799E0000000) #139
  %30 = call float @llvm.fma.f32(float %.08, float %21, float noundef 0x3FB32799E0000000)
  %.09 = select i1 %.not6, float %30, float %29
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %31, 0
  %32 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %21, float noundef 0x3FC5558080000000) #139
  %33 = call float @llvm.fma.f32(float %.09, float %21, float noundef 0x3FC5558080000000)
  %.04 = select i1 %.not7, float %33, float %32
  %34 = fmul float %21, %.04
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %34, float %20, float %20) #139
  %37 = call float @llvm.fma.f32(float %34, float %20, float %20)
  %.0 = select i1 %.not8, float %37, float %36
  %38 = fmul float %.0, -2.000000e+00
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FEDDCB020000000, float noundef 0x3FFAEE9D60000000, float %38) #139
  %41 = call float @llvm.fma.f32(float noundef 0x3FEDDCB020000000, float noundef 0x3FFAEE9D60000000, float %38)
  %.03 = select i1 %.not9, float %41, float %40
  %z.0 = select i1 %19, float %.03, float %.0
  %42 = fcmp ugt float %z.0, 0x7FF0000000000000
  br i1 %42, label %49, label %43

43:                                               ; preds = %__nv_fabsf.exit
  %44 = bitcast float %z.0 to i32
  %45 = bitcast float %a to i32
  %46 = and i32 %45, -2147483648
  %47 = or i32 %46, %44
  %48 = bitcast i32 %47 to float
  br label %49

49:                                               ; preds = %43, %__nv_fabsf.exit
  %z.1 = phi float [ %48, %43 ], [ %z.0, %__nv_fabsf.exit ]
  ret float %z.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_acosf(float %a) local_unnamed_addr #25 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fneg float %.01
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %4, 0
  %5 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 5.000000e-01, float %3, float noundef 5.000000e-01) #139
  %6 = call float @llvm.fma.f32(float %3, float noundef 5.000000e-01, float noundef 5.000000e-01)
  %.02 = select i1 %.not1, float %6, float %5
  %7 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.02) #139
  %8 = fmul float %7, %.02
  %9 = fmul float %7, 5.000000e-01
  %10 = fneg float %8
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %10, float %9, float noundef 5.000000e-01) #139
  %13 = call float @llvm.fma.f32(float %10, float %9, float noundef 5.000000e-01)
  %.05 = select i1 %.not2, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %8, float %.05, float %8) #139
  %16 = call float @llvm.fma.f32(float %8, float %.05, float %8)
  %.06 = select i1 %.not3, float %16, float %15
  %17 = fcmp oeq float %.01, 1.000000e+00
  %18 = select i1 %17, float 0.000000e+00, float %.06
  %19 = fcmp ogt float %.01, 0x3FE1EB8520000000
  %20 = select i1 %19, float %18, float %.01
  %21 = bitcast float %20 to i32
  %22 = bitcast float %a to i32
  %23 = and i32 %22, -2147483648
  %24 = or i32 %23, %21
  %25 = bitcast i32 %24 to float
  %26 = fmul float %25, %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FA21D9DE0000000, float %26, float noundef 0x3F91635760000000) #139
  %29 = call float @llvm.fma.f32(float %26, float noundef 0x3FA21D9DE0000000, float noundef 0x3F91635760000000)
  %.07 = select i1 %.not4, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %26, float noundef 0x3F9F805180000000) #139
  %32 = call float @llvm.fma.f32(float %.07, float %26, float noundef 0x3F9F805180000000)
  %.08 = select i1 %.not5, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %26, float noundef 0x3FA6E42720000000) #139
  %35 = call float @llvm.fma.f32(float %.08, float %26, float noundef 0x3FA6E42720000000)
  %.010 = select i1 %.not6, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %26, float noundef 0x3FB3327B60000000) #139
  %38 = call float @llvm.fma.f32(float %.010, float %26, float noundef 0x3FB3327B60000000)
  %.09 = select i1 %.not7, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %26, float noundef 0x3FC55558C0000000) #139
  %41 = call float @llvm.fma.f32(float %.09, float %26, float noundef 0x3FC55558C0000000)
  %.04 = select i1 %.not8, float %41, float %40
  %42 = fmul float %26, %.04
  %43 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %43, 0
  %44 = call float @llvm.nvvm.fma.rn.ftz.f(float %42, float %25, float %25) #139
  %45 = call float @llvm.fma.f32(float %42, float %25, float %25)
  %.0 = select i1 %.not9, float %45, float %44
  %46 = fneg float %.0
  %47 = select i1 %19, float %.0, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FEDDCB020000000, float noundef 0x3FFAEE9D60000000, float %47) #139
  %50 = call float @llvm.fma.f32(float noundef 0x3FEDDCB020000000, float noundef 0x3FFAEE9D60000000, float %47)
  %.03 = select i1 %.not10, float %50, float %49
  %51 = fcmp ogt float %a, 0x3FE1EB8520000000
  %52 = select i1 %51, float %.0, float %.03
  %53 = fmul float %52, 2.000000e+00
  %z.0 = select i1 %19, float %53, float %52
  ret float %z.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_logf(float %a) local_unnamed_addr #26 {
__nv_fmaf_rn.exit10.i:
  %0 = fcmp olt float %a, 0x3810000000000000
  %1 = fmul float %a, 0x4160000000000000
  %.02 = select i1 %0, float %1, float %a
  %i.i.0 = select i1 %0, float -2.300000e+01, float 0.000000e+00
  %2 = bitcast float %.02 to i32
  %3 = add i32 %2, -1059760811
  %4 = and i32 %3, -8388608
  %5 = sub i32 %2, %4
  %6 = bitcast i32 %5 to float
  %7 = sitofp i32 %4 to float
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %7, float noundef 0x3E80000000000000, float noundef %i.i.0) #139
  %10 = call float @llvm.fma.f32(float %7, float noundef 0x3E80000000000000, float noundef %i.i.0)
  %.08 = select i1 %.not, float %10, float %9
  %11 = fadd float %6, -1.000000e+00
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %11, float noundef 0x3FC2073EC0000000) #139
  %14 = call float @llvm.fma.f32(float %11, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.010 = select i1 %.not1, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %11, float noundef 0xBFBF19B980000000) #139
  %17 = call float @llvm.fma.f32(float %.010, float %11, float noundef 0xBFBF19B980000000)
  %.011 = select i1 %.not2, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %11, float noundef 0x3FC1E52AA0000000) #139
  %20 = call float @llvm.fma.f32(float %.011, float %11, float noundef 0x3FC1E52AA0000000)
  %.012 = select i1 %.not3, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %11, float noundef 0xBFC55B1720000000) #139
  %23 = call float @llvm.fma.f32(float %.012, float %11, float noundef 0xBFC55B1720000000)
  %.09 = select i1 %.not4, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %11, float noundef 0x3FC99DA160000000) #139
  %26 = call float @llvm.fma.f32(float %.09, float %11, float noundef 0x3FC99DA160000000)
  %.05 = select i1 %.not5, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %11, float noundef 0xBFCFFFE440000000) #139
  %29 = call float @llvm.fma.f32(float %.05, float %11, float noundef 0xBFCFFFE440000000)
  %.01 = select i1 %.not6, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %11, float noundef 0x3FD5554F00000000) #139
  %32 = call float @llvm.fma.f32(float %.01, float %11, float noundef 0x3FD5554F00000000)
  %.0 = select i1 %.not7, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0, float %11, float noundef -5.000000e-01) #139
  %35 = call float @llvm.fma.f32(float %.0, float %11, float noundef -5.000000e-01)
  %.07 = select i1 %.not8, float %35, float %34
  %36 = fmul float %11, %.07
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %36, float %11, float %11) #139
  %39 = call float @llvm.fma.f32(float %36, float %11, float %11)
  %.06 = select i1 %.not9, float %39, float %38
  %40 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %40, 0
  %41 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float noundef 0x3FE62E4300000000, float %.06) #139
  %42 = call float @llvm.fma.f32(float %.08, float noundef 0x3FE62E4300000000, float %.06)
  %.04 = select i1 %.not10, float %42, float %41
  %43 = icmp ugt i32 %2, 2139095039
  br i1 %43, label %__nv_fmaf_rn.exit.i, label %__internal_accurate_logf.exit

__nv_fmaf_rn.exit.i:                              ; preds = %__nv_fmaf_rn.exit10.i
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %46 = call float @llvm.fma.f32(float %.02, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.03 = select i1 %.not11, float %46, float %45
  br label %__internal_accurate_logf.exit

__internal_accurate_logf.exit:                    ; preds = %__nv_fmaf_rn.exit.i, %__nv_fmaf_rn.exit10.i
  %r.i.0 = phi float [ %.03, %__nv_fmaf_rn.exit.i ], [ %.04, %__nv_fmaf_rn.exit10.i ]
  %47 = fcmp oeq float %.02, 0.000000e+00
  %r.i.1 = select i1 %47, float 0xFFF0000000000000, float %r.i.0
  ret float %r.i.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_log10f(float %a) local_unnamed_addr #27 {
__nv_fmaf_rn.exit10.i:
  %0 = fcmp olt float %a, 0x3810000000000000
  %1 = fmul float %a, 0x4160000000000000
  %.02 = select i1 %0, float %1, float %a
  %i.i.0 = select i1 %0, float -2.300000e+01, float 0.000000e+00
  %2 = bitcast float %.02 to i32
  %3 = add i32 %2, -1059760811
  %4 = and i32 %3, -8388608
  %5 = sub i32 %2, %4
  %6 = bitcast i32 %5 to float
  %7 = sitofp i32 %4 to float
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %7, float noundef 0x3E80000000000000, float noundef %i.i.0) #139
  %10 = call float @llvm.fma.f32(float %7, float noundef 0x3E80000000000000, float noundef %i.i.0)
  %.08 = select i1 %.not, float %10, float %9
  %11 = fadd float %6, -1.000000e+00
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %11, float noundef 0x3FC2073EC0000000) #139
  %14 = call float @llvm.fma.f32(float %11, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.010 = select i1 %.not1, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %11, float noundef 0xBFBF19B980000000) #139
  %17 = call float @llvm.fma.f32(float %.010, float %11, float noundef 0xBFBF19B980000000)
  %.011 = select i1 %.not2, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %11, float noundef 0x3FC1E52AA0000000) #139
  %20 = call float @llvm.fma.f32(float %.011, float %11, float noundef 0x3FC1E52AA0000000)
  %.012 = select i1 %.not3, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %11, float noundef 0xBFC55B1720000000) #139
  %23 = call float @llvm.fma.f32(float %.012, float %11, float noundef 0xBFC55B1720000000)
  %.09 = select i1 %.not4, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %11, float noundef 0x3FC99DA160000000) #139
  %26 = call float @llvm.fma.f32(float %.09, float %11, float noundef 0x3FC99DA160000000)
  %.05 = select i1 %.not5, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %11, float noundef 0xBFCFFFE440000000) #139
  %29 = call float @llvm.fma.f32(float %.05, float %11, float noundef 0xBFCFFFE440000000)
  %.01 = select i1 %.not6, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %11, float noundef 0x3FD5554F00000000) #139
  %32 = call float @llvm.fma.f32(float %.01, float %11, float noundef 0x3FD5554F00000000)
  %.0 = select i1 %.not7, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0, float %11, float noundef -5.000000e-01) #139
  %35 = call float @llvm.fma.f32(float %.0, float %11, float noundef -5.000000e-01)
  %.07 = select i1 %.not8, float %35, float %34
  %36 = fmul float %11, %.07
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %36, float %11, float %11) #139
  %39 = call float @llvm.fma.f32(float %36, float %11, float %11)
  %.06 = select i1 %.not9, float %39, float %38
  %40 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %40, 0
  %41 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float noundef 0x3FE62E4300000000, float %.06) #139
  %42 = call float @llvm.fma.f32(float %.08, float noundef 0x3FE62E4300000000, float %.06)
  %.04 = select i1 %.not10, float %42, float %41
  %43 = icmp ugt i32 %2, 2139095039
  br i1 %43, label %__nv_fmaf_rn.exit.i, label %__internal_accurate_logf.exit

__nv_fmaf_rn.exit.i:                              ; preds = %__nv_fmaf_rn.exit10.i
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %46 = call float @llvm.fma.f32(float %.02, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.03 = select i1 %.not11, float %46, float %45
  br label %__internal_accurate_logf.exit

__internal_accurate_logf.exit:                    ; preds = %__nv_fmaf_rn.exit.i, %__nv_fmaf_rn.exit10.i
  %r.i.0 = phi float [ %.03, %__nv_fmaf_rn.exit.i ], [ %.04, %__nv_fmaf_rn.exit10.i ]
  %47 = fcmp oeq float %.02, 0.000000e+00
  %r.i.0.op = fmul float %r.i.0, 0x3FDBCB7B20000000
  %48 = select i1 %47, float 0xFFF0000000000000, float %r.i.0.op
  ret float %48
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_log1pf(float %a) local_unnamed_addr #28 {
__nv_fadd_rz.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.add.rz.ftz.f(float %a, float noundef 1.000000e+00) #139
  %2 = call float @llvm.nvvm.add.rz.f(float %a, float noundef 1.000000e+00) #139
  %.01 = select i1 %.not, float %2, float %1
  %3 = bitcast float %.01 to i32
  %4 = add i32 %3, -1061158912
  %5 = and i32 %4, -8388608
  %6 = bitcast float %a to i32
  %7 = sub i32 %6, %5
  %8 = bitcast i32 %7 to float
  %9 = sub i32 1082130432, %5
  %10 = bitcast i32 %9 to float
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.500000e-01, float %10, float noundef -1.000000e+00) #139
  %13 = call float @llvm.fma.f32(float %10, float noundef 2.500000e-01, float noundef -1.000000e+00)
  %.09 = select i1 %.not1, float %13, float %12
  %14 = fadd float %.09, %8
  %15 = sitofp i32 %5 to float
  %16 = fmul float %15, 0x3E80000000000000
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFA737EF00000000, float %14, float noundef 0x3FBB000240000000) #139
  %19 = call float @llvm.fma.f32(float %14, float noundef 0xBFA737EF00000000, float noundef 0x3FBB000240000000)
  %.012 = select i1 %.not2, float %19, float %18
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %14, float noundef 0xBFC0EF1C00000000) #139
  %22 = call float @llvm.fma.f32(float %.012, float %14, float noundef 0xBFC0EF1C00000000)
  %.010 = select i1 %.not3, float %22, float %21
  %23 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %23, 0
  %24 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %14, float noundef 0x3FC28C8EA0000000) #139
  %25 = call float @llvm.fma.f32(float %.010, float %14, float noundef 0x3FC28C8EA0000000)
  %.06 = select i1 %.not4, float %25, float %24
  %26 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %26, 0
  %27 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %14, float noundef 0xBFC54D1BA0000000) #139
  %28 = call float @llvm.fma.f32(float %.06, float %14, float noundef 0xBFC54D1BA0000000)
  %.02 = select i1 %.not5, float %28, float %27
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %29, 0
  %30 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %14, float noundef 0x3FC995F3C0000000) #139
  %31 = call float @llvm.fma.f32(float %.02, float %14, float noundef 0x3FC995F3C0000000)
  %.0 = select i1 %.not6, float %31, float %30
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0, float %14, float noundef 0xBFD0000840000000) #139
  %34 = call float @llvm.fma.f32(float %.0, float %14, float noundef 0xBFD0000840000000)
  %.011 = select i1 %.not7, float %34, float %33
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %14, float noundef 0x3FD5555CC0000000) #139
  %37 = call float @llvm.fma.f32(float %.011, float %14, float noundef 0x3FD5555CC0000000)
  %.08 = select i1 %.not8, float %37, float %36
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %38, 0
  %39 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %14, float noundef -5.000000e-01) #139
  %40 = call float @llvm.fma.f32(float %.08, float %14, float noundef -5.000000e-01)
  %.07 = select i1 %.not9, float %40, float %39
  %41 = fmul float %14, %.07
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %41, float %14, float %14) #139
  %44 = call float @llvm.fma.f32(float %41, float %14, float %14)
  %.05 = select i1 %.not10, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %16, float noundef 0x3FE62E4300000000, float %.05) #139
  %47 = call float @llvm.fma.f32(float %16, float noundef 0x3FE62E4300000000, float %.05)
  %.04 = select i1 %.not11, float %47, float %46
  %48 = icmp ugt i32 %6, 2139095039
  br i1 %48, label %49, label %56

49:                                               ; preds = %__nv_fadd_rz.exit
  %50 = icmp sgt i32 %6, -1082130432
  br i1 %50, label %__nv_fmaf_rn.exit, label %54

__nv_fmaf_rn.exit:                                ; preds = %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %53 = call float @llvm.fma.f32(float %a, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.03 = select i1 %.not12, float %53, float %52
  br label %54

54:                                               ; preds = %__nv_fmaf_rn.exit, %49
  %r.0 = phi float [ %.03, %__nv_fmaf_rn.exit ], [ %.04, %49 ]
  %55 = fcmp oeq float %a, 0.000000e+00
  %r.1 = select i1 %55, float -0.000000e+00, float %r.0
  br label %56

56:                                               ; preds = %54, %__nv_fadd_rz.exit
  %r.2 = phi float [ %r.1, %54 ], [ %.04, %__nv_fadd_rz.exit ]
  ret float %r.2
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_acoshf(float %a) local_unnamed_addr #29 {
  %1 = fadd float %a, -1.000000e+00
  %2 = bitcast float %1 to i32
  %3 = icmp ugt i32 %2, 1258291200
  br i1 %3, label %__nv_fmaf_rn.exit10.i13, label %__nv_fmul_rz.exit

__nv_fmaf_rn.exit10.i13:                          ; preds = %0
  %4 = fcmp olt float %1, 0x3810000000000000
  %5 = fmul float %1, 0x4160000000000000
  %.021 = select i1 %4, float %5, float %1
  %i.i11.0 = select i1 %4, float -2.300000e+01, float 0.000000e+00
  %6 = bitcast float %.021 to i32
  %7 = add i32 %6, -1059760811
  %8 = and i32 %7, -8388608
  %9 = sub i32 %6, %8
  %10 = bitcast i32 %9 to float
  %11 = sitofp i32 %8 to float
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float %11, float noundef 0x3E80000000000000, float noundef %i.i11.0) #139
  %14 = call float @llvm.fma.f32(float %11, float noundef 0x3E80000000000000, float noundef %i.i11.0)
  %.028 = select i1 %.not18, float %14, float %13
  %15 = fadd float %10, -1.000000e+00
  %16 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %16, 0
  %17 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %15, float noundef 0x3FC2073EC0000000) #139
  %18 = call float @llvm.fma.f32(float %15, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.027 = select i1 %.not19, float %18, float %17
  %19 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %19, 0
  %20 = call float @llvm.nvvm.fma.rn.ftz.f(float %.027, float %15, float noundef 0xBFBF19B980000000) #139
  %21 = call float @llvm.fma.f32(float %.027, float %15, float noundef 0xBFBF19B980000000)
  %.023 = select i1 %.not20, float %21, float %20
  %22 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %22, 0
  %23 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float %15, float noundef 0x3FC1E52AA0000000) #139
  %24 = call float @llvm.fma.f32(float %.023, float %15, float noundef 0x3FC1E52AA0000000)
  %.020 = select i1 %.not21, float %24, float %23
  %25 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %25, 0
  %26 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %15, float noundef 0xBFC55B1720000000) #139
  %27 = call float @llvm.fma.f32(float %.020, float %15, float noundef 0xBFC55B1720000000)
  %.018 = select i1 %.not22, float %27, float %26
  %28 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %28, 0
  %29 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %15, float noundef 0x3FC99DA160000000) #139
  %30 = call float @llvm.fma.f32(float %.018, float %15, float noundef 0x3FC99DA160000000)
  %.014 = select i1 %.not23, float %30, float %29
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %31, 0
  %32 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %15, float noundef 0xBFCFFFE440000000) #139
  %33 = call float @llvm.fma.f32(float %.014, float %15, float noundef 0xBFCFFFE440000000)
  %.010 = select i1 %.not24, float %33, float %32
  %34 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %34, 0
  %35 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %15, float noundef 0x3FD5554F00000000) #139
  %36 = call float @llvm.fma.f32(float %.010, float %15, float noundef 0x3FD5554F00000000)
  %.06 = select i1 %.not25, float %36, float %35
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not26 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %15, float noundef -5.000000e-01) #139
  %39 = call float @llvm.fma.f32(float %.06, float %15, float noundef -5.000000e-01)
  %.026 = select i1 %.not26, float %39, float %38
  %40 = fmul float %15, %.026
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %41, 0
  %42 = call float @llvm.nvvm.fma.rn.ftz.f(float %40, float %15, float %15) #139
  %43 = call float @llvm.fma.f32(float %40, float %15, float %15)
  %.025 = select i1 %.not27, float %43, float %42
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.028, float noundef 0x3FE62E4300000000, float %.025) #139
  %46 = call float @llvm.fma.f32(float %.028, float noundef 0x3FE62E4300000000, float %.025)
  %.024 = select i1 %.not28, float %46, float %45
  %47 = icmp ugt i32 %6, 2139095039
  br i1 %47, label %__nv_fmaf_rn.exit.i22, label %__internal_accurate_logf.exit

__nv_fmaf_rn.exit.i22:                            ; preds = %__nv_fmaf_rn.exit10.i13
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not29 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %50 = call float @llvm.fma.f32(float %.021, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.022 = select i1 %.not29, float %50, float %49
  br label %__internal_accurate_logf.exit

__internal_accurate_logf.exit:                    ; preds = %__nv_fmaf_rn.exit.i22, %__nv_fmaf_rn.exit10.i13
  %r.i10.0 = phi float [ %.022, %__nv_fmaf_rn.exit.i22 ], [ %.024, %__nv_fmaf_rn.exit10.i13 ]
  %51 = fcmp oeq float %.021, 0.000000e+00
  %r.i10.0.op = fadd float %r.i10.0, 0x3FE62E4300000000
  %52 = select i1 %51, float 0xFFF0000000000000, float %r.i10.0.op
  br label %__nv_log1pf.exit

__nv_fmul_rz.exit:                                ; preds = %0
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.mul.rz.ftz.f(float %a, float %1) #139
  %55 = call float @llvm.nvvm.mul.rz.f(float %a, float %1) #139
  %.03 = select i1 %.not, float %55, float %54
  %56 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %56, 0
  %57 = call float @llvm.nvvm.add.rn.ftz.f(float %.03, float %1) #139
  %58 = fadd float %1, %.03
  %.01 = select i1 %.not1, float %58, float %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %59, 0
  br i1 %.not2, label %66, label %60

60:                                               ; preds = %__nv_fmul_rz.exit
  %61 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %61, 0
  br i1 %.not17, label %64, label %62

62:                                               ; preds = %60
  %63 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %.01) #139
  br label %__nv_sqrtf.exit

64:                                               ; preds = %60
  %65 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %.01) #139
  br label %__nv_sqrtf.exit

66:                                               ; preds = %__nv_fmul_rz.exit
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %67, 0
  br i1 %.not3, label %70, label %68

68:                                               ; preds = %66
  %69 = call float @llvm.sqrt.f32(float %.01)
  br label %__nv_sqrtf.exit

70:                                               ; preds = %66
  %71 = call float @llvm.nvvm.sqrt.approx.f(float %.01) #139
  br label %__nv_sqrtf.exit

__nv_sqrtf.exit:                                  ; preds = %70, %68, %64, %62
  %.0 = phi float [ %63, %62 ], [ %65, %64 ], [ %69, %68 ], [ %71, %70 ]
  %72 = fadd float %1, %.0
  %73 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %73, 0
  %74 = call float @llvm.nvvm.add.rz.ftz.f(float %72, float noundef 1.000000e+00) #139
  %75 = call float @llvm.nvvm.add.rz.f(float %72, float noundef 1.000000e+00) #139
  %.02 = select i1 %.not4, float %75, float %74
  %76 = bitcast float %.02 to i32
  %77 = add i32 %76, -1061158912
  %78 = and i32 %77, -8388608
  %79 = bitcast float %72 to i32
  %80 = sub i32 %79, %78
  %81 = bitcast i32 %80 to float
  %82 = sub i32 1082130432, %78
  %83 = bitcast i32 %82 to float
  %84 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %84, 0
  %85 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.500000e-01, float %83, float noundef -1.000000e+00) #139
  %86 = call float @llvm.fma.f32(float %83, float noundef 2.500000e-01, float noundef -1.000000e+00)
  %.011 = select i1 %.not5, float %86, float %85
  %87 = fadd float %.011, %81
  %88 = sitofp i32 %78 to float
  %89 = fmul float %88, 0x3E80000000000000
  %90 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %90, 0
  %91 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFA737EF00000000, float %87, float noundef 0x3FBB000240000000) #139
  %92 = call float @llvm.fma.f32(float %87, float noundef 0xBFA737EF00000000, float noundef 0x3FBB000240000000)
  %.013 = select i1 %.not6, float %92, float %91
  %93 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %93, 0
  %94 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %87, float noundef 0xBFC0EF1C00000000) #139
  %95 = call float @llvm.fma.f32(float %.013, float %87, float noundef 0xBFC0EF1C00000000)
  %.015 = select i1 %.not7, float %95, float %94
  %96 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %96, 0
  %97 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %87, float noundef 0x3FC28C8EA0000000) #139
  %98 = call float @llvm.fma.f32(float %.015, float %87, float noundef 0x3FC28C8EA0000000)
  %.016 = select i1 %.not8, float %98, float %97
  %99 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %99, 0
  %100 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %87, float noundef 0xBFC54D1BA0000000) #139
  %101 = call float @llvm.fma.f32(float %.016, float %87, float noundef 0xBFC54D1BA0000000)
  %.017 = select i1 %.not9, float %101, float %100
  %102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %102, 0
  %103 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %87, float noundef 0x3FC995F3C0000000) #139
  %104 = call float @llvm.fma.f32(float %.017, float %87, float noundef 0x3FC995F3C0000000)
  %.019 = select i1 %.not10, float %104, float %103
  %105 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %105, 0
  %106 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %87, float noundef 0xBFD0000840000000) #139
  %107 = call float @llvm.fma.f32(float %.019, float %87, float noundef 0xBFD0000840000000)
  %.012 = select i1 %.not11, float %107, float %106
  %108 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %108, 0
  %109 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %87, float noundef 0x3FD5555CC0000000) #139
  %110 = call float @llvm.fma.f32(float %.012, float %87, float noundef 0x3FD5555CC0000000)
  %.09 = select i1 %.not12, float %110, float %109
  %111 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %111, 0
  %112 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %87, float noundef -5.000000e-01) #139
  %113 = call float @llvm.fma.f32(float %.09, float %87, float noundef -5.000000e-01)
  %.08 = select i1 %.not13, float %113, float %112
  %114 = fmul float %87, %.08
  %115 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %115, 0
  %116 = call float @llvm.nvvm.fma.rn.ftz.f(float %114, float %87, float %87) #139
  %117 = call float @llvm.fma.f32(float %114, float %87, float %87)
  %.07 = select i1 %.not14, float %117, float %116
  %118 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %118, 0
  %119 = call float @llvm.nvvm.fma.rn.ftz.f(float %89, float noundef 0x3FE62E4300000000, float %.07) #139
  %120 = call float @llvm.fma.f32(float %89, float noundef 0x3FE62E4300000000, float %.07)
  %.05 = select i1 %.not15, float %120, float %119
  %121 = icmp ugt i32 %79, 2139095039
  br i1 %121, label %122, label %__nv_log1pf.exit

122:                                              ; preds = %__nv_sqrtf.exit
  %123 = icmp sgt i32 %79, -1082130432
  br i1 %123, label %__nv_fmaf_rn.exit.i, label %127

__nv_fmaf_rn.exit.i:                              ; preds = %122
  %124 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %124, 0
  %125 = call float @llvm.nvvm.fma.rn.ftz.f(float %72, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %126 = call float @llvm.fma.f32(float %72, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.04 = select i1 %.not16, float %126, float %125
  br label %127

127:                                              ; preds = %__nv_fmaf_rn.exit.i, %122
  %r.i.0 = phi float [ %.04, %__nv_fmaf_rn.exit.i ], [ %.05, %122 ]
  %128 = fcmp oeq float %72, 0.000000e+00
  %r.i.1 = select i1 %128, float -0.000000e+00, float %r.i.0
  br label %__nv_log1pf.exit

__nv_log1pf.exit:                                 ; preds = %__nv_sqrtf.exit, %127, %__internal_accurate_logf.exit
  %t.0 = phi float [ %52, %__internal_accurate_logf.exit ], [ %r.i.1, %127 ], [ %.05, %__nv_sqrtf.exit ]
  ret float %t.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_asinhf(float %a) local_unnamed_addr #30 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = fcmp ogt float %.0, 0x47D0000000000000
  br i1 %3, label %4, label %__internal_fmad.exit

4:                                                ; preds = %__nv_fabsf.exit
  %5 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %5, 0
  br i1 %.not19, label %8, label %6

6:                                                ; preds = %4
  %7 = call float @llvm.nvvm.lg2.approx.ftz.f(float %.0) #139
  br label %__nv_fast_logf.exit

8:                                                ; preds = %4
  %9 = call float @llvm.nvvm.lg2.approx.f(float %.0) #139
  br label %__nv_fast_logf.exit

__nv_fast_logf.exit:                              ; preds = %8, %6
  %.01 = phi float [ %7, %6 ], [ %9, %8 ]
  %10 = fmul float %.01, 0x3FE62E4300000000
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.add.rn.ftz.f(float noundef 0x3FE62E4300000000, float %10) #139
  %13 = fadd float %10, 0x3FE62E4300000000
  %.02 = select i1 %.not20, float %13, float %12
  br label %__nv_log1pf.exit

__internal_fmad.exit:                             ; preds = %__nv_fabsf.exit
  %14 = fdiv float 1.000000e+00, %.0
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %14, float %14, float noundef 1.000000e+00) #139
  %17 = call float @llvm.fma.f32(float %14, float %14, float noundef 1.000000e+00)
  %.03 = select i1 %.not1, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %18, 0
  br i1 %.not2, label %25, label %19

19:                                               ; preds = %__internal_fmad.exit
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %20, 0
  br i1 %.not18, label %23, label %21

21:                                               ; preds = %19
  %22 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %.03) #139
  br label %__nv_sqrtf.exit

23:                                               ; preds = %19
  %24 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %.03) #139
  br label %__nv_sqrtf.exit

25:                                               ; preds = %__internal_fmad.exit
  %26 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %26, 0
  br i1 %.not3, label %29, label %27

27:                                               ; preds = %25
  %28 = call float @llvm.sqrt.f32(float %.03)
  br label %__nv_sqrtf.exit

29:                                               ; preds = %25
  %30 = call float @llvm.nvvm.sqrt.approx.f(float %.03) #139
  br label %__nv_sqrtf.exit

__nv_sqrtf.exit:                                  ; preds = %29, %27, %23, %21
  %.04 = phi float [ %22, %21 ], [ %24, %23 ], [ %28, %27 ], [ %30, %29 ]
  %31 = fadd float %14, %.04
  %32 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %31) #141, !srcloc !73
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0, float %32, float %.0) #139
  %35 = call float @llvm.fma.f32(float %.0, float %32, float %.0)
  %.05 = select i1 %.not4, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.add.rz.ftz.f(float %.05, float noundef 1.000000e+00) #139
  %38 = call float @llvm.nvvm.add.rz.f(float %.05, float noundef 1.000000e+00) #139
  %.06 = select i1 %.not5, float %38, float %37
  %39 = bitcast float %.06 to i32
  %40 = add i32 %39, -1061158912
  %41 = and i32 %40, -8388608
  %42 = bitcast float %.05 to i32
  %43 = sub i32 %42, %41
  %44 = bitcast i32 %43 to float
  %45 = sub i32 1082130432, %41
  %46 = bitcast i32 %45 to float
  %47 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %47, 0
  %48 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.500000e-01, float %46, float noundef -1.000000e+00) #139
  %49 = call float @llvm.fma.f32(float %46, float noundef 2.500000e-01, float noundef -1.000000e+00)
  %.014 = select i1 %.not6, float %49, float %48
  %50 = fadd float %.014, %44
  %51 = sitofp i32 %41 to float
  %52 = fmul float %51, 0x3E80000000000000
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFA737EF00000000, float %50, float noundef 0x3FBB000240000000) #139
  %55 = call float @llvm.fma.f32(float %50, float noundef 0xBFA737EF00000000, float noundef 0x3FBB000240000000)
  %.016 = select i1 %.not7, float %55, float %54
  %56 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %56, 0
  %57 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %50, float noundef 0xBFC0EF1C00000000) #139
  %58 = call float @llvm.fma.f32(float %.016, float %50, float noundef 0xBFC0EF1C00000000)
  %.018 = select i1 %.not8, float %58, float %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %50, float noundef 0x3FC28C8EA0000000) #139
  %61 = call float @llvm.fma.f32(float %.018, float %50, float noundef 0x3FC28C8EA0000000)
  %.017 = select i1 %.not9, float %61, float %60
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %50, float noundef 0xBFC54D1BA0000000) #139
  %64 = call float @llvm.fma.f32(float %.017, float %50, float noundef 0xBFC54D1BA0000000)
  %.013 = select i1 %.not10, float %64, float %63
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %50, float noundef 0x3FC995F3C0000000) #139
  %67 = call float @llvm.fma.f32(float %.013, float %50, float noundef 0x3FC995F3C0000000)
  %.09 = select i1 %.not11, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %50, float noundef 0xBFD0000840000000) #139
  %70 = call float @llvm.fma.f32(float %.09, float %50, float noundef 0xBFD0000840000000)
  %.015 = select i1 %.not12, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %50, float noundef 0x3FD5555CC0000000) #139
  %73 = call float @llvm.fma.f32(float %.015, float %50, float noundef 0x3FD5555CC0000000)
  %.012 = select i1 %.not13, float %73, float %72
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %74, 0
  %75 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %50, float noundef -5.000000e-01) #139
  %76 = call float @llvm.fma.f32(float %.012, float %50, float noundef -5.000000e-01)
  %.011 = select i1 %.not14, float %76, float %75
  %77 = fmul float %50, %.011
  %78 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %78, 0
  %79 = call float @llvm.nvvm.fma.rn.ftz.f(float %77, float %50, float %50) #139
  %80 = call float @llvm.fma.f32(float %77, float %50, float %50)
  %.010 = select i1 %.not15, float %80, float %79
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %81, 0
  %82 = call float @llvm.nvvm.fma.rn.ftz.f(float %52, float noundef 0x3FE62E4300000000, float %.010) #139
  %83 = call float @llvm.fma.f32(float %52, float noundef 0x3FE62E4300000000, float %.010)
  %.08 = select i1 %.not16, float %83, float %82
  %84 = icmp ugt i32 %42, 2139095039
  br i1 %84, label %85, label %__nv_log1pf.exit

85:                                               ; preds = %__nv_sqrtf.exit
  %86 = icmp sgt i32 %42, -1082130432
  br i1 %86, label %__nv_fmaf_rn.exit.i, label %90

__nv_fmaf_rn.exit.i:                              ; preds = %85
  %87 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %87, 0
  %88 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %89 = call float @llvm.fma.f32(float %.05, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.07 = select i1 %.not17, float %89, float %88
  br label %90

90:                                               ; preds = %__nv_fmaf_rn.exit.i, %85
  %r.i2.0 = phi float [ %.07, %__nv_fmaf_rn.exit.i ], [ %.08, %85 ]
  %91 = fcmp oeq float %.05, 0.000000e+00
  %r.i2.1 = select i1 %91, float -0.000000e+00, float %r.i2.0
  br label %__nv_log1pf.exit

__nv_log1pf.exit:                                 ; preds = %__nv_sqrtf.exit, %90, %__nv_fast_logf.exit
  %t.0 = phi float [ %.02, %__nv_fast_logf.exit ], [ %r.i2.1, %90 ], [ %.08, %__nv_sqrtf.exit ]
  %92 = fcmp ugt float %.0, 0x7FF0000000000000
  br i1 %92, label %99, label %93

93:                                               ; preds = %__nv_log1pf.exit
  %94 = bitcast float %t.0 to i32
  %95 = bitcast float %a to i32
  %96 = and i32 %95, -2147483648
  %97 = or i32 %96, %94
  %98 = bitcast i32 %97 to float
  br label %99

99:                                               ; preds = %93, %__nv_log1pf.exit
  %t.1 = phi float [ %98, %93 ], [ %t.0, %__nv_log1pf.exit ]
  ret float %t.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_atanhf(float %a) local_unnamed_addr #31 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = fsub float 1.000000e+00, %.0
  %4 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %3) #141, !srcloc !73
  %5 = fmul float %4, 2.000000e+00
  %6 = fmul float %5, %.0
  %7 = fcmp ogt float %.0, 0x47D0000000000000
  %t.0 = select i1 %7, float -2.000000e+00, float %6
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.add.rz.ftz.f(float %t.0, float noundef 1.000000e+00) #139
  %10 = call float @llvm.nvvm.add.rz.f(float %t.0, float noundef 1.000000e+00) #139
  %.01 = select i1 %.not1, float %10, float %9
  %11 = bitcast float %.01 to i32
  %12 = add i32 %11, -1061158912
  %13 = and i32 %12, -8388608
  %14 = bitcast float %t.0 to i32
  %15 = sub i32 %14, %13
  %16 = bitcast i32 %15 to float
  %17 = sub i32 1082130432, %13
  %18 = bitcast i32 %17 to float
  %19 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %19, 0
  %20 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.500000e-01, float %18, float noundef -1.000000e+00) #139
  %21 = call float @llvm.fma.f32(float %18, float noundef 2.500000e-01, float noundef -1.000000e+00)
  %.07 = select i1 %.not2, float %21, float %20
  %22 = fadd float %.07, %16
  %23 = sitofp i32 %13 to float
  %24 = fmul float %23, 0x3E80000000000000
  %25 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %25, 0
  %26 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFA737EF00000000, float %22, float noundef 0x3FBB000240000000) #139
  %27 = call float @llvm.fma.f32(float %22, float noundef 0xBFA737EF00000000, float noundef 0x3FBB000240000000)
  %.010 = select i1 %.not3, float %27, float %26
  %28 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %28, 0
  %29 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %22, float noundef 0xBFC0EF1C00000000) #139
  %30 = call float @llvm.fma.f32(float %.010, float %22, float noundef 0xBFC0EF1C00000000)
  %.012 = select i1 %.not4, float %30, float %29
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %31, 0
  %32 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %22, float noundef 0x3FC28C8EA0000000) #139
  %33 = call float @llvm.fma.f32(float %.012, float %22, float noundef 0x3FC28C8EA0000000)
  %.013 = select i1 %.not5, float %33, float %32
  %34 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %34, 0
  %35 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %22, float noundef 0xBFC54D1BA0000000) #139
  %36 = call float @llvm.fma.f32(float %.013, float %22, float noundef 0xBFC54D1BA0000000)
  %.014 = select i1 %.not6, float %36, float %35
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %22, float noundef 0x3FC995F3C0000000) #139
  %39 = call float @llvm.fma.f32(float %.014, float %22, float noundef 0x3FC995F3C0000000)
  %.011 = select i1 %.not7, float %39, float %38
  %40 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %40, 0
  %41 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %22, float noundef 0xBFD0000840000000) #139
  %42 = call float @llvm.fma.f32(float %.011, float %22, float noundef 0xBFD0000840000000)
  %.08 = select i1 %.not8, float %42, float %41
  %43 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %43, 0
  %44 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %22, float noundef 0x3FD5555CC0000000) #139
  %45 = call float @llvm.fma.f32(float %.08, float %22, float noundef 0x3FD5555CC0000000)
  %.06 = select i1 %.not9, float %45, float %44
  %46 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %46, 0
  %47 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %22, float noundef -5.000000e-01) #139
  %48 = call float @llvm.fma.f32(float %.06, float %22, float noundef -5.000000e-01)
  %.05 = select i1 %.not10, float %48, float %47
  %49 = fmul float %22, %.05
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float %49, float %22, float %22) #139
  %52 = call float @llvm.fma.f32(float %49, float %22, float %22)
  %.04 = select i1 %.not11, float %52, float %51
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fma.rn.ftz.f(float %24, float noundef 0x3FE62E4300000000, float %.04) #139
  %55 = call float @llvm.fma.f32(float %24, float noundef 0x3FE62E4300000000, float %.04)
  %.03 = select i1 %.not12, float %55, float %54
  %56 = icmp ugt i32 %14, 2139095039
  br i1 %56, label %57, label %__nv_log1pf.exit

57:                                               ; preds = %__nv_fabsf.exit
  %58 = icmp sgt i32 %14, -1082130432
  br i1 %58, label %__nv_fmaf_rn.exit.i, label %62

__nv_fmaf_rn.exit.i:                              ; preds = %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float %t.0, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %61 = call float @llvm.fma.f32(float %t.0, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.02 = select i1 %.not14, float %61, float %60
  br label %62

62:                                               ; preds = %__nv_fmaf_rn.exit.i, %57
  %r.i1.0 = phi float [ %.02, %__nv_fmaf_rn.exit.i ], [ %.03, %57 ]
  %63 = fcmp oeq float %t.0, 0.000000e+00
  %r.i1.1 = select i1 %63, float -0.000000e+00, float %r.i1.0
  br label %__nv_log1pf.exit

__nv_log1pf.exit:                                 ; preds = %62, %__nv_fabsf.exit
  %r.i1.2 = phi float [ %r.i1.1, %62 ], [ %.03, %__nv_fabsf.exit ]
  %64 = fmul float %r.i1.2, 5.000000e-01
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fabs.ftz.f(float %64) #139
  %67 = call float @llvm.fabs.f32(float %64)
  %.09 = select i1 %.not13, float %67, float %66
  %68 = fcmp ugt float %.09, 0x7FF0000000000000
  br i1 %68, label %75, label %69

69:                                               ; preds = %__nv_log1pf.exit
  %70 = bitcast float %64 to i32
  %71 = bitcast float %a to i32
  %72 = and i32 %71, -2147483648
  %73 = or i32 %72, %70
  %74 = bitcast i32 %73 to float
  br label %75

75:                                               ; preds = %69, %__nv_log1pf.exit
  %t.1 = phi float [ %74, %69 ], [ %64, %__nv_log1pf.exit ]
  ret float %t.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_expm1f(float %a) local_unnamed_addr #32 {
__nv_fabsf.exit:
  %0 = fmul float %a, 0x3FF7154760000000
  %1 = call float @llvm.round.f32(float %0) #140
  %2 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %2, 0
  %3 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %4 = call float @llvm.fabs.f32(float %a)
  %.03 = select i1 %.not, float %4, float %3
  %5 = fcmp olt float %.03, 0x3FDA3D70A0000000
  %t.0 = select i1 %5, float 0.000000e+00, float %1
  %6 = fneg float %t.0
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %7, 0
  %8 = call float @llvm.nvvm.fma.rn.ftz.f(float %6, float noundef 0x3FE62E4000000000, float %a) #139
  %9 = call float @llvm.fma.f32(float %6, float noundef 0x3FE62E4000000000, float %a)
  %.04 = select i1 %.not1, float %9, float %8
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %10, 0
  %11 = call float @llvm.nvvm.fma.rn.ftz.f(float %6, float noundef 0x3EB7F7D1C0000000, float %.04) #139
  %12 = call float @llvm.fma.f32(float %6, float noundef 0x3EB7F7D1C0000000, float %.04)
  %.05 = select i1 %.not2, float %12, float %11
  %13 = fcmp oeq float %t.0, 1.280000e+02
  %14 = fadd float %t.0, -1.000000e+00
  %j.0 = select i1 %13, float %14, float %t.0
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F56BD7CC0000000, float %.05, float noundef 0x3F812ACC60000000) #139
  %17 = call float @llvm.fma.f32(float %.05, float noundef 0x3F56BD7CC0000000, float noundef 0x3F812ACC60000000)
  %.07 = select i1 %.not3, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %.05, float noundef 0x3FA5557C60000000) #139
  %20 = call float @llvm.fma.f32(float %.07, float %.05, float noundef 0x3FA5557C60000000)
  %.08 = select i1 %.not4, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %.05, float noundef 0x3FC5553EC0000000) #139
  %23 = call float @llvm.fma.f32(float %.08, float %.05, float noundef 0x3FC5553EC0000000)
  %.09 = select i1 %.not5, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %.05, float noundef 0x3FDFFFFFC0000000) #139
  %26 = call float @llvm.fma.f32(float %.09, float %.05, float noundef 0x3FDFFFFFC0000000)
  %.06 = select i1 %.not6, float %26, float %25
  %27 = fmul float %.05, %.06
  %28 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %28, 0
  %29 = call float @llvm.nvvm.fma.rn.ftz.f(float %27, float %.05, float %.05) #139
  %30 = call float @llvm.fma.f32(float %27, float %.05, float %.05)
  %.01 = select i1 %.not7, float %30, float %29
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %31, 0
  br i1 %.not8, label %34, label %32

32:                                               ; preds = %__nv_fabsf.exit
  %33 = call float @llvm.nvvm.ex2.approx.ftz.f(float %j.0) #139
  br label %__nv_exp2f.exit

34:                                               ; preds = %__nv_fabsf.exit
  %35 = call float @llvm.nvvm.ex2.approx.f(float %j.0) #139
  br label %__nv_exp2f.exit

__nv_exp2f.exit:                                  ; preds = %34, %32
  %.0 = phi float [ %33, %32 ], [ %35, %34 ]
  %36 = fadd float %.0, -1.000000e+00
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %.0, float %36) #139
  %39 = call float @llvm.fma.f32(float %.01, float %.0, float %36)
  %.02 = select i1 %.not9, float %39, float %38
  %40 = select i1 %13, float %.02, float -0.000000e+00
  %u.0 = fadd float %.02, %40
  %41 = fcmp ogt float %j.0, 1.280000e+02
  %u.1 = select i1 %41, float 0x7FF0000000000000, float %u.0
  %42 = fcmp olt float %j.0, -2.500000e+01
  %u.2 = select i1 %42, float -1.000000e+00, float %u.1
  %43 = fcmp oeq float %a, 0.000000e+00
  %44 = fadd float %a, %a
  %u.3 = select i1 %43, float %44, float %u.2
  ret float %u.3
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_hypotf(float %a, float %b) local_unnamed_addr #33 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %5 = call float @llvm.fabs.f32(float %b)
  %.01 = select i1 %.not1, float %5, float %4
  %6 = bitcast float %.01 to i32
  %7 = bitcast float %.0 to i32
  %min.cond.not = icmp sgt i32 %6, %7
  %min = select i1 %min.cond.not, i32 %7, i32 %6
  %8 = bitcast i32 %min to float
  %max = select i1 %min.cond.not, i32 %6, i32 %7
  %9 = bitcast i32 %max to float
  %10 = and i32 %max, -33554432
  %11 = sub i32 2122317824, %10
  %12 = bitcast i32 %11 to float
  %13 = fmul float %8, %12
  %14 = fmul float %9, %12
  %15 = fmul float %13, %13
  %16 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %16, 0
  %17 = call float @llvm.nvvm.fma.rn.ftz.f(float %14, float %14, float %15) #139
  %18 = call float @llvm.fma.f32(float %14, float %14, float %15)
  %.03 = select i1 %.not2, float %18, float %17
  %19 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %19, 0
  br i1 %.not3, label %26, label %20

20:                                               ; preds = %__nv_fabsf.exit
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %21, 0
  br i1 %.not5, label %24, label %22

22:                                               ; preds = %20
  %23 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %.03) #139
  br label %__nv_sqrtf.exit

24:                                               ; preds = %20
  %25 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %.03) #139
  br label %__nv_sqrtf.exit

26:                                               ; preds = %__nv_fabsf.exit
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %27, 0
  br i1 %.not4, label %30, label %28

28:                                               ; preds = %26
  %29 = call float @llvm.sqrt.f32(float %.03)
  br label %__nv_sqrtf.exit

30:                                               ; preds = %26
  %31 = call float @llvm.nvvm.sqrt.approx.f(float %.03) #139
  br label %__nv_sqrtf.exit

__nv_sqrtf.exit:                                  ; preds = %30, %28, %24, %22
  %.02 = phi float [ %23, %22 ], [ %25, %24 ], [ %29, %28 ], [ %31, %30 ]
  %32 = or i32 %10, 8388608
  %33 = bitcast i32 %32 to float
  %34 = fmul float %.02, %33
  %35 = fcmp oeq float %8, 0.000000e+00
  %res.0 = select i1 %35, float %9, float %34
  %36 = fcmp oeq float %8, 0x7FF0000000000000
  %res.1 = select i1 %36, float 0x7FF0000000000000, float %res.0
  ret float %res.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_rhypotf(float %a, float %b) local_unnamed_addr #34 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %5 = call float @llvm.fabs.f32(float %b)
  %.01 = select i1 %.not1, float %5, float %4
  %6 = bitcast float %.01 to i32
  %7 = bitcast float %.0 to i32
  %min.cond.not = icmp sgt i32 %6, %7
  %min = select i1 %min.cond.not, i32 %7, i32 %6
  %8 = bitcast i32 %min to float
  %max = select i1 %min.cond.not, i32 %6, i32 %7
  %9 = bitcast i32 %max to float
  %10 = and i32 %max, -33554432
  %11 = sub i32 2122317824, %10
  %12 = bitcast i32 %11 to float
  %13 = fmul float %8, %12
  %14 = fmul float %9, %12
  %15 = fmul float %13, %13
  %16 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %16, 0
  %17 = call float @llvm.nvvm.fma.rn.ftz.f(float %14, float %14, float %15) #139
  %18 = call float @llvm.fma.f32(float %14, float %14, float %15)
  %.03 = select i1 %.not2, float %18, float %17
  %19 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %19, 0
  br i1 %.not3, label %22, label %20

20:                                               ; preds = %__nv_fabsf.exit
  %21 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.03) #139
  br label %__nv_rsqrtf.exit

22:                                               ; preds = %__nv_fabsf.exit
  %23 = call float @llvm.nvvm.rsqrt.approx.f(float %.03) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %22, %20
  %.02 = phi float [ %21, %20 ], [ %23, %22 ]
  %24 = fmul float %.02, %12
  %25 = fcmp oeq float %8, 0x7FF0000000000000
  %res.0 = select i1 %25, float 0.000000e+00, float %24
  ret float %res.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_norm3df(float %a, float %b, float %c) local_unnamed_addr #35 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.02 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %5 = call float @llvm.fabs.f32(float %b)
  %.04 = select i1 %.not1, float %5, float %4
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fabs.ftz.f(float %c) #139
  %8 = call float @llvm.fabs.f32(float %c)
  %.05 = select i1 %.not2, float %8, float %7
  %9 = fadd float %.02, %.04
  %10 = fadd float %9, %.05
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fmin.ftz.f(float %.02, float %.04) #139
  %13 = call float @llvm.minnum.f32(float %.02, float %.04)
  %.06 = select i1 %.not3, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fmax.ftz.f(float %.02, float %.04) #139
  %16 = call float @llvm.maxnum.f32(float %.02, float %.04)
  %.07 = select i1 %.not4, float %16, float %15
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fmin.ftz.f(float %.07, float %.05) #139
  %19 = call float @llvm.minnum.f32(float %.07, float %.05)
  %.08 = select i1 %.not5, float %19, float %18
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fmax.ftz.f(float %.07, float %.05) #139
  %22 = call float @llvm.maxnum.f32(float %.07, float %.05)
  %.09 = select i1 %.not6, float %22, float %21
  %23 = bitcast float %.09 to i32
  %24 = and i32 %23, -33554432
  %25 = sub i32 2122317824, %24
  %26 = bitcast i32 %25 to float
  %27 = fmul float %.08, %26
  %28 = fmul float %.06, %26
  %29 = fmul float %.09, %26
  %30 = fmul float %27, %27
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %31, 0
  %32 = call float @llvm.nvvm.fma.rn.ftz.f(float %28, float %28, float %30) #139
  %33 = call float @llvm.fma.f32(float %28, float %28, float %30)
  %.01 = select i1 %.not7, float %33, float %32
  %34 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %34, 0
  %35 = call float @llvm.nvvm.fma.rn.ftz.f(float %29, float %29, float %.01) #139
  %36 = call float @llvm.fma.f32(float %29, float %29, float %.01)
  %.0 = select i1 %.not8, float %36, float %35
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %37, 0
  br i1 %.not9, label %44, label %38

38:                                               ; preds = %__nv_fabsf.exit
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %39, 0
  br i1 %.not11, label %42, label %40

40:                                               ; preds = %38
  %41 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %.0) #139
  br label %__nv_sqrtf.exit

42:                                               ; preds = %38
  %43 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %.0) #139
  br label %__nv_sqrtf.exit

44:                                               ; preds = %__nv_fabsf.exit
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %45, 0
  br i1 %.not10, label %48, label %46

46:                                               ; preds = %44
  %47 = call float @llvm.sqrt.f32(float %.0)
  br label %__nv_sqrtf.exit

48:                                               ; preds = %44
  %49 = call float @llvm.nvvm.sqrt.approx.f(float %.0) #139
  br label %__nv_sqrtf.exit

__nv_sqrtf.exit:                                  ; preds = %48, %46, %42, %40
  %.03 = phi float [ %41, %40 ], [ %43, %42 ], [ %47, %46 ], [ %49, %48 ]
  %50 = or i32 %24, 8388608
  %51 = bitcast i32 %50 to float
  %52 = fmul float %.03, %51
  %53 = fcmp ogt float %10, %.09
  %res.0 = select i1 %53, float %52, float %10
  %54 = fcmp oeq float %.09, 0x7FF0000000000000
  %res.1 = select i1 %54, float 0x7FF0000000000000, float %res.0
  ret float %res.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_rnorm3df(float %a, float %b, float %c) local_unnamed_addr #36 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.02 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %5 = call float @llvm.fabs.f32(float %b)
  %.03 = select i1 %.not1, float %5, float %4
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fabs.ftz.f(float %c) #139
  %8 = call float @llvm.fabs.f32(float %c)
  %.04 = select i1 %.not2, float %8, float %7
  %9 = fadd float %.02, %.03
  %10 = fadd float %9, %.04
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fmin.ftz.f(float %.02, float %.03) #139
  %13 = call float @llvm.minnum.f32(float %.02, float %.03)
  %.05 = select i1 %.not3, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fmax.ftz.f(float %.02, float %.03) #139
  %16 = call float @llvm.maxnum.f32(float %.02, float %.03)
  %.06 = select i1 %.not4, float %16, float %15
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fmin.ftz.f(float %.06, float %.04) #139
  %19 = call float @llvm.minnum.f32(float %.06, float %.04)
  %.08 = select i1 %.not5, float %19, float %18
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fmax.ftz.f(float %.06, float %.04) #139
  %22 = call float @llvm.maxnum.f32(float %.06, float %.04)
  %.09 = select i1 %.not6, float %22, float %21
  %23 = bitcast float %.09 to i32
  %24 = and i32 %23, -33554432
  %25 = sub i32 2122317824, %24
  %26 = bitcast i32 %25 to float
  %27 = fmul float %.08, %26
  %28 = fmul float %.05, %26
  %29 = fmul float %.09, %26
  %30 = fmul float %27, %27
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %31, 0
  %32 = call float @llvm.nvvm.fma.rn.ftz.f(float %28, float %28, float %30) #139
  %33 = call float @llvm.fma.f32(float %28, float %28, float %30)
  %.01 = select i1 %.not7, float %33, float %32
  %34 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %34, 0
  %35 = call float @llvm.nvvm.fma.rn.ftz.f(float %29, float %29, float %.01) #139
  %36 = call float @llvm.fma.f32(float %29, float %29, float %.01)
  %.0 = select i1 %.not8, float %36, float %35
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %37, 0
  br i1 %.not9, label %40, label %38

38:                                               ; preds = %__nv_fabsf.exit
  %39 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0) #139
  br label %__nv_rsqrtf.exit

40:                                               ; preds = %__nv_fabsf.exit
  %41 = call float @llvm.nvvm.rsqrt.approx.f(float %.0) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %40, %38
  %.07 = phi float [ %39, %38 ], [ %41, %40 ]
  %42 = fmul float %.07, %26
  %43 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %43, 0
  %44 = call float @llvm.nvvm.fabs.ftz.f(float %10) #139
  %45 = call float @llvm.fabs.f32(float %10)
  %.010 = select i1 %.not10, float %45, float %44
  %46 = fcmp ugt float %.010, 0x7FF0000000000000
  %res.0 = select i1 %46, float %10, float %42
  %47 = fcmp oeq float %.09, 0x7FF0000000000000
  %res.1 = select i1 %47, float 0.000000e+00, float %res.0
  ret float %res.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_norm4df(float %a, float %b, float %c, float %d) local_unnamed_addr #37 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.02 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %5 = call float @llvm.fabs.f32(float %b)
  %.03 = select i1 %.not1, float %5, float %4
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fabs.ftz.f(float %c) #139
  %8 = call float @llvm.fabs.f32(float %c)
  %.04 = select i1 %.not2, float %8, float %7
  %9 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %9, 0
  %10 = call float @llvm.nvvm.fabs.ftz.f(float %d) #139
  %11 = call float @llvm.fabs.f32(float %d)
  %.06 = select i1 %.not3, float %11, float %10
  %12 = fadd float %.02, %.03
  %13 = fadd float %12, %.06
  %14 = fadd float %.04, %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fmin.ftz.f(float %.02, float %.03) #139
  %17 = call float @llvm.minnum.f32(float %.02, float %.03)
  %.07 = select i1 %.not4, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fmax.ftz.f(float %.02, float %.03) #139
  %20 = call float @llvm.maxnum.f32(float %.02, float %.03)
  %.08 = select i1 %.not5, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fmin.ftz.f(float %.08, float %.04) #139
  %23 = call float @llvm.minnum.f32(float %.08, float %.04)
  %.09 = select i1 %.not6, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fmax.ftz.f(float %.08, float %.04) #139
  %26 = call float @llvm.maxnum.f32(float %.08, float %.04)
  %.010 = select i1 %.not7, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fmin.ftz.f(float %.010, float %.06) #139
  %29 = call float @llvm.minnum.f32(float %.010, float %.06)
  %.011 = select i1 %.not8, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fmax.ftz.f(float %.010, float %.06) #139
  %32 = call float @llvm.maxnum.f32(float %.010, float %.06)
  %.012 = select i1 %.not9, float %32, float %31
  %33 = bitcast float %.012 to i32
  %34 = and i32 %33, -33554432
  %35 = sub i32 2122317824, %34
  %36 = bitcast i32 %35 to float
  %37 = fmul float %.011, %36
  %38 = fmul float %.09, %36
  %39 = fmul float %.07, %36
  %40 = fmul float %.012, %36
  %41 = fmul float %37, %37
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %38, float %38, float %41) #139
  %44 = call float @llvm.fma.f32(float %38, float %38, float %41)
  %.05 = select i1 %.not10, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %39, float %39, float %.05) #139
  %47 = call float @llvm.fma.f32(float %39, float %39, float %.05)
  %.01 = select i1 %.not11, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %40, float %40, float %.01) #139
  %50 = call float @llvm.fma.f32(float %40, float %40, float %.01)
  %.0 = select i1 %.not12, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %51, 0
  br i1 %.not13, label %58, label %52

52:                                               ; preds = %__nv_fabsf.exit
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %53, 0
  br i1 %.not15, label %56, label %54

54:                                               ; preds = %52
  %55 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %.0) #139
  br label %__nv_sqrtf.exit

56:                                               ; preds = %52
  %57 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %.0) #139
  br label %__nv_sqrtf.exit

58:                                               ; preds = %__nv_fabsf.exit
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %59, 0
  br i1 %.not14, label %62, label %60

60:                                               ; preds = %58
  %61 = call float @llvm.sqrt.f32(float %.0)
  br label %__nv_sqrtf.exit

62:                                               ; preds = %58
  %63 = call float @llvm.nvvm.sqrt.approx.f(float %.0) #139
  br label %__nv_sqrtf.exit

__nv_sqrtf.exit:                                  ; preds = %62, %60, %56, %54
  %.013 = phi float [ %55, %54 ], [ %57, %56 ], [ %61, %60 ], [ %63, %62 ]
  %64 = or i32 %34, 8388608
  %65 = bitcast i32 %64 to float
  %66 = fmul float %.013, %65
  %67 = fcmp ogt float %14, %.012
  %res.0 = select i1 %67, float %66, float %14
  %68 = fcmp oeq float %.012, 0x7FF0000000000000
  %res.1 = select i1 %68, float 0x7FF0000000000000, float %res.0
  ret float %res.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_rnorm4df(float %a, float %b, float %c, float %d) local_unnamed_addr #38 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.04 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %5 = call float @llvm.fabs.f32(float %b)
  %.05 = select i1 %.not1, float %5, float %4
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fabs.ftz.f(float %c) #139
  %8 = call float @llvm.fabs.f32(float %c)
  %.06 = select i1 %.not2, float %8, float %7
  %9 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %9, 0
  %10 = call float @llvm.nvvm.fabs.ftz.f(float %d) #139
  %11 = call float @llvm.fabs.f32(float %d)
  %.07 = select i1 %.not3, float %11, float %10
  %12 = fadd float %.04, %.05
  %13 = fadd float %12, %.07
  %14 = fadd float %.06, %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fmin.ftz.f(float %.04, float %.05) #139
  %17 = call float @llvm.minnum.f32(float %.04, float %.05)
  %.08 = select i1 %.not4, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fmax.ftz.f(float %.04, float %.05) #139
  %20 = call float @llvm.maxnum.f32(float %.04, float %.05)
  %.010 = select i1 %.not5, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fmin.ftz.f(float %.010, float %.06) #139
  %23 = call float @llvm.minnum.f32(float %.010, float %.06)
  %.011 = select i1 %.not6, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fmax.ftz.f(float %.010, float %.06) #139
  %26 = call float @llvm.maxnum.f32(float %.010, float %.06)
  %.012 = select i1 %.not7, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fmin.ftz.f(float %.012, float %.07) #139
  %29 = call float @llvm.minnum.f32(float %.012, float %.07)
  %.013 = select i1 %.not8, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fmax.ftz.f(float %.012, float %.07) #139
  %32 = call float @llvm.maxnum.f32(float %.012, float %.07)
  %.014 = select i1 %.not9, float %32, float %31
  %33 = bitcast float %.014 to i32
  %34 = and i32 %33, -33554432
  %35 = sub i32 2122317824, %34
  %36 = bitcast i32 %35 to float
  %37 = fmul float %.013, %36
  %38 = fmul float %.011, %36
  %39 = fmul float %.08, %36
  %40 = fmul float %.014, %36
  %41 = fmul float %37, %37
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %38, float %38, float %41) #139
  %44 = call float @llvm.fma.f32(float %38, float %38, float %41)
  %.02 = select i1 %.not10, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %39, float %39, float %.02) #139
  %47 = call float @llvm.fma.f32(float %39, float %39, float %.02)
  %.01 = select i1 %.not11, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %40, float %40, float %.01) #139
  %50 = call float @llvm.fma.f32(float %40, float %40, float %.01)
  %.0 = select i1 %.not12, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %51, 0
  br i1 %.not13, label %54, label %52

52:                                               ; preds = %__nv_fabsf.exit
  %53 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0) #139
  br label %__nv_rsqrtf.exit

54:                                               ; preds = %__nv_fabsf.exit
  %55 = call float @llvm.nvvm.rsqrt.approx.f(float %.0) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %54, %52
  %.03 = phi float [ %53, %52 ], [ %55, %54 ]
  %56 = fmul float %.03, %36
  %57 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %57, 0
  %58 = call float @llvm.nvvm.fabs.ftz.f(float %14) #139
  %59 = call float @llvm.fabs.f32(float %14)
  %.09 = select i1 %.not14, float %59, float %58
  %60 = fcmp ugt float %.09, 0x7FF0000000000000
  %res.0 = select i1 %60, float %14, float %56
  %61 = fcmp oeq float %.014, 0x7FF0000000000000
  %res.1 = select i1 %61, float 0.000000e+00, float %res.0
  ret float %res.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_normf(i32 %dim, float* nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) %t) local_unnamed_addr #39 {
__nv_fabsf.exit:
  %0 = load float, float* %t, align 4
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call float @llvm.nvvm.fabs.ftz.f(float %0) #139
  %3 = call float @llvm.fabs.f32(float %0)
  %.02 = select i1 %.not, float %3, float %2
  %4 = icmp sgt i32 %dim, 1
  br i1 %4, label %__nv_fabsf.exit1.lr.ph, label %._crit_edge

__nv_fabsf.exit1.lr.ph:                           ; preds = %__nv_fabsf.exit
  %5 = add nsw i32 %dim, -1
  %xtraiter = and i32 %5, 1
  %6 = icmp eq i32 %dim, 2
  br i1 %6, label %._crit_edge.unr-lcssa, label %__nv_fabsf.exit1.lr.ph.new

__nv_fabsf.exit1.lr.ph.new:                       ; preds = %__nv_fabsf.exit1.lr.ph
  %unroll_iter = and i32 %5, -2
  br label %__nv_fabsf.exit1

__nv_fabsf.exit1:                                 ; preds = %__nv_fabsf.exit1, %__nv_fabsf.exit1.lr.ph.new
  %i.015 = phi i32 [ 1, %__nv_fabsf.exit1.lr.ph.new ], [ %26, %__nv_fabsf.exit1 ]
  %largest.014 = phi float [ %.02, %__nv_fabsf.exit1.lr.ph.new ], [ %.05.1, %__nv_fabsf.exit1 ]
  %niter = phi i32 [ 0, %__nv_fabsf.exit1.lr.ph.new ], [ %niter.next.1, %__nv_fabsf.exit1 ]
  %7 = zext i32 %i.015 to i64
  %8 = getelementptr inbounds float, float* %t, i64 %7
  %9 = load float, float* %8, align 4
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %10, 0
  %11 = call float @llvm.nvvm.fabs.ftz.f(float %9) #139
  %12 = call float @llvm.fabs.f32(float %9)
  %.04 = select i1 %.not9, float %12, float %11
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %13, 0
  %14 = call float @llvm.nvvm.fmax.ftz.f(float %.04, float %largest.014) #139
  %15 = call float @llvm.maxnum.f32(float %.04, float %largest.014)
  %.05 = select i1 %.not10, float %15, float %14
  %16 = add nuw nsw i32 %i.015, 1
  %17 = zext i32 %16 to i64
  %18 = getelementptr inbounds float, float* %t, i64 %17
  %19 = load float, float* %18, align 4
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9.1 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fabs.ftz.f(float %19) #139
  %22 = call float @llvm.fabs.f32(float %19)
  %.04.1 = select i1 %.not9.1, float %22, float %21
  %23 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10.1 = icmp eq i32 %23, 0
  %24 = call float @llvm.nvvm.fmax.ftz.f(float %.04.1, float %.05) #139
  %25 = call float @llvm.maxnum.f32(float %.04.1, float %.05)
  %.05.1 = select i1 %.not10.1, float %25, float %24
  %26 = add nuw nsw i32 %i.015, 2
  %niter.next.1 = add i32 %niter, 2
  %niter.ncmp.1.not = icmp eq i32 %niter.next.1, %unroll_iter
  br i1 %niter.ncmp.1.not, label %._crit_edge.unr-lcssa.loopexit, label %__nv_fabsf.exit1

._crit_edge.unr-lcssa.loopexit:                   ; preds = %__nv_fabsf.exit1
  %phi.cast = zext i32 %26 to i64
  br label %._crit_edge.unr-lcssa

._crit_edge.unr-lcssa:                            ; preds = %._crit_edge.unr-lcssa.loopexit, %__nv_fabsf.exit1.lr.ph
  %.05.lcssa.ph = phi float [ undef, %__nv_fabsf.exit1.lr.ph ], [ %.05.1, %._crit_edge.unr-lcssa.loopexit ]
  %i.015.unr = phi i64 [ 1, %__nv_fabsf.exit1.lr.ph ], [ %phi.cast, %._crit_edge.unr-lcssa.loopexit ]
  %largest.014.unr = phi float [ %.02, %__nv_fabsf.exit1.lr.ph ], [ %.05.1, %._crit_edge.unr-lcssa.loopexit ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %._crit_edge, label %__nv_fabsf.exit1.epil.preheader

__nv_fabsf.exit1.epil.preheader:                  ; preds = %._crit_edge.unr-lcssa
  %27 = getelementptr inbounds float, float* %t, i64 %i.015.unr
  %28 = load float, float* %27, align 4
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9.epil = icmp eq i32 %29, 0
  %30 = call float @llvm.nvvm.fabs.ftz.f(float %28) #139
  %31 = call float @llvm.fabs.f32(float %28)
  %.04.epil = select i1 %.not9.epil, float %31, float %30
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10.epil = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fmax.ftz.f(float %.04.epil, float %largest.014.unr) #139
  %34 = call float @llvm.maxnum.f32(float %.04.epil, float %largest.014.unr)
  %.05.epil = select i1 %.not10.epil, float %34, float %33
  br label %._crit_edge

._crit_edge:                                      ; preds = %__nv_fabsf.exit1.epil.preheader, %._crit_edge.unr-lcssa, %__nv_fabsf.exit
  %largest.0.lcssa = phi float [ %.02, %__nv_fabsf.exit ], [ %.05.lcssa.ph, %._crit_edge.unr-lcssa ], [ %.05.epil, %__nv_fabsf.exit1.epil.preheader ]
  %35 = bitcast float %largest.0.lcssa to i32
  %36 = and i32 %35, -33554432
  %37 = sub i32 2122317824, %36
  %38 = bitcast i32 %37 to float
  %39 = fmul float %largest.0.lcssa, %38
  %40 = icmp sgt i32 %dim, 0
  br i1 %40, label %__nv_fabsf.exit5, label %__nv_fmaf_rn.exit6

__nv_fabsf.exit5:                                 ; preds = %._crit_edge, %54
  %j.013 = phi i32 [ %j.1, %54 ], [ 0, %._crit_edge ]
  %i.112 = phi i32 [ %55, %54 ], [ 0, %._crit_edge ]
  %res0.011 = phi float [ %res0.1, %54 ], [ 0.000000e+00, %._crit_edge ]
  %41 = zext i32 %i.112 to i64
  %42 = getelementptr inbounds float, float* %t, i64 %41
  %43 = load float, float* %42, align 4
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fabs.ftz.f(float %43) #139
  %46 = call float @llvm.fabs.f32(float %43)
  %.03 = select i1 %.not7, float %46, float %45
  %47 = fmul float %.03, %38
  %48 = fcmp une float %47, %39
  %49 = icmp ne i32 %i.112, %j.013
  %or.cond = select i1 %48, i1 true, i1 %49
  br i1 %or.cond, label %__nv_fmaf_rn.exit, label %54

__nv_fmaf_rn.exit:                                ; preds = %__nv_fabsf.exit5
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float %47, float %47, float %res0.011) #139
  %52 = call float @llvm.fma.f32(float %47, float %47, float %res0.011)
  %.01 = select i1 %.not8, float %52, float %51
  %53 = add nsw i32 %j.013, 1
  br label %54

54:                                               ; preds = %__nv_fmaf_rn.exit, %__nv_fabsf.exit5
  %res0.1 = phi float [ %.01, %__nv_fmaf_rn.exit ], [ %res0.011, %__nv_fabsf.exit5 ]
  %j.1 = phi i32 [ %53, %__nv_fmaf_rn.exit ], [ %j.013, %__nv_fabsf.exit5 ]
  %55 = add nuw nsw i32 %i.112, 1
  %exitcond.not = icmp eq i32 %55, %dim
  br i1 %exitcond.not, label %__nv_fmaf_rn.exit6, label %__nv_fabsf.exit5

__nv_fmaf_rn.exit6:                               ; preds = %54, %._crit_edge
  %res0.0.lcssa = phi float [ 0.000000e+00, %._crit_edge ], [ %res0.1, %54 ]
  %56 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %56, 0
  %57 = call float @llvm.nvvm.fma.rn.ftz.f(float %39, float %39, float %res0.0.lcssa) #139
  %58 = call float @llvm.fma.f32(float %39, float %39, float %res0.0.lcssa)
  %.0 = select i1 %.not3, float %58, float %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %59, 0
  br i1 %.not4, label %66, label %60

60:                                               ; preds = %__nv_fmaf_rn.exit6
  %61 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %61, 0
  br i1 %.not6, label %64, label %62

62:                                               ; preds = %60
  %63 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %.0) #139
  br label %__nv_sqrtf.exit

64:                                               ; preds = %60
  %65 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %.0) #139
  br label %__nv_sqrtf.exit

66:                                               ; preds = %__nv_fmaf_rn.exit6
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %67, 0
  br i1 %.not5, label %70, label %68

68:                                               ; preds = %66
  %69 = call float @llvm.sqrt.f32(float %.0)
  br label %__nv_sqrtf.exit

70:                                               ; preds = %66
  %71 = call float @llvm.nvvm.sqrt.approx.f(float %.0) #139
  br label %__nv_sqrtf.exit

__nv_sqrtf.exit:                                  ; preds = %70, %68, %64, %62
  %.06 = phi float [ %63, %62 ], [ %65, %64 ], [ %69, %68 ], [ %71, %70 ]
  %72 = or i32 %36, 8388608
  %73 = bitcast i32 %72 to float
  %74 = fmul float %.06, %73
  %75 = fcmp oeq float %res0.0.lcssa, 0.000000e+00
  %76 = fcmp oeq float %largest.0.lcssa, 0x7FF0000000000000
  %or.cond2 = select i1 %75, i1 true, i1 %76
  %res.0 = select i1 %or.cond2, float %largest.0.lcssa, float %74
  ret float %res.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_rnormf(i32 %dim, float* nocapture nofree noundef nonnull readonly align 4 dereferenceable(4) %t) local_unnamed_addr #40 {
__nv_fabsf.exit:
  %0 = load float, float* %t, align 4
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call float @llvm.nvvm.fabs.ftz.f(float %0) #139
  %3 = call float @llvm.fabs.f32(float %0)
  %.03 = select i1 %.not, float %3, float %2
  %4 = icmp sgt i32 %dim, 1
  br i1 %4, label %__nv_fabsf.exit1.lr.ph, label %._crit_edge

__nv_fabsf.exit1.lr.ph:                           ; preds = %__nv_fabsf.exit
  %5 = add nsw i32 %dim, -1
  %xtraiter = and i32 %5, 1
  %6 = icmp eq i32 %dim, 2
  br i1 %6, label %._crit_edge.unr-lcssa, label %__nv_fabsf.exit1.lr.ph.new

__nv_fabsf.exit1.lr.ph.new:                       ; preds = %__nv_fabsf.exit1.lr.ph
  %unroll_iter = and i32 %5, -2
  br label %__nv_fabsf.exit1

__nv_fabsf.exit1:                                 ; preds = %__nv_fabsf.exit1, %__nv_fabsf.exit1.lr.ph.new
  %i.011 = phi i32 [ 1, %__nv_fabsf.exit1.lr.ph.new ], [ %26, %__nv_fabsf.exit1 ]
  %largest.010 = phi float [ %.03, %__nv_fabsf.exit1.lr.ph.new ], [ %.05.1, %__nv_fabsf.exit1 ]
  %niter = phi i32 [ 0, %__nv_fabsf.exit1.lr.ph.new ], [ %niter.next.1, %__nv_fabsf.exit1 ]
  %7 = zext i32 %i.011 to i64
  %8 = getelementptr inbounds float, float* %t, i64 %7
  %9 = load float, float* %8, align 4
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %10, 0
  %11 = call float @llvm.nvvm.fabs.ftz.f(float %9) #139
  %12 = call float @llvm.fabs.f32(float %9)
  %.04 = select i1 %.not5, float %12, float %11
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %13, 0
  %14 = call float @llvm.nvvm.fmax.ftz.f(float %.04, float %largest.010) #139
  %15 = call float @llvm.maxnum.f32(float %.04, float %largest.010)
  %.05 = select i1 %.not6, float %15, float %14
  %16 = add nuw nsw i32 %i.011, 1
  %17 = zext i32 %16 to i64
  %18 = getelementptr inbounds float, float* %t, i64 %17
  %19 = load float, float* %18, align 4
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5.1 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fabs.ftz.f(float %19) #139
  %22 = call float @llvm.fabs.f32(float %19)
  %.04.1 = select i1 %.not5.1, float %22, float %21
  %23 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6.1 = icmp eq i32 %23, 0
  %24 = call float @llvm.nvvm.fmax.ftz.f(float %.04.1, float %.05) #139
  %25 = call float @llvm.maxnum.f32(float %.04.1, float %.05)
  %.05.1 = select i1 %.not6.1, float %25, float %24
  %26 = add nuw nsw i32 %i.011, 2
  %niter.next.1 = add i32 %niter, 2
  %niter.ncmp.1.not = icmp eq i32 %niter.next.1, %unroll_iter
  br i1 %niter.ncmp.1.not, label %._crit_edge.unr-lcssa.loopexit, label %__nv_fabsf.exit1

._crit_edge.unr-lcssa.loopexit:                   ; preds = %__nv_fabsf.exit1
  %phi.cast = zext i32 %26 to i64
  br label %._crit_edge.unr-lcssa

._crit_edge.unr-lcssa:                            ; preds = %._crit_edge.unr-lcssa.loopexit, %__nv_fabsf.exit1.lr.ph
  %.05.lcssa.ph = phi float [ undef, %__nv_fabsf.exit1.lr.ph ], [ %.05.1, %._crit_edge.unr-lcssa.loopexit ]
  %i.011.unr = phi i64 [ 1, %__nv_fabsf.exit1.lr.ph ], [ %phi.cast, %._crit_edge.unr-lcssa.loopexit ]
  %largest.010.unr = phi float [ %.03, %__nv_fabsf.exit1.lr.ph ], [ %.05.1, %._crit_edge.unr-lcssa.loopexit ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %._crit_edge, label %__nv_fabsf.exit1.epil.preheader

__nv_fabsf.exit1.epil.preheader:                  ; preds = %._crit_edge.unr-lcssa
  %27 = getelementptr inbounds float, float* %t, i64 %i.011.unr
  %28 = load float, float* %27, align 4
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5.epil = icmp eq i32 %29, 0
  %30 = call float @llvm.nvvm.fabs.ftz.f(float %28) #139
  %31 = call float @llvm.fabs.f32(float %28)
  %.04.epil = select i1 %.not5.epil, float %31, float %30
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6.epil = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fmax.ftz.f(float %.04.epil, float %largest.010.unr) #139
  %34 = call float @llvm.maxnum.f32(float %.04.epil, float %largest.010.unr)
  %.05.epil = select i1 %.not6.epil, float %34, float %33
  br label %._crit_edge

._crit_edge:                                      ; preds = %__nv_fabsf.exit1.epil.preheader, %._crit_edge.unr-lcssa, %__nv_fabsf.exit
  %largest.0.lcssa = phi float [ %.03, %__nv_fabsf.exit ], [ %.05.lcssa.ph, %._crit_edge.unr-lcssa ], [ %.05.epil, %__nv_fabsf.exit1.epil.preheader ]
  %35 = bitcast float %largest.0.lcssa to i32
  %36 = and i32 %35, -33554432
  %37 = sub i32 2122317824, %36
  %38 = bitcast i32 %37 to float
  %39 = fmul float %largest.0.lcssa, %38
  %40 = icmp sgt i32 %dim, 0
  br i1 %40, label %__nv_fabsf.exit4, label %__nv_fmaf_rn.exit

__nv_fabsf.exit4:                                 ; preds = %._crit_edge, %54
  %j.09 = phi i32 [ %j.1, %54 ], [ 0, %._crit_edge ]
  %i.18 = phi i32 [ %55, %54 ], [ 0, %._crit_edge ]
  %res0.07 = phi float [ %res0.1, %54 ], [ 0.000000e+00, %._crit_edge ]
  %41 = zext i32 %i.18 to i64
  %42 = getelementptr inbounds float, float* %t, i64 %41
  %43 = load float, float* %42, align 4
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fabs.ftz.f(float %43) #139
  %46 = call float @llvm.fabs.f32(float %43)
  %.01 = select i1 %.not3, float %46, float %45
  %47 = fmul float %.01, %38
  %48 = fcmp une float %47, %39
  %49 = icmp ne i32 %i.18, %j.09
  %or.cond = select i1 %48, i1 true, i1 %49
  br i1 %or.cond, label %__nv_fmaf_rn.exit5, label %54

__nv_fmaf_rn.exit5:                               ; preds = %__nv_fabsf.exit4
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float %47, float %47, float %res0.07) #139
  %52 = call float @llvm.fma.f32(float %47, float %47, float %res0.07)
  %.0 = select i1 %.not4, float %52, float %51
  %53 = add nsw i32 %j.09, 1
  br label %54

54:                                               ; preds = %__nv_fmaf_rn.exit5, %__nv_fabsf.exit4
  %res0.1 = phi float [ %.0, %__nv_fmaf_rn.exit5 ], [ %res0.07, %__nv_fabsf.exit4 ]
  %j.1 = phi i32 [ %53, %__nv_fmaf_rn.exit5 ], [ %j.09, %__nv_fabsf.exit4 ]
  %55 = add nuw nsw i32 %i.18, 1
  %exitcond.not = icmp eq i32 %55, %dim
  br i1 %exitcond.not, label %__nv_fmaf_rn.exit, label %__nv_fabsf.exit4

__nv_fmaf_rn.exit:                                ; preds = %54, %._crit_edge
  %res0.0.lcssa = phi float [ 0.000000e+00, %._crit_edge ], [ %res0.1, %54 ]
  %56 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %56, 0
  %57 = call float @llvm.nvvm.fma.rn.ftz.f(float %39, float %39, float %res0.0.lcssa) #139
  %58 = call float @llvm.fma.f32(float %39, float %39, float %res0.0.lcssa)
  %.02 = select i1 %.not1, float %58, float %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %59, 0
  br i1 %.not2, label %62, label %60

60:                                               ; preds = %__nv_fmaf_rn.exit
  %61 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.02) #139
  br label %__nv_rsqrtf.exit

62:                                               ; preds = %__nv_fmaf_rn.exit
  %63 = call float @llvm.nvvm.rsqrt.approx.f(float %.02) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %62, %60
  %.06 = phi float [ %61, %60 ], [ %63, %62 ]
  %64 = fmul float %.06, %38
  %65 = fcmp oeq float %largest.0.lcssa, 0x7FF0000000000000
  %res.0 = select i1 %65, float 0.000000e+00, float %64
  ret float %res.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_cbrtf(float %a) local_unnamed_addr #41 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  br i1 %.not1, label %6, label %4

4:                                                ; preds = %__nv_fabsf.exit
  %5 = call float @llvm.nvvm.lg2.approx.ftz.f(float %.01) #139
  br label %__nv_fast_log2f.exit

6:                                                ; preds = %__nv_fabsf.exit
  %7 = call float @llvm.nvvm.lg2.approx.f(float %.01) #139
  br label %__nv_fast_log2f.exit

__nv_fast_log2f.exit:                             ; preds = %6, %4
  %.02 = phi float [ %5, %4 ], [ %7, %6 ]
  %8 = fmul float %.02, 0x3FD5555560000000
  %9 = call float @llvm.nvvm.ex2.approx.ftz.f(float %8) #139
  %10 = fmul float %9, %9
  %11 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %10) #141, !srcloc !73
  %12 = fneg float %.01
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %13, 0
  %14 = call float @llvm.nvvm.fma.rn.ftz.f(float %11, float %12, float %9) #139
  %15 = call float @llvm.fma.f32(float %11, float %12, float %9)
  %.0 = select i1 %.not2, float %15, float %14
  %16 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %16, 0
  %17 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0, float noundef 0xBFD5555560000000, float %9) #139
  %18 = call float @llvm.fma.f32(float %.0, float noundef 0xBFD5555560000000, float %9)
  %.03 = select i1 %.not3, float %18, float %17
  %19 = fcmp olt float %a, 0.000000e+00
  %20 = fneg float %.03
  %t.0 = select i1 %19, float %20, float %.03
  %21 = fadd float %a, %a
  %22 = fcmp oeq float %21, %a
  %t.1 = select i1 %22, float %21, float %t.0
  ret float %t.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_rcbrtf(float %a) local_unnamed_addr #42 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %3, 0
  br i1 %.not1, label %6, label %4

4:                                                ; preds = %__nv_fabsf.exit
  %5 = call float @llvm.nvvm.lg2.approx.ftz.f(float %.01) #139
  br label %__nv_fast_log2f.exit

6:                                                ; preds = %__nv_fabsf.exit
  %7 = call float @llvm.nvvm.lg2.approx.f(float %.01) #139
  br label %__nv_fast_log2f.exit

__nv_fast_log2f.exit:                             ; preds = %6, %4
  %.02 = phi float [ %5, %4 ], [ %7, %6 ]
  %8 = fmul float %.02, 0xBFD5555560000000
  %9 = call float @llvm.nvvm.ex2.approx.ftz.f(float %8) #139
  %10 = fmul float %9, %9
  %11 = fneg float %.01
  %12 = fmul float %9, %11
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %13, 0
  %14 = call float @llvm.nvvm.fma.rn.ftz.f(float %10, float %12, float noundef 1.000000e+00) #139
  %15 = call float @llvm.fma.f32(float %10, float %12, float noundef 1.000000e+00)
  %.03 = select i1 %.not2, float %15, float %14
  %16 = fmul float %9, 0x3FD5555560000000
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %16, float %9) #139
  %19 = call float @llvm.fma.f32(float %.03, float %16, float %9)
  %.0 = select i1 %.not3, float %19, float %18
  %20 = fcmp olt float %a, 0.000000e+00
  %21 = fneg float %.0
  %t.0 = select i1 %20, float %21, float %.0
  %22 = fadd float %a, %a
  %23 = fcmp oeq float %22, %a
  br i1 %23, label %24, label %26

24:                                               ; preds = %__nv_fast_log2f.exit
  %25 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %a) #141, !srcloc !73
  br label %26

26:                                               ; preds = %24, %__nv_fast_log2f.exit
  %t.1 = phi float [ %25, %24 ], [ %t.0, %__nv_fast_log2f.exit ]
  ret float %t.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_j0f(float %a) local_unnamed_addr #43 {
__nv_fabsf.exit:
  %result.i.i.i = alloca [7 x i32], align 4
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp ugt float %.01, 8.000000e+00
  br i1 %3, label %__nv_isinff.exit, label %__internal_fmad.exit

__internal_fmad.exit:                             ; preds = %__nv_fabsf.exit
  %4 = fadd float %.01, 0xC0033D1520000000
  %5 = fadd float %4, 0xBE7D2E3660000000
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not39 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBCD6771CE0000000, float %5, float noundef 0xBD35953660000000) #139
  %8 = call float @llvm.fma.f32(float %5, float noundef 0xBCD6771CE0000000, float noundef 0xBD35953660000000)
  %.02 = select i1 %.not39, float %8, float %7
  %9 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not40 = icmp eq i32 %9, 0
  %10 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %5, float noundef 0x3D87E1C300000000) #139
  %11 = call float @llvm.fma.f32(float %.02, float %5, float noundef 0x3D87E1C300000000)
  %.03 = select i1 %.not40, float %11, float %10
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not41 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %5, float noundef 0xBD9A82F020000000) #139
  %14 = call float @llvm.fma.f32(float %.03, float %5, float noundef 0xBD9A82F020000000)
  %.05 = select i1 %.not41, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %5, float noundef 0xBDFD21E700000000) #139
  %17 = call float @llvm.fma.f32(float %.05, float %5, float noundef 0xBDFD21E700000000)
  %.06 = select i1 %.not42, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %5, float noundef 0x3E04060B60000000) #139
  %20 = call float @llvm.fma.f32(float %.06, float %5, float noundef 0x3E04060B60000000)
  %.07 = select i1 %.not43, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %5, float noundef 0x3E6F2E2860000000) #139
  %23 = call float @llvm.fma.f32(float %.07, float %5, float noundef 0x3E6F2E2860000000)
  %.08 = select i1 %.not44, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %5, float noundef 0x3E1EEDF0A0000000) #139
  %26 = call float @llvm.fma.f32(float %.08, float %5, float noundef 0x3E1EEDF0A0000000)
  %.09 = select i1 %.not45, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %5, float noundef 0xBED6DBF8C0000000) #139
  %29 = call float @llvm.fma.f32(float %.09, float %5, float noundef 0xBED6DBF8C0000000)
  %.010 = select i1 %.not46, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %5, float noundef 0xBEDECCB920000000) #139
  %32 = call float @llvm.fma.f32(float %.010, float %5, float noundef 0xBEDECCB920000000)
  %.012 = select i1 %.not47, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %5, float noundef 0x3F33C5BD60000000) #139
  %35 = call float @llvm.fma.f32(float %.012, float %5, float noundef 0x3F33C5BD60000000)
  %.013 = select i1 %.not48, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %5, float noundef 0x3F495C6680000000) #139
  %38 = call float @llvm.fma.f32(float %.013, float %5, float noundef 0x3F495C6680000000)
  %.014 = select i1 %.not49, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %5, float noundef 0xBF7DD54360000000) #139
  %41 = call float @llvm.fma.f32(float %.014, float %5, float noundef 0xBF7DD54360000000)
  %.015 = select i1 %.not50, float %41, float %40
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not51 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %5, float noundef 0xBF9B4EE8E0000000) #139
  %44 = call float @llvm.fma.f32(float %.015, float %5, float noundef 0xBF9B4EE8E0000000)
  %.016 = select i1 %.not51, float %44, float %43
  %45 = fmul float %5, %.016
  %46 = fadd float %.01, 0xC016148F60000000
  %47 = fadd float %46, 0x3E734F46E0000000
  %48 = fmul float %47, %45
  %49 = fadd float %.01, 0xC0214EB560000000
  %50 = fadd float %49, 0xBE9999BDA0000000
  %51 = fmul float %50, %48
  br label %215

__nv_isinff.exit:                                 ; preds = %__nv_fabsf.exit
  %52 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %52, 0
  %53 = call float @llvm.nvvm.fabs.ftz.f(float %.01) #139
  %54 = call float @llvm.fabs.f32(float %.01)
  %.017 = select i1 %.not1, float %54, float %53
  %55 = fcmp oeq float %.017, 0x7FF0000000000000
  br i1 %55, label %215, label %__internal_fmad.exit13

__internal_fmad.exit13:                           ; preds = %__nv_isinff.exit
  %56 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.01) #141, !srcloc !73
  %57 = fmul float %56, %56
  %58 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %58, 0
  %59 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400ADFD260000000, float %57, float noundef 0xBFE076F840000000) #139
  %60 = call float @llvm.fma.f32(float %57, float noundef 0x400ADFD260000000, float noundef 0xBFE076F840000000)
  %.019 = select i1 %.not2, float %60, float %59
  %61 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %61, 0
  %62 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %57, float noundef 0x3FBA767E60000000) #139
  %63 = call float @llvm.fma.f32(float %.019, float %57, float noundef 0x3FBA767E60000000)
  %.020 = select i1 %.not3, float %63, float %62
  %64 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %64, 0
  %65 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %57, float noundef 0xBFAFFFF6C0000000) #139
  %66 = call float @llvm.fma.f32(float %.020, float %57, float noundef 0xBFAFFFF6C0000000)
  %.021 = select i1 %.not4, float %66, float %65
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %67, 0
  %68 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %57, float noundef 1.000000e+00) #139
  %69 = call float @llvm.fma.f32(float %.021, float %57, float noundef 1.000000e+00)
  %.022 = select i1 %.not5, float %69, float %68
  %70 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %70, 0
  %71 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF23C0120000000, float %57, float noundef 0xBFCA4825A0000000) #139
  %72 = call float @llvm.fma.f32(float %57, float noundef 0x3FF23C0120000000, float noundef 0xBFCA4825A0000000)
  %.023 = select i1 %.not6, float %72, float %71
  %73 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %73, 0
  %74 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float %57, float noundef 0x3FB0A9DA20000000) #139
  %75 = call float @llvm.fma.f32(float %.023, float %57, float noundef 0x3FB0A9DA20000000)
  %.024 = select i1 %.not7, float %75, float %74
  %76 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %76, 0
  %77 = call float @llvm.nvvm.fma.rn.ftz.f(float %.024, float %57, float noundef 0xBFBFFFFFE0000000) #139
  %78 = call float @llvm.fma.f32(float %.024, float %57, float noundef 0xBFBFFFFFE0000000)
  %.026 = select i1 %.not8, float %78, float %77
  %79 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %79, 0
  %80 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %56, float %.01) #139
  %81 = call float @llvm.fma.f32(float %.026, float %56, float %.01)
  %.027 = select i1 %.not9, float %81, float %80
  %82 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %82, 0
  br i1 %.not10, label %85, label %83

83:                                               ; preds = %__internal_fmad.exit13
  %84 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.01) #139
  br label %__nv_rsqrtf.exit

85:                                               ; preds = %__internal_fmad.exit13
  %86 = call float @llvm.nvvm.rsqrt.approx.f(float %.01) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %85, %83
  %.028 = phi float [ %84, %83 ], [ %86, %85 ]
  %87 = fmul float %.028, 0x3FE9884540000000
  %88 = fmul float %.022, %87
  %89 = fmul float %.027, 0x3FE45F3060000000
  %90 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %90, 0
  %91 = call i32 @llvm.nvvm.f2i.rn.ftz(float %89) #139
  %92 = call i32 @llvm.nvvm.f2i.rn(float %89) #139
  %.030 = select i1 %.not11, i32 %92, i32 %91
  %93 = sitofp i32 %.030 to float
  %94 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %94, 0
  %95 = call float @llvm.nvvm.fma.rn.ftz.f(float %93, float noundef 0xBFF921FB40000000, float %.027) #139
  %96 = call float @llvm.fma.f32(float %93, float noundef 0xBFF921FB40000000, float %.027)
  %.031 = select i1 %.not12, float %96, float %95
  %97 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %97, 0
  %98 = call float @llvm.nvvm.fma.rn.ftz.f(float %93, float noundef 0xBE74442D00000000, float %.031) #139
  %99 = call float @llvm.fma.f32(float %93, float noundef 0xBE74442D00000000, float %.031)
  %.032 = select i1 %.not13, float %99, float %98
  %100 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %100, 0
  %101 = call float @llvm.nvvm.fma.rn.ftz.f(float %93, float noundef 0xBCF84698A0000000, float %.032) #139
  %102 = call float @llvm.fma.f32(float %93, float noundef 0xBCF84698A0000000, float %.032)
  %.033 = select i1 %.not14, float %102, float %101
  %103 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %103, 0
  %104 = call float @llvm.nvvm.fabs.ftz.f(float %.027) #139
  %105 = call float @llvm.fabs.f32(float %.027)
  %.035 = select i1 %.not15, float %105, float %104
  %106 = fcmp ult float %.035, 1.056150e+05
  br i1 %106, label %__internal_trig_reduction_kernel.exit.i, label %__nv_isinff.exit.i.i

__nv_isinff.exit.i.i:                             ; preds = %__nv_rsqrtf.exit
  %107 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %107, 0
  %.036 = select i1 %.not32, float %105, float %104
  %108 = fcmp oeq float %.036, 0x7FF0000000000000
  br i1 %108, label %__nv_fmul_rn.exit.i.i, label %112

__nv_fmul_rn.exit.i.i:                            ; preds = %__nv_isinff.exit.i.i
  %109 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not38 = icmp eq i32 %109, 0
  %110 = call float @llvm.nvvm.mul.rn.ftz.f(float %.027, float noundef 0.000000e+00) #139
  %111 = fmul float %.027, 0.000000e+00
  %.037 = select i1 %.not38, float %111, float %110
  br label %__internal_trig_reduction_kernel.exit.i

112:                                              ; preds = %__nv_isinff.exit.i.i
  %113 = bitcast float %.027 to i32
  %114 = lshr i32 %113, 23
  %115 = and i32 %114, 255
  %116 = add nsw i32 %115, -128
  %117 = shl i32 %113, 8
  %118 = or i32 %117, -2147483648
  %119 = lshr i32 %116, 5
  br label %120

120:                                              ; preds = %112, %120
  %iq.i.i.i.053 = phi i32 [ 0, %112 ], [ %128, %120 ]
  %hi.i.i.i.052 = phi i32 [ 0, %112 ], [ %126, %120 ]
  %121 = zext i32 %iq.i.i.i.053 to i64
  %122 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %121
  %123 = load i32, i32 addrspace(1)* %122, align 4
  %124 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %123, i32 %118, i32 %hi.i.i.i.052) #141, !srcloc !67
  %125 = extractvalue { i32, i32 } %124, 0
  %126 = extractvalue { i32, i32 } %124, 1
  %127 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %121
  store i32 %125, i32* %127, align 4, !noalias !68
  %128 = add nuw nsw i32 %iq.i.i.i.053, 1
  %exitcond.not = icmp eq i32 %128, 6
  br i1 %exitcond.not, label %129, label %120, !llvm.loop !71

129:                                              ; preds = %120
  %130 = and i32 %113, -2147483648
  %131 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 6
  store i32 %126, i32* %131, align 4, !noalias !68
  %132 = and i32 %114, 31
  %133 = sub nsw i32 6, %119
  %134 = sext i32 %133 to i64
  %135 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %134
  %136 = load i32, i32* %135, align 4
  %137 = sub nsw i32 5, %119
  %138 = sext i32 %137 to i64
  %139 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %138
  %140 = load i32, i32* %139, align 4
  %.not33 = icmp eq i32 %132, 0
  br i1 %.not33, label %__internal_trig_reduction_slowpath.exit.i.i, label %141

141:                                              ; preds = %129
  %142 = sub nsw i32 4, %119
  %143 = sub nuw nsw i32 32, %132
  %144 = shl i32 %136, %132
  %145 = lshr i32 %140, %143
  %146 = add i32 %145, %144
  %147 = shl i32 %140, %132
  %148 = sext i32 %142 to i64
  %149 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %148
  %150 = load i32, i32* %149, align 4
  %151 = lshr i32 %150, %143
  %152 = add i32 %151, %147
  br label %__internal_trig_reduction_slowpath.exit.i.i

__internal_trig_reduction_slowpath.exit.i.i:      ; preds = %141, %129
  %hi.i.i.i.1 = phi i32 [ %146, %141 ], [ %136, %129 ]
  %lo.i.i.i.0 = phi i32 [ %152, %141 ], [ %140, %129 ]
  %153 = lshr i32 %hi.i.i.i.1, 30
  %154 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.1, i32 %lo.i.i.i.0, i32 noundef 2)
  %155 = shl i32 %lo.i.i.i.0, 2
  %156 = lshr i32 %154, 31
  %157 = add nuw nsw i32 %156, %153
  %.not34 = icmp eq i32 %130, 0
  %158 = sub nsw i32 0, %157
  %spec.select = select i1 %.not34, i32 %157, i32 %158
  %.not35 = icmp sgt i32 %154, -1
  %159 = xor i32 %130, -2147483648
  %s.i.i.i.0 = select i1 %.not35, i32 %130, i32 %159
  %not..not35 = xor i1 %.not35, true
  %160 = sext i1 %not..not35 to i32
  %hi.i.i.i.2 = xor i32 %154, %160
  %lo.i.i.i.1 = xor i32 %155, %160
  %161 = zext i32 %hi.i.i.i.2 to i64
  %162 = shl nuw i64 %161, 32
  %163 = zext i32 %lo.i.i.i.1 to i64
  %164 = or i64 %162, %163
  %165 = sitofp i64 %164 to double
  %166 = fmul double %165, 0x3BF921FB54442D19
  %167 = fptrunc double %166 to float
  %.not37 = icmp eq i32 %s.i.i.i.0, 0
  %168 = fneg float %167
  %r.i.i.i.0 = select i1 %.not37, float %167, float %168
  br label %__internal_trig_reduction_kernel.exit.i

__internal_trig_reduction_kernel.exit.i:          ; preds = %__nv_fmul_rn.exit.i.i, %__internal_trig_reduction_slowpath.exit.i.i, %__nv_rsqrtf.exit
  %i.i.1 = phi i32 [ %.030, %__nv_rsqrtf.exit ], [ 0, %__nv_fmul_rn.exit.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i ]
  %t.i.i.1 = phi float [ %.033, %__nv_rsqrtf.exit ], [ %.037, %__nv_fmul_rn.exit.i.i ], [ %r.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i ]
  %169 = and i32 %i.i.1, 3
  %170 = uitofp i32 %169 to float
  %171 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %171, 0
  %172 = call float @llvm.nvvm.fma.rn.ftz.f(float %170, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000) #139
  %173 = call float @llvm.fma.f32(float %170, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000)
  %.040 = select i1 %.not16, float %173, float %172
  %174 = fadd float %t.i.i.1, %.040
  %175 = fmul float %174, 0x3FE45F3060000000
  %176 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %176, 0
  %177 = call i32 @llvm.nvvm.f2i.rn.ftz(float %175) #139
  %178 = call i32 @llvm.nvvm.f2i.rn(float %175) #139
  %.041 = select i1 %.not17, i32 %178, i32 %177
  %179 = sitofp i32 %.041 to float
  %180 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %180, 0
  %181 = call float @llvm.nvvm.fma.rn.ftz.f(float %179, float noundef 0xBFF921FB40000000, float %174) #139
  %182 = call float @llvm.fma.f32(float %179, float noundef 0xBFF921FB40000000, float %174)
  %.039 = select i1 %.not18, float %182, float %181
  %183 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %183, 0
  %184 = call float @llvm.nvvm.fma.rn.ftz.f(float %179, float noundef 0xBE74442D00000000, float %.039) #139
  %185 = call float @llvm.fma.f32(float %179, float noundef 0xBE74442D00000000, float %.039)
  %.038 = select i1 %.not19, float %185, float %184
  %186 = add i32 %.041, 1
  %187 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %187, 0
  %188 = call float @llvm.nvvm.mul.rn.ftz.f(float %.038, float %.038) #139
  %189 = fmul float %.038, %.038
  %.034 = select i1 %.not20, float %189, float %188
  %190 = and i32 %186, 1
  %.not21 = icmp eq i32 %190, 0
  %191 = select i1 %.not21, float %.038, float 1.000000e+00
  %192 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %192, 0
  %193 = call float @llvm.nvvm.fma.rn.ftz.f(float %.034, float %191, float noundef 0.000000e+00) #139
  %194 = call float @llvm.fma.f32(float %.034, float %191, float noundef 0.000000e+00)
  %.029 = select i1 %.not22, float %194, float %193
  br i1 %.not21, label %__internal_fmad.exit2.i.i, label %__internal_fmad.exit1.i.i

__internal_fmad.exit1.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i
  %195 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %195, 0
  %196 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.034, float noundef 0xBF56C0FDA0000000) #139
  %197 = call float @llvm.fma.f32(float %.034, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.025 = select i1 %.not31, float %197, float %196
  br label %__internal_fmad.exit2.i.i

__internal_fmad.exit2.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i, %__internal_fmad.exit1.i.i
  %198 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i ]
  %199 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i ]
  %200 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i ], [ %.025, %__internal_fmad.exit1.i.i ]
  %201 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %201, 0
  %202 = call float @llvm.nvvm.fma.rn.ftz.f(float %200, float %.034, float noundef %199) #139
  %203 = call float @llvm.fma.f32(float %200, float %.034, float noundef %199)
  %.018 = select i1 %.not25, float %203, float %202
  %204 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %204, 0
  %205 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %.034, float noundef %198) #139
  %206 = call float @llvm.fma.f32(float %.018, float %.034, float noundef %198)
  %.011 = select i1 %.not27, float %206, float %205
  %207 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %207, 0
  %208 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %.029, float %191) #139
  %209 = call float @llvm.fma.f32(float %.011, float %.029, float %191)
  %.04 = select i1 %.not28, float %209, float %208
  %210 = and i32 %186, 2
  %.not29 = icmp eq i32 %210, 0
  br i1 %.not29, label %__internal_cos_offset_f.exit, label %__internal_fmad.exit5.i.i

__internal_fmad.exit5.i.i:                        ; preds = %__internal_fmad.exit2.i.i
  %211 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %211, 0
  %212 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %213 = call float @llvm.fma.f32(float %.04, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0 = select i1 %.not30, float %213, float %212
  br label %__internal_cos_offset_f.exit

__internal_cos_offset_f.exit:                     ; preds = %__internal_fmad.exit5.i.i, %__internal_fmad.exit2.i.i
  %z.i.i.0 = phi float [ %.0, %__internal_fmad.exit5.i.i ], [ %.04, %__internal_fmad.exit2.i.i ]
  %214 = fmul float %88, %z.i.i.0
  br label %215

215:                                              ; preds = %__internal_cos_offset_f.exit, %__nv_isinff.exit, %__internal_fmad.exit
  %r.1 = phi float [ %51, %__internal_fmad.exit ], [ %214, %__internal_cos_offset_f.exit ], [ 0.000000e+00, %__nv_isinff.exit ]
  ret float %r.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_j1f(float %a) local_unnamed_addr #44 {
__nv_fabsf.exit:
  %result.i.i.i = alloca [7 x i32], align 4
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = fcmp ugt float %.0, 8.000000e+00
  br i1 %3, label %__nv_isinff.exit, label %__internal_fmad.exit

__internal_fmad.exit:                             ; preds = %__nv_fabsf.exit
  %4 = fadd float %.0, 0xC00EA75580000000
  %5 = fadd float %4, 0x3E74A121E0000000
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not39 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3D35E68C60000000, float %5, float noundef 0x3D7037E840000000) #139
  %8 = call float @llvm.fma.f32(float %5, float noundef 0x3D35E68C60000000, float noundef 0x3D7037E840000000)
  %.01 = select i1 %.not39, float %8, float %7
  %9 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not40 = icmp eq i32 %9, 0
  %10 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %5, float noundef 0xBDBC43D820000000) #139
  %11 = call float @llvm.fma.f32(float %.01, float %5, float noundef 0xBDBC43D820000000)
  %.02 = select i1 %.not40, float %11, float %10
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not41 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %5, float noundef 0xBDEBBBDFE0000000) #139
  %14 = call float @llvm.fma.f32(float %.02, float %5, float noundef 0xBDEBBBDFE0000000)
  %.03 = select i1 %.not41, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %5, float noundef 0x3E336193A0000000) #139
  %17 = call float @llvm.fma.f32(float %.03, float %5, float noundef 0x3E336193A0000000)
  %.04 = select i1 %.not42, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %5, float noundef 0x3E5D022E60000000) #139
  %20 = call float @llvm.fma.f32(float %.04, float %5, float noundef 0x3E5D022E60000000)
  %.06 = select i1 %.not43, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %5, float noundef 0xBEA1F1B900000000) #139
  %23 = call float @llvm.fma.f32(float %.06, float %5, float noundef 0xBEA1F1B900000000)
  %.07 = select i1 %.not44, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %5, float noundef 0xBEC3CCA7A0000000) #139
  %26 = call float @llvm.fma.f32(float %.07, float %5, float noundef 0xBEC3CCA7A0000000)
  %.08 = select i1 %.not45, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %5, float noundef 0x3F059B38A0000000) #139
  %29 = call float @llvm.fma.f32(float %.08, float %5, float noundef 0x3F059B38A0000000)
  %.09 = select i1 %.not46, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %5, float noundef 0x3F1F3D6200000000) #139
  %32 = call float @llvm.fma.f32(float %.09, float %5, float noundef 0x3F1F3D6200000000)
  %.010 = select i1 %.not47, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %5, float noundef 0xBF5D9D7380000000) #139
  %35 = call float @llvm.fma.f32(float %.010, float %5, float noundef 0xBF5D9D7380000000)
  %.011 = select i1 %.not48, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %5, float noundef 0xBF64EDFFA0000000) #139
  %38 = call float @llvm.fma.f32(float %.011, float %5, float noundef 0xBF64EDFFA0000000)
  %.013 = select i1 %.not49, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %5, float noundef 0x3FA0E73260000000) #139
  %41 = call float @llvm.fma.f32(float %.013, float %5, float noundef 0x3FA0E73260000000)
  %.014 = select i1 %.not50, float %41, float %40
  %42 = fadd float %.0, 0xC01C0FF600000000
  %43 = fadd float %42, 0x3E88971B60000000
  %44 = fmul float %43, %.014
  %45 = fmul float %5, %44
  %46 = fmul float %.0, %45
  br label %210

__nv_isinff.exit:                                 ; preds = %__nv_fabsf.exit
  %47 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %47, 0
  %48 = call float @llvm.nvvm.fabs.ftz.f(float %.0) #139
  %49 = call float @llvm.fabs.f32(float %.0)
  %.015 = select i1 %.not1, float %49, float %48
  %50 = fcmp oeq float %.015, 0x7FF0000000000000
  br i1 %50, label %210, label %__internal_fmad.exit12

__internal_fmad.exit12:                           ; preds = %__nv_isinff.exit
  %51 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0) #141, !srcloc !73
  %52 = fmul float %51, %51
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC0105966E0000000, float %52, float noundef 0x3FE7FEFD20000000) #139
  %55 = call float @llvm.fma.f32(float %52, float noundef 0xC0105966E0000000, float noundef 0x3FE7FEFD20000000)
  %.016 = select i1 %.not2, float %55, float %54
  %56 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %56, 0
  %57 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %52, float noundef 0xBFC8B175C0000000) #139
  %58 = call float @llvm.fma.f32(float %.016, float %52, float noundef 0xBFC8B175C0000000)
  %.017 = select i1 %.not3, float %58, float %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %52, float noundef 0x3FC7FFF160000000) #139
  %61 = call float @llvm.fma.f32(float %.017, float %52, float noundef 0x3FC7FFF160000000)
  %.018 = select i1 %.not4, float %61, float %60
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %52, float noundef 1.000000e+00) #139
  %64 = call float @llvm.fma.f32(float %.018, float %52, float noundef 1.000000e+00)
  %.020 = select i1 %.not5, float %64, float %63
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFF9477440000000, float %52, float noundef 0x3FD72295A0000000) #139
  %67 = call float @llvm.fma.f32(float %52, float noundef 0xBFF9477440000000, float noundef 0x3FD72295A0000000)
  %.021 = select i1 %.not6, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %52, float noundef 0xBFC4FE5D80000000) #139
  %70 = call float @llvm.fma.f32(float %.021, float %52, float noundef 0xBFC4FE5D80000000)
  %.022 = select i1 %.not7, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rn.ftz.f(float %.022, float %52, float noundef 0x3FD7FFFFA0000000) #139
  %73 = call float @llvm.fma.f32(float %.022, float %52, float noundef 0x3FD7FFFFA0000000)
  %.023 = select i1 %.not8, float %73, float %72
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %74, 0
  %75 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float %51, float %.0) #139
  %76 = call float @llvm.fma.f32(float %.023, float %51, float %.0)
  %.024 = select i1 %.not9, float %76, float %75
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %77, 0
  br i1 %.not10, label %80, label %78

78:                                               ; preds = %__internal_fmad.exit12
  %79 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0) #139
  br label %__nv_rsqrtf.exit

80:                                               ; preds = %__internal_fmad.exit12
  %81 = call float @llvm.nvvm.rsqrt.approx.f(float %.0) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %80, %78
  %.025 = phi float [ %79, %78 ], [ %81, %80 ]
  %82 = fmul float %.025, 0x3FE9884540000000
  %83 = fmul float %.020, %82
  %84 = fmul float %.024, 0x3FE45F3060000000
  %85 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %85, 0
  %86 = call i32 @llvm.nvvm.f2i.rn.ftz(float %84) #139
  %87 = call i32 @llvm.nvvm.f2i.rn(float %84) #139
  %.027 = select i1 %.not11, i32 %87, i32 %86
  %88 = sitofp i32 %.027 to float
  %89 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %89, 0
  %90 = call float @llvm.nvvm.fma.rn.ftz.f(float %88, float noundef 0xBFF921FB40000000, float %.024) #139
  %91 = call float @llvm.fma.f32(float %88, float noundef 0xBFF921FB40000000, float %.024)
  %.029 = select i1 %.not12, float %91, float %90
  %92 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %92, 0
  %93 = call float @llvm.nvvm.fma.rn.ftz.f(float %88, float noundef 0xBE74442D00000000, float %.029) #139
  %94 = call float @llvm.fma.f32(float %88, float noundef 0xBE74442D00000000, float %.029)
  %.030 = select i1 %.not13, float %94, float %93
  %95 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %95, 0
  %96 = call float @llvm.nvvm.fma.rn.ftz.f(float %88, float noundef 0xBCF84698A0000000, float %.030) #139
  %97 = call float @llvm.fma.f32(float %88, float noundef 0xBCF84698A0000000, float %.030)
  %.031 = select i1 %.not14, float %97, float %96
  %98 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %98, 0
  %99 = call float @llvm.nvvm.fabs.ftz.f(float %.024) #139
  %100 = call float @llvm.fabs.f32(float %.024)
  %.032 = select i1 %.not15, float %100, float %99
  %101 = fcmp ult float %.032, 1.056150e+05
  br i1 %101, label %__internal_trig_reduction_kernel.exit.i, label %__nv_isinff.exit.i.i

__nv_isinff.exit.i.i:                             ; preds = %__nv_rsqrtf.exit
  %102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %102, 0
  %.033 = select i1 %.not32, float %100, float %99
  %103 = fcmp oeq float %.033, 0x7FF0000000000000
  br i1 %103, label %__nv_fmul_rn.exit.i.i, label %107

__nv_fmul_rn.exit.i.i:                            ; preds = %__nv_isinff.exit.i.i
  %104 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not38 = icmp eq i32 %104, 0
  %105 = call float @llvm.nvvm.mul.rn.ftz.f(float %.024, float noundef 0.000000e+00) #139
  %106 = fmul float %.024, 0.000000e+00
  %.034 = select i1 %.not38, float %106, float %105
  br label %__internal_trig_reduction_kernel.exit.i

107:                                              ; preds = %__nv_isinff.exit.i.i
  %108 = bitcast float %.024 to i32
  %109 = lshr i32 %108, 23
  %110 = and i32 %109, 255
  %111 = add nsw i32 %110, -128
  %112 = shl i32 %108, 8
  %113 = or i32 %112, -2147483648
  %114 = lshr i32 %111, 5
  br label %115

115:                                              ; preds = %107, %115
  %iq.i.i.i.052 = phi i32 [ 0, %107 ], [ %123, %115 ]
  %hi.i.i.i.051 = phi i32 [ 0, %107 ], [ %121, %115 ]
  %116 = zext i32 %iq.i.i.i.052 to i64
  %117 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %116
  %118 = load i32, i32 addrspace(1)* %117, align 4
  %119 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %118, i32 %113, i32 %hi.i.i.i.051) #141, !srcloc !67
  %120 = extractvalue { i32, i32 } %119, 0
  %121 = extractvalue { i32, i32 } %119, 1
  %122 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %116
  store i32 %120, i32* %122, align 4, !noalias !68
  %123 = add nuw nsw i32 %iq.i.i.i.052, 1
  %exitcond.not = icmp eq i32 %123, 6
  br i1 %exitcond.not, label %124, label %115, !llvm.loop !71

124:                                              ; preds = %115
  %125 = and i32 %108, -2147483648
  %126 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 6
  store i32 %121, i32* %126, align 4, !noalias !68
  %127 = and i32 %109, 31
  %128 = sub nsw i32 6, %114
  %129 = sext i32 %128 to i64
  %130 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %129
  %131 = load i32, i32* %130, align 4
  %132 = sub nsw i32 5, %114
  %133 = sext i32 %132 to i64
  %134 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %133
  %135 = load i32, i32* %134, align 4
  %.not33 = icmp eq i32 %127, 0
  br i1 %.not33, label %__internal_trig_reduction_slowpath.exit.i.i, label %136

136:                                              ; preds = %124
  %137 = sub nsw i32 4, %114
  %138 = sub nuw nsw i32 32, %127
  %139 = shl i32 %131, %127
  %140 = lshr i32 %135, %138
  %141 = add i32 %140, %139
  %142 = shl i32 %135, %127
  %143 = sext i32 %137 to i64
  %144 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i, i64 0, i64 %143
  %145 = load i32, i32* %144, align 4
  %146 = lshr i32 %145, %138
  %147 = add i32 %146, %142
  br label %__internal_trig_reduction_slowpath.exit.i.i

__internal_trig_reduction_slowpath.exit.i.i:      ; preds = %136, %124
  %hi.i.i.i.1 = phi i32 [ %141, %136 ], [ %131, %124 ]
  %lo.i.i.i.0 = phi i32 [ %147, %136 ], [ %135, %124 ]
  %148 = lshr i32 %hi.i.i.i.1, 30
  %149 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.1, i32 %lo.i.i.i.0, i32 noundef 2)
  %150 = shl i32 %lo.i.i.i.0, 2
  %151 = lshr i32 %149, 31
  %152 = add nuw nsw i32 %151, %148
  %.not34 = icmp eq i32 %125, 0
  %153 = sub nsw i32 0, %152
  %spec.select = select i1 %.not34, i32 %152, i32 %153
  %.not35 = icmp sgt i32 %149, -1
  %154 = xor i32 %125, -2147483648
  %s.i.i.i.0 = select i1 %.not35, i32 %125, i32 %154
  %not..not35 = xor i1 %.not35, true
  %155 = sext i1 %not..not35 to i32
  %hi.i.i.i.2 = xor i32 %149, %155
  %lo.i.i.i.1 = xor i32 %150, %155
  %156 = zext i32 %hi.i.i.i.2 to i64
  %157 = shl nuw i64 %156, 32
  %158 = zext i32 %lo.i.i.i.1 to i64
  %159 = or i64 %157, %158
  %160 = sitofp i64 %159 to double
  %161 = fmul double %160, 0x3BF921FB54442D19
  %162 = fptrunc double %161 to float
  %.not37 = icmp eq i32 %s.i.i.i.0, 0
  %163 = fneg float %162
  %r.i.i.i.0 = select i1 %.not37, float %162, float %163
  br label %__internal_trig_reduction_kernel.exit.i

__internal_trig_reduction_kernel.exit.i:          ; preds = %__nv_fmul_rn.exit.i.i, %__internal_trig_reduction_slowpath.exit.i.i, %__nv_rsqrtf.exit
  %i.i.1 = phi i32 [ %.027, %__nv_rsqrtf.exit ], [ 0, %__nv_fmul_rn.exit.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i ]
  %t.i.i.1 = phi float [ %.031, %__nv_rsqrtf.exit ], [ %.034, %__nv_fmul_rn.exit.i.i ], [ %r.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i ]
  %164 = and i32 %i.i.1, 3
  %165 = uitofp i32 %164 to float
  %166 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %166, 0
  %167 = call float @llvm.nvvm.fma.rn.ftz.f(float %165, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %168 = call float @llvm.fma.f32(float %165, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.037 = select i1 %.not16, float %168, float %167
  %169 = fadd float %t.i.i.1, %.037
  %170 = fmul float %169, 0x3FE45F3060000000
  %171 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %171, 0
  %172 = call i32 @llvm.nvvm.f2i.rn.ftz(float %170) #139
  %173 = call i32 @llvm.nvvm.f2i.rn(float %170) #139
  %.038 = select i1 %.not17, i32 %173, i32 %172
  %174 = sitofp i32 %.038 to float
  %175 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %175, 0
  %176 = call float @llvm.nvvm.fma.rn.ftz.f(float %174, float noundef 0xBFF921FB40000000, float %169) #139
  %177 = call float @llvm.fma.f32(float %174, float noundef 0xBFF921FB40000000, float %169)
  %.039 = select i1 %.not18, float %177, float %176
  %178 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %178, 0
  %179 = call float @llvm.nvvm.fma.rn.ftz.f(float %174, float noundef 0xBE74442D00000000, float %.039) #139
  %180 = call float @llvm.fma.f32(float %174, float noundef 0xBE74442D00000000, float %.039)
  %.040 = select i1 %.not19, float %180, float %179
  %181 = add i32 %.038, 1
  %182 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %182, 0
  %183 = call float @llvm.nvvm.mul.rn.ftz.f(float %.040, float %.040) #139
  %184 = fmul float %.040, %.040
  %.036 = select i1 %.not20, float %184, float %183
  %185 = and i32 %181, 1
  %.not21 = icmp eq i32 %185, 0
  %186 = select i1 %.not21, float %.040, float 1.000000e+00
  %187 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %187, 0
  %188 = call float @llvm.nvvm.fma.rn.ftz.f(float %.036, float %186, float noundef 0.000000e+00) #139
  %189 = call float @llvm.fma.f32(float %.036, float %186, float noundef 0.000000e+00)
  %.035 = select i1 %.not22, float %189, float %188
  br i1 %.not21, label %__internal_fmad.exit2.i.i, label %__internal_fmad.exit1.i.i

__internal_fmad.exit1.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i
  %190 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %190, 0
  %191 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.036, float noundef 0xBF56C0FDA0000000) #139
  %192 = call float @llvm.fma.f32(float %.036, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.028 = select i1 %.not31, float %192, float %191
  br label %__internal_fmad.exit2.i.i

__internal_fmad.exit2.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i, %__internal_fmad.exit1.i.i
  %193 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i ]
  %194 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i ]
  %195 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i ], [ %.028, %__internal_fmad.exit1.i.i ]
  %196 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %196, 0
  %197 = call float @llvm.nvvm.fma.rn.ftz.f(float %195, float %.036, float noundef %194) #139
  %198 = call float @llvm.fma.f32(float %195, float %.036, float noundef %194)
  %.026 = select i1 %.not25, float %198, float %197
  %199 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %199, 0
  %200 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %.036, float noundef %193) #139
  %201 = call float @llvm.fma.f32(float %.026, float %.036, float noundef %193)
  %.019 = select i1 %.not27, float %201, float %200
  %202 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %202, 0
  %203 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %.035, float %186) #139
  %204 = call float @llvm.fma.f32(float %.019, float %.035, float %186)
  %.012 = select i1 %.not28, float %204, float %203
  %205 = and i32 %181, 2
  %.not29 = icmp eq i32 %205, 0
  br i1 %.not29, label %__internal_cos_offset_f.exit, label %__internal_fmad.exit5.i.i

__internal_fmad.exit5.i.i:                        ; preds = %__internal_fmad.exit2.i.i
  %206 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %206, 0
  %207 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %208 = call float @llvm.fma.f32(float %.012, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.05 = select i1 %.not30, float %208, float %207
  br label %__internal_cos_offset_f.exit

__internal_cos_offset_f.exit:                     ; preds = %__internal_fmad.exit5.i.i, %__internal_fmad.exit2.i.i
  %z.i.i.0 = phi float [ %.05, %__internal_fmad.exit5.i.i ], [ %.012, %__internal_fmad.exit2.i.i ]
  %209 = fmul float %83, %z.i.i.0
  br label %210

210:                                              ; preds = %__internal_cos_offset_f.exit, %__nv_isinff.exit, %__internal_fmad.exit
  %r.1 = phi float [ %46, %__internal_fmad.exit ], [ %209, %__internal_cos_offset_f.exit ], [ 0.000000e+00, %__nv_isinff.exit ]
  %211 = fcmp olt float %a, 0.000000e+00
  %212 = fneg float %r.1
  %r.2 = select i1 %211, float %212, float %r.1
  %213 = fcmp olt float %.0, 0x39B4484C00000000
  br i1 %213, label %214, label %221

214:                                              ; preds = %210
  %215 = bitcast float %a to i32
  %216 = and i32 %215, -2147483648
  %217 = bitcast float %r.2 to i32
  %218 = and i32 %217, 2147483647
  %219 = or i32 %218, %216
  %220 = bitcast i32 %219 to float
  br label %221

221:                                              ; preds = %214, %210
  %r.3 = phi float [ %220, %214 ], [ %r.2, %210 ]
  ret float %r.3
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_y0f(float %a) local_unnamed_addr #45 {
__nv_fabsf.exit:
  %result.i.i.i.i = alloca [7 x i32], align 4
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp ugt float %.01, 0x3FDC982EC0000000
  br i1 %3, label %287, label %__internal_fmad.exit

__internal_fmad.exit:                             ; preds = %__nv_fabsf.exit
  %4 = fmul float %.01, %.01
  %5 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not68 = icmp eq i32 %5, 0
  %6 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E7B7CB580000000, float %4, float noundef 0xBEE3E936C0000000) #139
  %7 = call float @llvm.fma.f32(float %4, float noundef 0x3E7B7CB580000000, float noundef 0xBEE3E936C0000000)
  %.02 = select i1 %.not68, float %7, float %6
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not69 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %4, float noundef 0x3F41A62000000000) #139
  %10 = call float @llvm.fma.f32(float %.02, float %4, float noundef 0x3F41A62000000000)
  %.03 = select i1 %.not69, float %10, float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not70 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %4, float noundef 0xBF9075B1C0000000) #139
  %13 = call float @llvm.fma.f32(float %.03, float %4, float noundef 0xBF9075B1C0000000)
  %.05 = select i1 %.not70, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not71 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %4, float noundef 0x3FC6BBCB40000000) #139
  %16 = call float @llvm.fma.f32(float %.05, float %4, float noundef 0x3FC6BBCB40000000)
  %.06 = select i1 %.not71, float %16, float %15
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not72 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %4, float noundef 0xBFB2E4D6A0000000) #139
  %19 = call float @llvm.fma.f32(float %.06, float %4, float noundef 0xBFB2E4D6A0000000)
  %.07 = select i1 %.not72, float %19, float %18
  %20 = fcmp olt float %.01, 0x3810000000000000
  %21 = fmul float %.01, 0x4160000000000000
  %.09 = select i1 %20, float %21, float %.01
  %i.i.0 = select i1 %20, float -2.300000e+01, float 0.000000e+00
  %22 = bitcast float %.09 to i32
  %23 = add i32 %22, -1059760811
  %24 = and i32 %23, -8388608
  %25 = sub i32 %22, %24
  %26 = bitcast i32 %25 to float
  %27 = sitofp i32 %24 to float
  %28 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not73 = icmp eq i32 %28, 0
  %29 = call float @llvm.nvvm.fma.rn.ftz.f(float %27, float noundef 0x3E80000000000000, float noundef %i.i.0) #139
  %30 = call float @llvm.fma.f32(float %27, float noundef 0x3E80000000000000, float noundef %i.i.0)
  %.014 = select i1 %.not73, float %30, float %29
  %31 = fadd float %26, -1.000000e+00
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not74 = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %31, float noundef 0x3FC2073EC0000000) #139
  %34 = call float @llvm.fma.f32(float %31, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.016 = select i1 %.not74, float %34, float %33
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not75 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %31, float noundef 0xBFBF19B980000000) #139
  %37 = call float @llvm.fma.f32(float %.016, float %31, float noundef 0xBFBF19B980000000)
  %.017 = select i1 %.not75, float %37, float %36
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not76 = icmp eq i32 %38, 0
  %39 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %31, float noundef 0x3FC1E52AA0000000) #139
  %40 = call float @llvm.fma.f32(float %.017, float %31, float noundef 0x3FC1E52AA0000000)
  %.018 = select i1 %.not76, float %40, float %39
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not77 = icmp eq i32 %41, 0
  %42 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %31, float noundef 0xBFC55B1720000000) #139
  %43 = call float @llvm.fma.f32(float %.018, float %31, float noundef 0xBFC55B1720000000)
  %.019 = select i1 %.not77, float %43, float %42
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not78 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %31, float noundef 0x3FC99DA160000000) #139
  %46 = call float @llvm.fma.f32(float %.019, float %31, float noundef 0x3FC99DA160000000)
  %.020 = select i1 %.not78, float %46, float %45
  %47 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not79 = icmp eq i32 %47, 0
  %48 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %31, float noundef 0xBFCFFFE440000000) #139
  %49 = call float @llvm.fma.f32(float %.020, float %31, float noundef 0xBFCFFFE440000000)
  %.021 = select i1 %.not79, float %49, float %48
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not80 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %31, float noundef 0x3FD5554F00000000) #139
  %52 = call float @llvm.fma.f32(float %.021, float %31, float noundef 0x3FD5554F00000000)
  %.023 = select i1 %.not80, float %52, float %51
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not81 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float %31, float noundef -5.000000e-01) #139
  %55 = call float @llvm.fma.f32(float %.023, float %31, float noundef -5.000000e-01)
  %.013 = select i1 %.not81, float %55, float %54
  %56 = fmul float %31, %.013
  %57 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not82 = icmp eq i32 %57, 0
  %58 = call float @llvm.nvvm.fma.rn.ftz.f(float %56, float %31, float %31) #139
  %59 = call float @llvm.fma.f32(float %56, float %31, float %31)
  %.012 = select i1 %.not82, float %59, float %58
  %60 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not83 = icmp eq i32 %60, 0
  %61 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float noundef 0x3FE62E4300000000, float %.012) #139
  %62 = call float @llvm.fma.f32(float %.014, float noundef 0x3FE62E4300000000, float %.012)
  %.011 = select i1 %.not83, float %62, float %61
  %63 = icmp ugt i32 %22, 2139095039
  br i1 %63, label %__nv_fmaf_rn.exit.i, label %__internal_accurate_logf.exit

__nv_fmaf_rn.exit.i:                              ; preds = %__internal_fmad.exit
  %64 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not137 = icmp eq i32 %64, 0
  %65 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %66 = call float @llvm.fma.f32(float %.09, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.010 = select i1 %.not137, float %66, float %65
  br label %__internal_accurate_logf.exit

__internal_accurate_logf.exit:                    ; preds = %__nv_fmaf_rn.exit.i, %__internal_fmad.exit
  %r.i.0 = phi float [ %.010, %__nv_fmaf_rn.exit.i ], [ %.011, %__internal_fmad.exit ]
  %67 = fcmp oeq float %.09, 0.000000e+00
  %r.i.0.op = fmul float %r.i.0, 0x3FE45F3060000000
  %68 = select i1 %67, float 0xFFF0000000000000, float %r.i.0.op
  %69 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not84 = icmp eq i32 %69, 0
  %70 = call float @llvm.nvvm.fabs.ftz.f(float %.01) #139
  %71 = call float @llvm.fabs.f32(float %.01)
  %.024 = select i1 %.not84, float %71, float %70
  %72 = fcmp ugt float %.024, 8.000000e+00
  br i1 %72, label %__nv_isinff.exit.i, label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %__internal_accurate_logf.exit
  %73 = fadd float %.024, 0xC0033D1520000000
  %74 = fadd float %73, 0xBE7D2E3660000000
  %75 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not124 = icmp eq i32 %75, 0
  %76 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBCD6771CE0000000, float %74, float noundef 0xBD35953660000000) #139
  %77 = call float @llvm.fma.f32(float %74, float noundef 0xBCD6771CE0000000, float noundef 0xBD35953660000000)
  %.025 = select i1 %.not124, float %77, float %76
  %78 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not125 = icmp eq i32 %78, 0
  %79 = call float @llvm.nvvm.fma.rn.ftz.f(float %.025, float %74, float noundef 0x3D87E1C300000000) #139
  %80 = call float @llvm.fma.f32(float %.025, float %74, float noundef 0x3D87E1C300000000)
  %.027 = select i1 %.not125, float %80, float %79
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not126 = icmp eq i32 %81, 0
  %82 = call float @llvm.nvvm.fma.rn.ftz.f(float %.027, float %74, float noundef 0xBD9A82F020000000) #139
  %83 = call float @llvm.fma.f32(float %.027, float %74, float noundef 0xBD9A82F020000000)
  %.028 = select i1 %.not126, float %83, float %82
  %84 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not127 = icmp eq i32 %84, 0
  %85 = call float @llvm.nvvm.fma.rn.ftz.f(float %.028, float %74, float noundef 0xBDFD21E700000000) #139
  %86 = call float @llvm.fma.f32(float %.028, float %74, float noundef 0xBDFD21E700000000)
  %.029 = select i1 %.not127, float %86, float %85
  %87 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not128 = icmp eq i32 %87, 0
  %88 = call float @llvm.nvvm.fma.rn.ftz.f(float %.029, float %74, float noundef 0x3E04060B60000000) #139
  %89 = call float @llvm.fma.f32(float %.029, float %74, float noundef 0x3E04060B60000000)
  %.030 = select i1 %.not128, float %89, float %88
  %90 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not129 = icmp eq i32 %90, 0
  %91 = call float @llvm.nvvm.fma.rn.ftz.f(float %.030, float %74, float noundef 0x3E6F2E2860000000) #139
  %92 = call float @llvm.fma.f32(float %.030, float %74, float noundef 0x3E6F2E2860000000)
  %.031 = select i1 %.not129, float %92, float %91
  %93 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not130 = icmp eq i32 %93, 0
  %94 = call float @llvm.nvvm.fma.rn.ftz.f(float %.031, float %74, float noundef 0x3E1EEDF0A0000000) #139
  %95 = call float @llvm.fma.f32(float %.031, float %74, float noundef 0x3E1EEDF0A0000000)
  %.032 = select i1 %.not130, float %95, float %94
  %96 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not131 = icmp eq i32 %96, 0
  %97 = call float @llvm.nvvm.fma.rn.ftz.f(float %.032, float %74, float noundef 0xBED6DBF8C0000000) #139
  %98 = call float @llvm.fma.f32(float %.032, float %74, float noundef 0xBED6DBF8C0000000)
  %.034 = select i1 %.not131, float %98, float %97
  %99 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not132 = icmp eq i32 %99, 0
  %100 = call float @llvm.nvvm.fma.rn.ftz.f(float %.034, float %74, float noundef 0xBEDECCB920000000) #139
  %101 = call float @llvm.fma.f32(float %.034, float %74, float noundef 0xBEDECCB920000000)
  %.035 = select i1 %.not132, float %101, float %100
  %102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not133 = icmp eq i32 %102, 0
  %103 = call float @llvm.nvvm.fma.rn.ftz.f(float %.035, float %74, float noundef 0x3F33C5BD60000000) #139
  %104 = call float @llvm.fma.f32(float %.035, float %74, float noundef 0x3F33C5BD60000000)
  %.036 = select i1 %.not133, float %104, float %103
  %105 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not134 = icmp eq i32 %105, 0
  %106 = call float @llvm.nvvm.fma.rn.ftz.f(float %.036, float %74, float noundef 0x3F495C6680000000) #139
  %107 = call float @llvm.fma.f32(float %.036, float %74, float noundef 0x3F495C6680000000)
  %.037 = select i1 %.not134, float %107, float %106
  %108 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not135 = icmp eq i32 %108, 0
  %109 = call float @llvm.nvvm.fma.rn.ftz.f(float %.037, float %74, float noundef 0xBF7DD54360000000) #139
  %110 = call float @llvm.fma.f32(float %.037, float %74, float noundef 0xBF7DD54360000000)
  %.038 = select i1 %.not135, float %110, float %109
  %111 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not136 = icmp eq i32 %111, 0
  %112 = call float @llvm.nvvm.fma.rn.ftz.f(float %.038, float %74, float noundef 0xBF9B4EE8E0000000) #139
  %113 = call float @llvm.fma.f32(float %.038, float %74, float noundef 0xBF9B4EE8E0000000)
  %.039 = select i1 %.not136, float %113, float %112
  %114 = fmul float %74, %.039
  %115 = fadd float %.024, 0xC016148F60000000
  %116 = fadd float %115, 0x3E734F46E0000000
  %117 = fmul float %116, %114
  %118 = fadd float %.024, 0xC0214EB560000000
  %119 = fadd float %118, 0xBE9999BDA0000000
  %120 = fmul float %119, %117
  br label %__nv_j0f.exit

__nv_isinff.exit.i:                               ; preds = %__internal_accurate_logf.exit
  %121 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not85 = icmp eq i32 %121, 0
  %122 = call float @llvm.nvvm.fabs.ftz.f(float %.024) #139
  %123 = call float @llvm.fabs.f32(float %.024)
  %.040 = select i1 %.not85, float %123, float %122
  %124 = fcmp oeq float %.040, 0x7FF0000000000000
  br i1 %124, label %__nv_j0f.exit, label %__internal_fmad.exit13.i

__internal_fmad.exit13.i:                         ; preds = %__nv_isinff.exit.i
  %125 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.024) #141, !srcloc !73
  %126 = fmul float %125, %125
  %127 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not86 = icmp eq i32 %127, 0
  %128 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400ADFD260000000, float %126, float noundef 0xBFE076F840000000) #139
  %129 = call float @llvm.fma.f32(float %126, float noundef 0x400ADFD260000000, float noundef 0xBFE076F840000000)
  %.041 = select i1 %.not86, float %129, float %128
  %130 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not87 = icmp eq i32 %130, 0
  %131 = call float @llvm.nvvm.fma.rn.ftz.f(float %.041, float %126, float noundef 0x3FBA767E60000000) #139
  %132 = call float @llvm.fma.f32(float %.041, float %126, float noundef 0x3FBA767E60000000)
  %.043 = select i1 %.not87, float %132, float %131
  %133 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not88 = icmp eq i32 %133, 0
  %134 = call float @llvm.nvvm.fma.rn.ftz.f(float %.043, float %126, float noundef 0xBFAFFFF6C0000000) #139
  %135 = call float @llvm.fma.f32(float %.043, float %126, float noundef 0xBFAFFFF6C0000000)
  %.044 = select i1 %.not88, float %135, float %134
  %136 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not89 = icmp eq i32 %136, 0
  %137 = call float @llvm.nvvm.fma.rn.ftz.f(float %.044, float %126, float noundef 1.000000e+00) #139
  %138 = call float @llvm.fma.f32(float %.044, float %126, float noundef 1.000000e+00)
  %.045 = select i1 %.not89, float %138, float %137
  %139 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not90 = icmp eq i32 %139, 0
  %140 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF23C0120000000, float %126, float noundef 0xBFCA4825A0000000) #139
  %141 = call float @llvm.fma.f32(float %126, float noundef 0x3FF23C0120000000, float noundef 0xBFCA4825A0000000)
  %.047 = select i1 %.not90, float %141, float %140
  %142 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not91 = icmp eq i32 %142, 0
  %143 = call float @llvm.nvvm.fma.rn.ftz.f(float %.047, float %126, float noundef 0x3FB0A9DA20000000) #139
  %144 = call float @llvm.fma.f32(float %.047, float %126, float noundef 0x3FB0A9DA20000000)
  %.048 = select i1 %.not91, float %144, float %143
  %145 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not92 = icmp eq i32 %145, 0
  %146 = call float @llvm.nvvm.fma.rn.ftz.f(float %.048, float %126, float noundef 0xBFBFFFFFE0000000) #139
  %147 = call float @llvm.fma.f32(float %.048, float %126, float noundef 0xBFBFFFFFE0000000)
  %.049 = select i1 %.not92, float %147, float %146
  %148 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not93 = icmp eq i32 %148, 0
  %149 = call float @llvm.nvvm.fma.rn.ftz.f(float %.049, float %125, float %.024) #139
  %150 = call float @llvm.fma.f32(float %.049, float %125, float %.024)
  %.051 = select i1 %.not93, float %150, float %149
  %151 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not94 = icmp eq i32 %151, 0
  br i1 %.not94, label %154, label %152

152:                                              ; preds = %__internal_fmad.exit13.i
  %153 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.024) #139
  br label %__nv_rsqrtf.exit.i

154:                                              ; preds = %__internal_fmad.exit13.i
  %155 = call float @llvm.nvvm.rsqrt.approx.f(float %.024) #139
  br label %__nv_rsqrtf.exit.i

__nv_rsqrtf.exit.i:                               ; preds = %154, %152
  %.053 = phi float [ %153, %152 ], [ %155, %154 ]
  %156 = fmul float %.053, 0x3FE9884540000000
  %157 = fmul float %.045, %156
  %158 = fmul float %.051, 0x3FE45F3060000000
  %159 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not95 = icmp eq i32 %159, 0
  %160 = call i32 @llvm.nvvm.f2i.rn.ftz(float %158) #139
  %161 = call i32 @llvm.nvvm.f2i.rn(float %158) #139
  %.054 = select i1 %.not95, i32 %161, i32 %160
  %162 = sitofp i32 %.054 to float
  %163 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not96 = icmp eq i32 %163, 0
  %164 = call float @llvm.nvvm.fma.rn.ftz.f(float %162, float noundef 0xBFF921FB40000000, float %.051) #139
  %165 = call float @llvm.fma.f32(float %162, float noundef 0xBFF921FB40000000, float %.051)
  %.055 = select i1 %.not96, float %165, float %164
  %166 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not97 = icmp eq i32 %166, 0
  %167 = call float @llvm.nvvm.fma.rn.ftz.f(float %162, float noundef 0xBE74442D00000000, float %.055) #139
  %168 = call float @llvm.fma.f32(float %162, float noundef 0xBE74442D00000000, float %.055)
  %.056 = select i1 %.not97, float %168, float %167
  %169 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not98 = icmp eq i32 %169, 0
  %170 = call float @llvm.nvvm.fma.rn.ftz.f(float %162, float noundef 0xBCF84698A0000000, float %.056) #139
  %171 = call float @llvm.fma.f32(float %162, float noundef 0xBCF84698A0000000, float %.056)
  %.057 = select i1 %.not98, float %171, float %170
  %172 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not99 = icmp eq i32 %172, 0
  %173 = call float @llvm.nvvm.fabs.ftz.f(float %.051) #139
  %174 = call float @llvm.fabs.f32(float %.051)
  %.058 = select i1 %.not99, float %174, float %173
  %175 = fcmp ult float %.058, 1.056150e+05
  br i1 %175, label %__internal_trig_reduction_kernel.exit.i.i, label %__nv_isinff.exit.i.i.i

__nv_isinff.exit.i.i.i:                           ; preds = %__nv_rsqrtf.exit.i
  %176 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not117 = icmp eq i32 %176, 0
  %.059 = select i1 %.not117, float %174, float %173
  %177 = fcmp oeq float %.059, 0x7FF0000000000000
  br i1 %177, label %__nv_fmul_rn.exit.i.i.i, label %181

__nv_fmul_rn.exit.i.i.i:                          ; preds = %__nv_isinff.exit.i.i.i
  %178 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not123 = icmp eq i32 %178, 0
  %179 = call float @llvm.nvvm.mul.rn.ftz.f(float %.051, float noundef 0.000000e+00) #139
  %180 = fmul float %.051, 0.000000e+00
  %.060 = select i1 %.not123, float %180, float %179
  br label %__internal_trig_reduction_kernel.exit.i.i

181:                                              ; preds = %__nv_isinff.exit.i.i.i
  %182 = bitcast float %.051 to i32
  %183 = lshr i32 %182, 23
  %184 = and i32 %183, 255
  %185 = add nsw i32 %184, -128
  %186 = shl i32 %182, 8
  %187 = or i32 %186, -2147483648
  %188 = lshr i32 %185, 5
  br label %189

189:                                              ; preds = %181, %189
  %iq.i.i.i.i.0141 = phi i32 [ 0, %181 ], [ %197, %189 ]
  %hi.i.i.i.i.0140 = phi i32 [ 0, %181 ], [ %195, %189 ]
  %190 = zext i32 %iq.i.i.i.i.0141 to i64
  %191 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %190
  %192 = load i32, i32 addrspace(1)* %191, align 4
  %193 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %192, i32 %187, i32 %hi.i.i.i.i.0140) #141, !srcloc !67
  %194 = extractvalue { i32, i32 } %193, 0
  %195 = extractvalue { i32, i32 } %193, 1
  %196 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %190
  store i32 %194, i32* %196, align 4, !noalias !68
  %197 = add nuw nsw i32 %iq.i.i.i.i.0141, 1
  %exitcond146.not = icmp eq i32 %197, 6
  br i1 %exitcond146.not, label %198, label %189, !llvm.loop !71

198:                                              ; preds = %189
  %199 = and i32 %182, -2147483648
  %200 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %195, i32* %200, align 4, !noalias !68
  %201 = and i32 %183, 31
  %202 = sub nsw i32 6, %188
  %203 = sext i32 %202 to i64
  %204 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %203
  %205 = load i32, i32* %204, align 4
  %206 = sub nsw i32 5, %188
  %207 = sext i32 %206 to i64
  %208 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %207
  %209 = load i32, i32* %208, align 4
  %.not118 = icmp eq i32 %201, 0
  br i1 %.not118, label %__internal_trig_reduction_slowpath.exit.i.i.i, label %210

210:                                              ; preds = %198
  %211 = sub nsw i32 4, %188
  %212 = sub nuw nsw i32 32, %201
  %213 = shl i32 %205, %201
  %214 = lshr i32 %209, %212
  %215 = add i32 %214, %213
  %216 = shl i32 %209, %201
  %217 = sext i32 %211 to i64
  %218 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %217
  %219 = load i32, i32* %218, align 4
  %220 = lshr i32 %219, %212
  %221 = add i32 %220, %216
  br label %__internal_trig_reduction_slowpath.exit.i.i.i

__internal_trig_reduction_slowpath.exit.i.i.i:    ; preds = %210, %198
  %hi.i.i.i.i.1 = phi i32 [ %215, %210 ], [ %205, %198 ]
  %lo.i.i.i.i.0 = phi i32 [ %221, %210 ], [ %209, %198 ]
  %222 = lshr i32 %hi.i.i.i.i.1, 30
  %223 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.1, i32 %lo.i.i.i.i.0, i32 noundef 2)
  %224 = shl i32 %lo.i.i.i.i.0, 2
  %225 = lshr i32 %223, 31
  %226 = add nuw nsw i32 %225, %222
  %.not119 = icmp eq i32 %199, 0
  %227 = sub nsw i32 0, %226
  %spec.select = select i1 %.not119, i32 %226, i32 %227
  %.not120 = icmp sgt i32 %223, -1
  %228 = xor i32 %199, -2147483648
  %s.i.i.i.i.0 = select i1 %.not120, i32 %199, i32 %228
  %not..not120 = xor i1 %.not120, true
  %229 = sext i1 %not..not120 to i32
  %hi.i.i.i.i.2 = xor i32 %223, %229
  %lo.i.i.i.i.1 = xor i32 %224, %229
  %230 = zext i32 %hi.i.i.i.i.2 to i64
  %231 = shl nuw i64 %230, 32
  %232 = zext i32 %lo.i.i.i.i.1 to i64
  %233 = or i64 %231, %232
  %234 = sitofp i64 %233 to double
  %235 = fmul double %234, 0x3BF921FB54442D19
  %236 = fptrunc double %235 to float
  %.not122 = icmp eq i32 %s.i.i.i.i.0, 0
  %237 = fneg float %236
  %r.i.i.i.i.0 = select i1 %.not122, float %236, float %237
  br label %__internal_trig_reduction_kernel.exit.i.i

__internal_trig_reduction_kernel.exit.i.i:        ; preds = %__nv_fmul_rn.exit.i.i.i, %__internal_trig_reduction_slowpath.exit.i.i.i, %__nv_rsqrtf.exit.i
  %i.i.i.1 = phi i32 [ %.054, %__nv_rsqrtf.exit.i ], [ 0, %__nv_fmul_rn.exit.i.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %t.i.i.i.1 = phi float [ %.057, %__nv_rsqrtf.exit.i ], [ %.060, %__nv_fmul_rn.exit.i.i.i ], [ %r.i.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %238 = and i32 %i.i.i.1, 3
  %239 = uitofp i32 %238 to float
  %240 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not100 = icmp eq i32 %240, 0
  %241 = call float @llvm.nvvm.fma.rn.ftz.f(float %239, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000) #139
  %242 = call float @llvm.fma.f32(float %239, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000)
  %.061 = select i1 %.not100, float %242, float %241
  %243 = fadd float %t.i.i.i.1, %.061
  %244 = fmul float %243, 0x3FE45F3060000000
  %245 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not101 = icmp eq i32 %245, 0
  %246 = call i32 @llvm.nvvm.f2i.rn.ftz(float %244) #139
  %247 = call i32 @llvm.nvvm.f2i.rn(float %244) #139
  %.062 = select i1 %.not101, i32 %247, i32 %246
  %248 = sitofp i32 %.062 to float
  %249 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not102 = icmp eq i32 %249, 0
  %250 = call float @llvm.nvvm.fma.rn.ftz.f(float %248, float noundef 0xBFF921FB40000000, float %243) #139
  %251 = call float @llvm.fma.f32(float %248, float noundef 0xBFF921FB40000000, float %243)
  %.064 = select i1 %.not102, float %251, float %250
  %252 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not103 = icmp eq i32 %252, 0
  %253 = call float @llvm.nvvm.fma.rn.ftz.f(float %248, float noundef 0xBE74442D00000000, float %.064) #139
  %254 = call float @llvm.fma.f32(float %248, float noundef 0xBE74442D00000000, float %.064)
  %.065 = select i1 %.not103, float %254, float %253
  %255 = add i32 %.062, 1
  %256 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not104 = icmp eq i32 %256, 0
  %257 = call float @llvm.nvvm.mul.rn.ftz.f(float %.065, float %.065) #139
  %258 = fmul float %.065, %.065
  %.066 = select i1 %.not104, float %258, float %257
  %259 = and i32 %255, 1
  %.not105 = icmp eq i32 %259, 0
  %260 = select i1 %.not105, float %.065, float 1.000000e+00
  %261 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not106 = icmp eq i32 %261, 0
  %262 = call float @llvm.nvvm.fma.rn.ftz.f(float %.066, float %260, float noundef 0.000000e+00) #139
  %263 = call float @llvm.fma.f32(float %.066, float %260, float noundef 0.000000e+00)
  %.068 = select i1 %.not106, float %263, float %262
  br i1 %.not105, label %__internal_fmad.exit2.i.i.i, label %__internal_fmad.exit1.i.i.i

__internal_fmad.exit1.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i
  %264 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not116 = icmp eq i32 %264, 0
  %265 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.066, float noundef 0xBF56C0FDA0000000) #139
  %266 = call float @llvm.fma.f32(float %.066, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.069 = select i1 %.not116, float %266, float %265
  br label %__internal_fmad.exit2.i.i.i

__internal_fmad.exit2.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i, %__internal_fmad.exit1.i.i.i
  %267 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i ]
  %268 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i ]
  %269 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i ], [ %.069, %__internal_fmad.exit1.i.i.i ]
  %270 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not109 = icmp eq i32 %270, 0
  %271 = call float @llvm.nvvm.fma.rn.ftz.f(float %269, float %.066, float noundef %268) #139
  %272 = call float @llvm.fma.f32(float %269, float %.066, float noundef %268)
  %.071 = select i1 %.not109, float %272, float %271
  %273 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not111 = icmp eq i32 %273, 0
  %274 = call float @llvm.nvvm.fma.rn.ftz.f(float %.071, float %.066, float noundef %267) #139
  %275 = call float @llvm.fma.f32(float %.071, float %.066, float noundef %267)
  %.073 = select i1 %.not111, float %275, float %274
  %276 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not112 = icmp eq i32 %276, 0
  %277 = call float @llvm.nvvm.fma.rn.ftz.f(float %.073, float %.068, float %260) #139
  %278 = call float @llvm.fma.f32(float %.073, float %.068, float %260)
  %.074 = select i1 %.not112, float %278, float %277
  %279 = and i32 %255, 2
  %.not113 = icmp eq i32 %279, 0
  br i1 %.not113, label %__internal_cos_offset_f.exit.i, label %__internal_fmad.exit5.i.i.i

__internal_fmad.exit5.i.i.i:                      ; preds = %__internal_fmad.exit2.i.i.i
  %280 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not115 = icmp eq i32 %280, 0
  %281 = call float @llvm.nvvm.fma.rn.ftz.f(float %.074, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %282 = call float @llvm.fma.f32(float %.074, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.075 = select i1 %.not115, float %282, float %281
  br label %__internal_cos_offset_f.exit.i

__internal_cos_offset_f.exit.i:                   ; preds = %__internal_fmad.exit5.i.i.i, %__internal_fmad.exit2.i.i.i
  %z.i.i.i.0 = phi float [ %.075, %__internal_fmad.exit5.i.i.i ], [ %.074, %__internal_fmad.exit2.i.i.i ]
  %283 = fmul float %157, %z.i.i.i.0
  br label %__nv_j0f.exit

__nv_j0f.exit:                                    ; preds = %__internal_cos_offset_f.exit.i, %__nv_isinff.exit.i, %__internal_fmad.exit.i
  %r.i5.1 = phi float [ %120, %__internal_fmad.exit.i ], [ %283, %__internal_cos_offset_f.exit.i ], [ 0.000000e+00, %__nv_isinff.exit.i ]
  %284 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not114 = icmp eq i32 %284, 0
  %285 = call float @llvm.nvvm.fma.rn.ftz.f(float %68, float %r.i5.1, float %.07) #139
  %286 = call float @llvm.fma.f32(float %68, float %r.i5.1, float %.07)
  %.077 = select i1 %.not114, float %286, float %285
  br label %549

287:                                              ; preds = %__nv_fabsf.exit
  %288 = fcmp ugt float %.01, 0x3FFECF5F00000000
  br i1 %288, label %328, label %__internal_fmad.exit9

__internal_fmad.exit9:                            ; preds = %287
  %289 = fadd float %.01, 0xBFEC982EC0000000
  %290 = fadd float %289, 0x3E4CAFA060000000
  %291 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not56 = icmp eq i32 %291, 0
  %292 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F9BB91660000000, float %290, float noundef 0xBFC04AE980000000) #139
  %293 = call float @llvm.fma.f32(float %290, float noundef 0x3F9BB91660000000, float noundef 0xBFC04AE980000000)
  %.078 = select i1 %.not56, float %293, float %292
  %294 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not57 = icmp eq i32 %294, 0
  %295 = call float @llvm.nvvm.fma.rn.ftz.f(float %.078, float %290, float noundef 0x3FCFE59920000000) #139
  %296 = call float @llvm.fma.f32(float %.078, float %290, float noundef 0x3FCFE59920000000)
  %.079 = select i1 %.not57, float %296, float %295
  %297 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not58 = icmp eq i32 %297, 0
  %298 = call float @llvm.nvvm.fma.rn.ftz.f(float %.079, float %290, float noundef 0xBFD17E5360000000) #139
  %299 = call float @llvm.fma.f32(float %.079, float %290, float noundef 0xBFD17E5360000000)
  %.081 = select i1 %.not58, float %299, float %298
  %300 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not59 = icmp eq i32 %300, 0
  %301 = call float @llvm.nvvm.fma.rn.ftz.f(float %.081, float %290, float noundef 0x3FCB79D260000000) #139
  %302 = call float @llvm.fma.f32(float %.081, float %290, float noundef 0x3FCB79D260000000)
  %.082 = select i1 %.not59, float %302, float %301
  %303 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not60 = icmp eq i32 %303, 0
  %304 = call float @llvm.nvvm.fma.rn.ftz.f(float %.082, float %290, float noundef 0xBFC7189FE0000000) #139
  %305 = call float @llvm.fma.f32(float %.082, float %290, float noundef 0xBFC7189FE0000000)
  %.083 = select i1 %.not60, float %305, float %304
  %306 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not61 = icmp eq i32 %306, 0
  %307 = call float @llvm.nvvm.fma.rn.ftz.f(float %.083, float %290, float noundef 0x3FC84EE9A0000000) #139
  %308 = call float @llvm.fma.f32(float %.083, float %290, float noundef 0x3FC84EE9A0000000)
  %.085 = select i1 %.not61, float %308, float %307
  %309 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not62 = icmp eq i32 %309, 0
  %310 = call float @llvm.nvvm.fma.rn.ftz.f(float %.085, float %290, float noundef 0xBFCA4B9640000000) #139
  %311 = call float @llvm.fma.f32(float %.085, float %290, float noundef 0xBFCA4B9640000000)
  %.086 = select i1 %.not62, float %311, float %310
  %312 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not63 = icmp eq i32 %312, 0
  %313 = call float @llvm.nvvm.fma.rn.ftz.f(float %.086, float %290, float noundef 0x3FCC1E87A0000000) #139
  %314 = call float @llvm.fma.f32(float %.086, float %290, float noundef 0x3FCC1E87A0000000)
  %.087 = select i1 %.not63, float %314, float %313
  %315 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not64 = icmp eq i32 %315, 0
  %316 = call float @llvm.nvvm.fma.rn.ftz.f(float %.087, float %290, float noundef 0xBFCCF228A0000000) #139
  %317 = call float @llvm.fma.f32(float %.087, float %290, float noundef 0xBFCCF228A0000000)
  %.088 = select i1 %.not64, float %317, float %316
  %318 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not65 = icmp eq i32 %318, 0
  %319 = call float @llvm.nvvm.fma.rn.ftz.f(float %.088, float %290, float noundef 0x3FCC3A4940000000) #139
  %320 = call float @llvm.fma.f32(float %.088, float %290, float noundef 0x3FCC3A4940000000)
  %.089 = select i1 %.not65, float %320, float %319
  %321 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not66 = icmp eq i32 %321, 0
  %322 = call float @llvm.nvvm.fma.rn.ftz.f(float %.089, float %290, float noundef 0xBFDF7E35A0000000) #139
  %323 = call float @llvm.fma.f32(float %.089, float %290, float noundef 0xBFDF7E35A0000000)
  %.090 = select i1 %.not66, float %323, float %322
  %324 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not67 = icmp eq i32 %324, 0
  %325 = call float @llvm.nvvm.fma.rn.ftz.f(float %.090, float %290, float noundef 0x3FEC243760000000) #139
  %326 = call float @llvm.fma.f32(float %.090, float %290, float noundef 0x3FEC243760000000)
  %.092 = select i1 %.not67, float %326, float %325
  %327 = fmul float %290, %.092
  br label %549

328:                                              ; preds = %287
  %329 = fcmp ugt float %.01, 0x40161663C0000000
  br i1 %329, label %360, label %__internal_fmad.exit33

__internal_fmad.exit33:                           ; preds = %328
  %330 = fadd float %.01, 0xC00FA95340000000
  %331 = fadd float %330, 0xBE7B30AD40000000
  %332 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %332, 0
  %333 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBE893BA7E0000000, float %331, float noundef 0xBE8BC4C0E0000000) #139
  %334 = call float @llvm.fma.f32(float %331, float noundef 0xBE893BA7E0000000, float noundef 0xBE8BC4C0E0000000)
  %.094 = select i1 %.not47, float %334, float %333
  %335 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %335, 0
  %336 = call float @llvm.nvvm.fma.rn.ftz.f(float %.094, float %331, float noundef 0xBED0AE0C80000000) #139
  %337 = call float @llvm.fma.f32(float %.094, float %331, float noundef 0xBED0AE0C80000000)
  %.095 = select i1 %.not48, float %337, float %336
  %338 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %338, 0
  %339 = call float @llvm.nvvm.fma.rn.ftz.f(float %.095, float %331, float noundef 0x3F0AA8C200000000) #139
  %340 = call float @llvm.fma.f32(float %.095, float %331, float noundef 0x3F0AA8C200000000)
  %.096 = select i1 %.not49, float %340, float %339
  %341 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %341, 0
  %342 = call float @llvm.nvvm.fma.rn.ftz.f(float %.096, float %331, float noundef 0x3F2959DAE0000000) #139
  %343 = call float @llvm.fma.f32(float %.096, float %331, float noundef 0x3F2959DAE0000000)
  %.097 = select i1 %.not50, float %343, float %342
  %344 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not51 = icmp eq i32 %344, 0
  %345 = call float @llvm.nvvm.fma.rn.ftz.f(float %.097, float %331, float noundef 0xBF61E34180000000) #139
  %346 = call float @llvm.fma.f32(float %.097, float %331, float noundef 0xBF61E34180000000)
  %.098 = select i1 %.not51, float %346, float %345
  %347 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not52 = icmp eq i32 %347, 0
  %348 = call float @llvm.nvvm.fma.rn.ftz.f(float %.098, float %331, float noundef 0xBF7C0FE5C0000000) #139
  %349 = call float @llvm.fma.f32(float %.098, float %331, float noundef 0xBF7C0FE5C0000000)
  %.099 = select i1 %.not52, float %349, float %348
  %350 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not53 = icmp eq i32 %350, 0
  %351 = call float @llvm.nvvm.fma.rn.ftz.f(float %.099, float %331, float noundef 0x3FADF6D6A0000000) #139
  %352 = call float @llvm.fma.f32(float %.099, float %331, float noundef 0x3FADF6D6A0000000)
  %.0101 = select i1 %.not53, float %352, float %351
  %353 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not54 = icmp eq i32 %353, 0
  %354 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0101, float %331, float noundef 0x3FAA09BE20000000) #139
  %355 = call float @llvm.fma.f32(float %.0101, float %331, float noundef 0x3FAA09BE20000000)
  %.0102 = select i1 %.not54, float %355, float %354
  %356 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not55 = icmp eq i32 %356, 0
  %357 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0102, float %331, float noundef 0xBFD9C34260000000) #139
  %358 = call float @llvm.fma.f32(float %.0102, float %331, float noundef 0xBFD9C34260000000)
  %.0103 = select i1 %.not55, float %358, float %357
  %359 = fmul float %331, %.0103
  br label %549

360:                                              ; preds = %328
  %361 = fcmp ugt float %.01, 0x40214EF300000000
  br i1 %361, label %__nv_isinff.exit, label %__internal_fmad.exit51

__internal_fmad.exit51:                           ; preds = %360
  %362 = fadd float %.01, 0xC01C581DC0000000
  %363 = fadd float %362, 0xBE739C8400000000
  %364 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not39 = icmp eq i32 %364, 0
  %365 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EA219D7C0000000, float %363, float noundef 0x3EC53B4D80000000) #139
  %366 = call float @llvm.fma.f32(float %363, float noundef 0x3EA219D7C0000000, float noundef 0x3EC53B4D80000000)
  %.0104 = select i1 %.not39, float %366, float %365
  %367 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not40 = icmp eq i32 %367, 0
  %368 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0104, float %363, float noundef 0xBF080A9800000000) #139
  %369 = call float @llvm.fma.f32(float %.0104, float %363, float noundef 0xBF080A9800000000)
  %.0105 = select i1 %.not40, float %369, float %368
  %370 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not41 = icmp eq i32 %370, 0
  %371 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0105, float %363, float noundef 0xBF22631560000000) #139
  %372 = call float @llvm.fma.f32(float %.0105, float %363, float noundef 0xBF22631560000000)
  %.0106 = select i1 %.not41, float %372, float %371
  %373 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %373, 0
  %374 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0106, float %363, float noundef 0x3F61D32420000000) #139
  %375 = call float @llvm.fma.f32(float %.0106, float %363, float noundef 0x3F61D32420000000)
  %.0108 = select i1 %.not42, float %375, float %374
  %376 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %376, 0
  %377 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0108, float %363, float noundef 0x3F6B2E9AA0000000) #139
  %378 = call float @llvm.fma.f32(float %.0108, float %363, float noundef 0x3F6B2E9AA0000000)
  %.0109 = select i1 %.not43, float %378, float %377
  %379 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %379, 0
  %380 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0109, float %363, float noundef 0xBFA8969AE0000000) #139
  %381 = call float @llvm.fma.f32(float %.0109, float %363, float noundef 0xBFA8969AE0000000)
  %.0110 = select i1 %.not44, float %381, float %380
  %382 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %382, 0
  %383 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0110, float %363, float noundef 0xBF95AEF320000000) #139
  %384 = call float @llvm.fma.f32(float %.0110, float %363, float noundef 0xBF95AEF320000000)
  %.0111 = select i1 %.not45, float %384, float %383
  %385 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %385, 0
  %386 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0111, float %363, float noundef 0x3FD334CCA0000000) #139
  %387 = call float @llvm.fma.f32(float %.0111, float %363, float noundef 0x3FD334CCA0000000)
  %.0112 = select i1 %.not46, float %387, float %386
  %388 = fmul float %363, %.0112
  br label %549

__nv_isinff.exit:                                 ; preds = %360
  %389 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %389, 0
  %390 = call float @llvm.nvvm.fabs.ftz.f(float %.01) #139
  %391 = call float @llvm.fabs.f32(float %.01)
  %.0113 = select i1 %.not2, float %391, float %390
  %392 = fcmp oeq float %.0113, 0x7FF0000000000000
  br i1 %392, label %549, label %__internal_fmad.exit70

__internal_fmad.exit70:                           ; preds = %__nv_isinff.exit
  %393 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.01) #141, !srcloc !73
  %394 = fmul float %393, %393
  %395 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %395, 0
  %396 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFD98D3E60000000, float %394, float noundef 0x3FBA1ABE00000000) #139
  %397 = call float @llvm.fma.f32(float %394, float noundef 0xBFD98D3E60000000, float noundef 0x3FBA1ABE00000000)
  %.0115 = select i1 %.not3, float %397, float %396
  %398 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %398, 0
  %399 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0115, float %394, float noundef 0xBFAFFF0AA0000000) #139
  %400 = call float @llvm.fma.f32(float %.0115, float %394, float noundef 0xBFAFFF0AA0000000)
  %.0116 = select i1 %.not4, float %400, float %399
  %401 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %401, 0
  %402 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0116, float %394, float noundef 1.000000e+00) #139
  %403 = call float @llvm.fma.f32(float %.0116, float %394, float noundef 1.000000e+00)
  %.0117 = select i1 %.not5, float %403, float %402
  %404 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %404, 0
  %405 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF199AC20000000, float %394, float noundef 0xBFCA1A6380000000) #139
  %406 = call float @llvm.fma.f32(float %394, float noundef 0x3FF199AC20000000, float noundef 0xBFCA1A6380000000)
  %.0114 = select i1 %.not6, float %406, float %405
  %407 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %407, 0
  %408 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0114, float %394, float noundef 0x3FB0A8F060000000) #139
  %409 = call float @llvm.fma.f32(float %.0114, float %394, float noundef 0x3FB0A8F060000000)
  %.0107 = select i1 %.not7, float %409, float %408
  %410 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %410, 0
  %411 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0107, float %394, float noundef 0xBFBFFFFF60000000) #139
  %412 = call float @llvm.fma.f32(float %.0107, float %394, float noundef 0xBFBFFFFF60000000)
  %.0100 = select i1 %.not8, float %412, float %411
  %413 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %413, 0
  %414 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0100, float %393, float %.01) #139
  %415 = call float @llvm.fma.f32(float %.0100, float %393, float %.01)
  %.093 = select i1 %.not9, float %415, float %414
  %416 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %416, 0
  br i1 %.not10, label %419, label %417

417:                                              ; preds = %__internal_fmad.exit70
  %418 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.01) #139
  br label %__nv_rsqrtf.exit

419:                                              ; preds = %__internal_fmad.exit70
  %420 = call float @llvm.nvvm.rsqrt.approx.f(float %.01) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %419, %417
  %.091 = phi float [ %418, %417 ], [ %420, %419 ]
  %421 = fmul float %.091, 0x3FE9884540000000
  %422 = fmul float %.0117, %421
  %423 = fmul float %.093, 0x3FE45F3060000000
  %424 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %424, 0
  %425 = call i32 @llvm.nvvm.f2i.rn.ftz(float %423) #139
  %426 = call i32 @llvm.nvvm.f2i.rn(float %423) #139
  %.084 = select i1 %.not11, i32 %426, i32 %425
  %427 = sitofp i32 %.084 to float
  %428 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %428, 0
  %429 = call float @llvm.nvvm.fma.rn.ftz.f(float %427, float noundef 0xBFF921FB40000000, float %.093) #139
  %430 = call float @llvm.fma.f32(float %427, float noundef 0xBFF921FB40000000, float %.093)
  %.080 = select i1 %.not12, float %430, float %429
  %431 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %431, 0
  %432 = call float @llvm.nvvm.fma.rn.ftz.f(float %427, float noundef 0xBE74442D00000000, float %.080) #139
  %433 = call float @llvm.fma.f32(float %427, float noundef 0xBE74442D00000000, float %.080)
  %.076 = select i1 %.not13, float %433, float %432
  %434 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %434, 0
  %435 = call float @llvm.nvvm.fma.rn.ftz.f(float %427, float noundef 0xBCF84698A0000000, float %.076) #139
  %436 = call float @llvm.fma.f32(float %427, float noundef 0xBCF84698A0000000, float %.076)
  %.072 = select i1 %.not14, float %436, float %435
  %437 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %437, 0
  %438 = call float @llvm.nvvm.fabs.ftz.f(float %.093) #139
  %439 = call float @llvm.fabs.f32(float %.093)
  %.070 = select i1 %.not15, float %439, float %438
  %440 = fcmp ult float %.070, 1.056150e+05
  br i1 %440, label %__internal_trig_reduction_kernel.exit.i, label %__nv_isinff.exit.i.i

__nv_isinff.exit.i.i:                             ; preds = %__nv_rsqrtf.exit
  %441 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %441, 0
  %.067 = select i1 %.not32, float %439, float %438
  %442 = fcmp oeq float %.067, 0x7FF0000000000000
  br i1 %442, label %__nv_fmul_rn.exit.i.i, label %446

__nv_fmul_rn.exit.i.i:                            ; preds = %__nv_isinff.exit.i.i
  %443 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not38 = icmp eq i32 %443, 0
  %444 = call float @llvm.nvvm.mul.rn.ftz.f(float %.093, float noundef 0.000000e+00) #139
  %445 = fmul float %.093, 0.000000e+00
  %.063 = select i1 %.not38, float %445, float %444
  br label %__internal_trig_reduction_kernel.exit.i

446:                                              ; preds = %__nv_isinff.exit.i.i
  %447 = bitcast float %.093 to i32
  %448 = lshr i32 %447, 23
  %449 = and i32 %448, 255
  %450 = add nsw i32 %449, -128
  %451 = shl i32 %447, 8
  %452 = or i32 %451, -2147483648
  %453 = lshr i32 %450, 5
  br label %454

454:                                              ; preds = %446, %454
  %hi.i.i.i.0139 = phi i32 [ 0, %446 ], [ %460, %454 ]
  %iq.i.i.i.0138 = phi i32 [ 0, %446 ], [ %462, %454 ]
  %455 = zext i32 %iq.i.i.i.0138 to i64
  %456 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %455
  %457 = load i32, i32 addrspace(1)* %456, align 4
  %458 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %457, i32 %452, i32 %hi.i.i.i.0139) #141, !srcloc !67
  %459 = extractvalue { i32, i32 } %458, 0
  %460 = extractvalue { i32, i32 } %458, 1
  %461 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %455
  store i32 %459, i32* %461, align 4, !noalias !68
  %462 = add nuw nsw i32 %iq.i.i.i.0138, 1
  %exitcond.not = icmp eq i32 %462, 6
  br i1 %exitcond.not, label %463, label %454, !llvm.loop !71

463:                                              ; preds = %454
  %464 = and i32 %447, -2147483648
  %465 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %460, i32* %465, align 4, !noalias !68
  %466 = and i32 %448, 31
  %467 = sub nsw i32 6, %453
  %468 = sext i32 %467 to i64
  %469 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %468
  %470 = load i32, i32* %469, align 4
  %471 = sub nsw i32 5, %453
  %472 = sext i32 %471 to i64
  %473 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %472
  %474 = load i32, i32* %473, align 4
  %.not33 = icmp eq i32 %466, 0
  br i1 %.not33, label %__internal_trig_reduction_slowpath.exit.i.i, label %475

475:                                              ; preds = %463
  %476 = sub nsw i32 4, %453
  %477 = sub nuw nsw i32 32, %466
  %478 = shl i32 %470, %466
  %479 = lshr i32 %474, %477
  %480 = add i32 %479, %478
  %481 = shl i32 %474, %466
  %482 = sext i32 %476 to i64
  %483 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %482
  %484 = load i32, i32* %483, align 4
  %485 = lshr i32 %484, %477
  %486 = add i32 %485, %481
  br label %__internal_trig_reduction_slowpath.exit.i.i

__internal_trig_reduction_slowpath.exit.i.i:      ; preds = %475, %463
  %lo.i.i.i.0 = phi i32 [ %486, %475 ], [ %474, %463 ]
  %hi.i.i.i.1 = phi i32 [ %480, %475 ], [ %470, %463 ]
  %487 = lshr i32 %hi.i.i.i.1, 30
  %488 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.1, i32 %lo.i.i.i.0, i32 noundef 2)
  %489 = shl i32 %lo.i.i.i.0, 2
  %490 = lshr i32 %488, 31
  %491 = add nuw nsw i32 %490, %487
  %.not34 = icmp eq i32 %464, 0
  %492 = sub nsw i32 0, %491
  %spec.select1 = select i1 %.not34, i32 %491, i32 %492
  %.not35 = icmp sgt i32 %488, -1
  %493 = xor i32 %464, -2147483648
  %not..not35 = xor i1 %.not35, true
  %494 = sext i1 %not..not35 to i32
  %lo.i.i.i.1 = xor i32 %489, %494
  %hi.i.i.i.2 = xor i32 %488, %494
  %s.i.i.i.0 = select i1 %.not35, i32 %464, i32 %493
  %495 = zext i32 %hi.i.i.i.2 to i64
  %496 = shl nuw i64 %495, 32
  %497 = zext i32 %lo.i.i.i.1 to i64
  %498 = or i64 %496, %497
  %499 = sitofp i64 %498 to double
  %500 = fmul double %499, 0x3BF921FB54442D19
  %501 = fptrunc double %500 to float
  %.not37 = icmp eq i32 %s.i.i.i.0, 0
  %502 = fneg float %501
  %r.i.i.i.0 = select i1 %.not37, float %501, float %502
  br label %__internal_trig_reduction_kernel.exit.i

__internal_trig_reduction_kernel.exit.i:          ; preds = %__nv_fmul_rn.exit.i.i, %__internal_trig_reduction_slowpath.exit.i.i, %__nv_rsqrtf.exit
  %i.i85.1 = phi i32 [ %.084, %__nv_rsqrtf.exit ], [ 0, %__nv_fmul_rn.exit.i.i ], [ %spec.select1, %__internal_trig_reduction_slowpath.exit.i.i ]
  %t.i.i.1 = phi float [ %.072, %__nv_rsqrtf.exit ], [ %.063, %__nv_fmul_rn.exit.i.i ], [ %r.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i ]
  %503 = and i32 %i.i85.1, 3
  %504 = uitofp i32 %503 to float
  %505 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %505, 0
  %506 = call float @llvm.nvvm.fma.rn.ftz.f(float %504, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %507 = call float @llvm.fma.f32(float %504, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.052 = select i1 %.not16, float %507, float %506
  %508 = fadd float %t.i.i.1, %.052
  %509 = fmul float %508, 0x3FE45F3060000000
  %510 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %510, 0
  %511 = call i32 @llvm.nvvm.f2i.rn.ftz(float %509) #139
  %512 = call i32 @llvm.nvvm.f2i.rn(float %509) #139
  %.050 = select i1 %.not17, i32 %512, i32 %511
  %513 = sitofp i32 %.050 to float
  %514 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %514, 0
  %515 = call float @llvm.nvvm.fma.rn.ftz.f(float %513, float noundef 0xBFF921FB40000000, float %508) #139
  %516 = call float @llvm.fma.f32(float %513, float noundef 0xBFF921FB40000000, float %508)
  %.046 = select i1 %.not18, float %516, float %515
  %517 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %517, 0
  %518 = call float @llvm.nvvm.fma.rn.ftz.f(float %513, float noundef 0xBE74442D00000000, float %.046) #139
  %519 = call float @llvm.fma.f32(float %513, float noundef 0xBE74442D00000000, float %.046)
  %.042 = select i1 %.not19, float %519, float %518
  %520 = add i32 %.050, 1
  %521 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %521, 0
  %522 = call float @llvm.nvvm.mul.rn.ftz.f(float %.042, float %.042) #139
  %523 = fmul float %.042, %.042
  %.033 = select i1 %.not20, float %523, float %522
  %524 = and i32 %520, 1
  %.not21 = icmp eq i32 %524, 0
  %525 = select i1 %.not21, float %.042, float 1.000000e+00
  %526 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %526, 0
  %527 = call float @llvm.nvvm.fma.rn.ftz.f(float %.033, float %525, float noundef 0.000000e+00) #139
  %528 = call float @llvm.fma.f32(float %.033, float %525, float noundef 0.000000e+00)
  %.026 = select i1 %.not22, float %528, float %527
  br i1 %.not21, label %__internal_fmad.exit2.i.i, label %__internal_fmad.exit1.i.i

__internal_fmad.exit1.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i
  %529 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %529, 0
  %530 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.033, float noundef 0xBF56C0FDA0000000) #139
  %531 = call float @llvm.fma.f32(float %.033, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.022 = select i1 %.not31, float %531, float %530
  br label %__internal_fmad.exit2.i.i

__internal_fmad.exit2.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i, %__internal_fmad.exit1.i.i
  %532 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i ]
  %533 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i ]
  %534 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i ], [ %.022, %__internal_fmad.exit1.i.i ]
  %535 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %535, 0
  %536 = call float @llvm.nvvm.fma.rn.ftz.f(float %534, float %.033, float noundef %533) #139
  %537 = call float @llvm.fma.f32(float %534, float %.033, float noundef %533)
  %.015 = select i1 %.not25, float %537, float %536
  %538 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %538, 0
  %539 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %.033, float noundef %532) #139
  %540 = call float @llvm.fma.f32(float %.015, float %.033, float noundef %532)
  %.08 = select i1 %.not27, float %540, float %539
  %541 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %541, 0
  %542 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %.026, float %525) #139
  %543 = call float @llvm.fma.f32(float %.08, float %.026, float %525)
  %.04 = select i1 %.not28, float %543, float %542
  %544 = and i32 %520, 2
  %.not29 = icmp eq i32 %544, 0
  br i1 %.not29, label %__internal_cos_offset_f.exit, label %__internal_fmad.exit5.i.i

__internal_fmad.exit5.i.i:                        ; preds = %__internal_fmad.exit2.i.i
  %545 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %545, 0
  %546 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %547 = call float @llvm.fma.f32(float %.04, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0 = select i1 %.not30, float %547, float %546
  br label %__internal_cos_offset_f.exit

__internal_cos_offset_f.exit:                     ; preds = %__internal_fmad.exit5.i.i, %__internal_fmad.exit2.i.i
  %z.i.i83.0 = phi float [ %.0, %__internal_fmad.exit5.i.i ], [ %.04, %__internal_fmad.exit2.i.i ]
  %548 = fmul float %422, %z.i.i83.0
  br label %549

549:                                              ; preds = %__internal_fmad.exit9, %__internal_fmad.exit51, %__nv_isinff.exit, %__internal_cos_offset_f.exit, %__internal_fmad.exit33, %__nv_j0f.exit
  %r.4 = phi float [ %.077, %__nv_j0f.exit ], [ %327, %__internal_fmad.exit9 ], [ %359, %__internal_fmad.exit33 ], [ %388, %__internal_fmad.exit51 ], [ %548, %__internal_cos_offset_f.exit ], [ 0.000000e+00, %__nv_isinff.exit ]
  %550 = fcmp olt float %a, 0.000000e+00
  %r.5 = select i1 %550, float 0x7FFFFFFFE0000000, float %r.4
  ret float %r.5
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_y1f(float %a) local_unnamed_addr #46 {
__nv_fabsf.exit:
  %result.i.i.i.i = alloca [7 x i32], align 4
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp olt float %.01, 0x3810000000000000
  br i1 %3, label %4, label %6

4:                                                ; preds = %__nv_fabsf.exit
  %5 = fdiv float 0xBFE45F3060000000, %.01
  br label %551

6:                                                ; preds = %__nv_fabsf.exit
  %7 = fcmp ugt float %.01, 0x3FFB2D5880000000
  br i1 %7, label %298, label %__internal_fmad.exit

__internal_fmad.exit:                             ; preds = %6
  %8 = fmul float %.01, %.01
  %9 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not65 = icmp eq i32 %9, 0
  %10 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E428C5980000000, float %8, float noundef 0xBEB0A4FB40000000) #139
  %11 = call float @llvm.fma.f32(float %8, float noundef 0x3E428C5980000000, float noundef 0xBEB0A4FB40000000)
  %.02 = select i1 %.not65, float %11, float %10
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not66 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %8, float noundef 0x3F12C7D2A0000000) #139
  %14 = call float @llvm.fma.f32(float %.02, float %8, float noundef 0x3F12C7D2A0000000)
  %.03 = select i1 %.not66, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not67 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %8, float noundef 0xBF6835B960000000) #139
  %17 = call float @llvm.fma.f32(float %.03, float %8, float noundef 0xBF6835B960000000)
  %.04 = select i1 %.not67, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not68 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %8, float noundef 0x3FABD39760000000) #139
  %20 = call float @llvm.fma.f32(float %.04, float %8, float noundef 0x3FABD39760000000)
  %.06 = select i1 %.not68, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not69 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %8, float noundef 0xBFC9186620000000) #139
  %23 = call float @llvm.fma.f32(float %.06, float %8, float noundef 0xBFC9186620000000)
  %.07 = select i1 %.not69, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not70 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fabs.ftz.f(float %.01) #139
  %26 = call float @llvm.fabs.f32(float %.01)
  %.08 = select i1 %.not70, float %26, float %25
  %27 = fcmp ugt float %.08, 8.000000e+00
  br i1 %27, label %__nv_isinff.exit.i, label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %__internal_fmad.exit
  %28 = fadd float %.08, 0xC00EA75580000000
  %29 = fadd float %28, 0x3E74A121E0000000
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not123 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3D35E68C60000000, float %29, float noundef 0x3D7037E840000000) #139
  %32 = call float @llvm.fma.f32(float %29, float noundef 0x3D35E68C60000000, float noundef 0x3D7037E840000000)
  %.010 = select i1 %.not123, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not124 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %29, float noundef 0xBDBC43D820000000) #139
  %35 = call float @llvm.fma.f32(float %.010, float %29, float noundef 0xBDBC43D820000000)
  %.011 = select i1 %.not124, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not125 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %29, float noundef 0xBDEBBBDFE0000000) #139
  %38 = call float @llvm.fma.f32(float %.011, float %29, float noundef 0xBDEBBBDFE0000000)
  %.012 = select i1 %.not125, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not126 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %29, float noundef 0x3E336193A0000000) #139
  %41 = call float @llvm.fma.f32(float %.012, float %29, float noundef 0x3E336193A0000000)
  %.013 = select i1 %.not126, float %41, float %40
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not127 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %29, float noundef 0x3E5D022E60000000) #139
  %44 = call float @llvm.fma.f32(float %.013, float %29, float noundef 0x3E5D022E60000000)
  %.014 = select i1 %.not127, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not128 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %29, float noundef 0xBEA1F1B900000000) #139
  %47 = call float @llvm.fma.f32(float %.014, float %29, float noundef 0xBEA1F1B900000000)
  %.015 = select i1 %.not128, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not129 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %29, float noundef 0xBEC3CCA7A0000000) #139
  %50 = call float @llvm.fma.f32(float %.015, float %29, float noundef 0xBEC3CCA7A0000000)
  %.017 = select i1 %.not129, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not130 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %29, float noundef 0x3F059B38A0000000) #139
  %53 = call float @llvm.fma.f32(float %.017, float %29, float noundef 0x3F059B38A0000000)
  %.018 = select i1 %.not130, float %53, float %52
  %54 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not131 = icmp eq i32 %54, 0
  %55 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %29, float noundef 0x3F1F3D6200000000) #139
  %56 = call float @llvm.fma.f32(float %.018, float %29, float noundef 0x3F1F3D6200000000)
  %.019 = select i1 %.not131, float %56, float %55
  %57 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not132 = icmp eq i32 %57, 0
  %58 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %29, float noundef 0xBF5D9D7380000000) #139
  %59 = call float @llvm.fma.f32(float %.019, float %29, float noundef 0xBF5D9D7380000000)
  %.020 = select i1 %.not132, float %59, float %58
  %60 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not133 = icmp eq i32 %60, 0
  %61 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %29, float noundef 0xBF64EDFFA0000000) #139
  %62 = call float @llvm.fma.f32(float %.020, float %29, float noundef 0xBF64EDFFA0000000)
  %.021 = select i1 %.not133, float %62, float %61
  %63 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not134 = icmp eq i32 %63, 0
  %64 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %29, float noundef 0x3FA0E73260000000) #139
  %65 = call float @llvm.fma.f32(float %.021, float %29, float noundef 0x3FA0E73260000000)
  %.022 = select i1 %.not134, float %65, float %64
  %66 = fadd float %.08, 0xC01C0FF600000000
  %67 = fadd float %66, 0x3E88971B60000000
  %68 = fmul float %67, %.022
  %69 = fmul float %29, %68
  %70 = fmul float %.08, %69
  br label %234

__nv_isinff.exit.i:                               ; preds = %__internal_fmad.exit
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not71 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fabs.ftz.f(float %.08) #139
  %73 = call float @llvm.fabs.f32(float %.08)
  %.024 = select i1 %.not71, float %73, float %72
  %74 = fcmp oeq float %.024, 0x7FF0000000000000
  br i1 %74, label %234, label %__internal_fmad.exit12.i

__internal_fmad.exit12.i:                         ; preds = %__nv_isinff.exit.i
  %75 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.08) #141, !srcloc !73
  %76 = fmul float %75, %75
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not72 = icmp eq i32 %77, 0
  %78 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC0105966E0000000, float %76, float noundef 0x3FE7FEFD20000000) #139
  %79 = call float @llvm.fma.f32(float %76, float noundef 0xC0105966E0000000, float noundef 0x3FE7FEFD20000000)
  %.025 = select i1 %.not72, float %79, float %78
  %80 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not73 = icmp eq i32 %80, 0
  %81 = call float @llvm.nvvm.fma.rn.ftz.f(float %.025, float %76, float noundef 0xBFC8B175C0000000) #139
  %82 = call float @llvm.fma.f32(float %.025, float %76, float noundef 0xBFC8B175C0000000)
  %.026 = select i1 %.not73, float %82, float %81
  %83 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not74 = icmp eq i32 %83, 0
  %84 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %76, float noundef 0x3FC7FFF160000000) #139
  %85 = call float @llvm.fma.f32(float %.026, float %76, float noundef 0x3FC7FFF160000000)
  %.027 = select i1 %.not74, float %85, float %84
  %86 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not75 = icmp eq i32 %86, 0
  %87 = call float @llvm.nvvm.fma.rn.ftz.f(float %.027, float %76, float noundef 1.000000e+00) #139
  %88 = call float @llvm.fma.f32(float %.027, float %76, float noundef 1.000000e+00)
  %.028 = select i1 %.not75, float %88, float %87
  %89 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not76 = icmp eq i32 %89, 0
  %90 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFF9477440000000, float %76, float noundef 0x3FD72295A0000000) #139
  %91 = call float @llvm.fma.f32(float %76, float noundef 0xBFF9477440000000, float noundef 0x3FD72295A0000000)
  %.029 = select i1 %.not76, float %91, float %90
  %92 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not77 = icmp eq i32 %92, 0
  %93 = call float @llvm.nvvm.fma.rn.ftz.f(float %.029, float %76, float noundef 0xBFC4FE5D80000000) #139
  %94 = call float @llvm.fma.f32(float %.029, float %76, float noundef 0xBFC4FE5D80000000)
  %.031 = select i1 %.not77, float %94, float %93
  %95 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not78 = icmp eq i32 %95, 0
  %96 = call float @llvm.nvvm.fma.rn.ftz.f(float %.031, float %76, float noundef 0x3FD7FFFFA0000000) #139
  %97 = call float @llvm.fma.f32(float %.031, float %76, float noundef 0x3FD7FFFFA0000000)
  %.032 = select i1 %.not78, float %97, float %96
  %98 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not79 = icmp eq i32 %98, 0
  %99 = call float @llvm.nvvm.fma.rn.ftz.f(float %.032, float %75, float %.08) #139
  %100 = call float @llvm.fma.f32(float %.032, float %75, float %.08)
  %.033 = select i1 %.not79, float %100, float %99
  %101 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not80 = icmp eq i32 %101, 0
  br i1 %.not80, label %104, label %102

102:                                              ; preds = %__internal_fmad.exit12.i
  %103 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.08) #139
  br label %__nv_rsqrtf.exit.i

104:                                              ; preds = %__internal_fmad.exit12.i
  %105 = call float @llvm.nvvm.rsqrt.approx.f(float %.08) #139
  br label %__nv_rsqrtf.exit.i

__nv_rsqrtf.exit.i:                               ; preds = %104, %102
  %.034 = phi float [ %103, %102 ], [ %105, %104 ]
  %106 = fmul float %.034, 0x3FE9884540000000
  %107 = fmul float %.028, %106
  %108 = fmul float %.033, 0x3FE45F3060000000
  %109 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not81 = icmp eq i32 %109, 0
  %110 = call i32 @llvm.nvvm.f2i.rn.ftz(float %108) #139
  %111 = call i32 @llvm.nvvm.f2i.rn(float %108) #139
  %.036 = select i1 %.not81, i32 %111, i32 %110
  %112 = sitofp i32 %.036 to float
  %113 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not82 = icmp eq i32 %113, 0
  %114 = call float @llvm.nvvm.fma.rn.ftz.f(float %112, float noundef 0xBFF921FB40000000, float %.033) #139
  %115 = call float @llvm.fma.f32(float %112, float noundef 0xBFF921FB40000000, float %.033)
  %.037 = select i1 %.not82, float %115, float %114
  %116 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not83 = icmp eq i32 %116, 0
  %117 = call float @llvm.nvvm.fma.rn.ftz.f(float %112, float noundef 0xBE74442D00000000, float %.037) #139
  %118 = call float @llvm.fma.f32(float %112, float noundef 0xBE74442D00000000, float %.037)
  %.038 = select i1 %.not83, float %118, float %117
  %119 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not84 = icmp eq i32 %119, 0
  %120 = call float @llvm.nvvm.fma.rn.ftz.f(float %112, float noundef 0xBCF84698A0000000, float %.038) #139
  %121 = call float @llvm.fma.f32(float %112, float noundef 0xBCF84698A0000000, float %.038)
  %.039 = select i1 %.not84, float %121, float %120
  %122 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not85 = icmp eq i32 %122, 0
  %123 = call float @llvm.nvvm.fabs.ftz.f(float %.033) #139
  %124 = call float @llvm.fabs.f32(float %.033)
  %.040 = select i1 %.not85, float %124, float %123
  %125 = fcmp ult float %.040, 1.056150e+05
  br i1 %125, label %__internal_trig_reduction_kernel.exit.i.i, label %__nv_isinff.exit.i.i.i

__nv_isinff.exit.i.i.i:                           ; preds = %__nv_rsqrtf.exit.i
  %126 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not116 = icmp eq i32 %126, 0
  %.042 = select i1 %.not116, float %124, float %123
  %127 = fcmp oeq float %.042, 0x7FF0000000000000
  br i1 %127, label %__nv_fmul_rn.exit.i.i.i, label %131

__nv_fmul_rn.exit.i.i.i:                          ; preds = %__nv_isinff.exit.i.i.i
  %128 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not122 = icmp eq i32 %128, 0
  %129 = call float @llvm.nvvm.mul.rn.ftz.f(float %.033, float noundef 0.000000e+00) #139
  %130 = fmul float %.033, 0.000000e+00
  %.043 = select i1 %.not122, float %130, float %129
  br label %__internal_trig_reduction_kernel.exit.i.i

131:                                              ; preds = %__nv_isinff.exit.i.i.i
  %132 = bitcast float %.033 to i32
  %133 = lshr i32 %132, 23
  %134 = and i32 %133, 255
  %135 = add nsw i32 %134, -128
  %136 = shl i32 %132, 8
  %137 = or i32 %136, -2147483648
  %138 = lshr i32 %135, 5
  br label %139

139:                                              ; preds = %131, %139
  %iq.i.i.i.i.0138 = phi i32 [ 0, %131 ], [ %147, %139 ]
  %hi.i.i.i.i.0137 = phi i32 [ 0, %131 ], [ %145, %139 ]
  %140 = zext i32 %iq.i.i.i.i.0138 to i64
  %141 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %140
  %142 = load i32, i32 addrspace(1)* %141, align 4
  %143 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %142, i32 %137, i32 %hi.i.i.i.i.0137) #141, !srcloc !67
  %144 = extractvalue { i32, i32 } %143, 0
  %145 = extractvalue { i32, i32 } %143, 1
  %146 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %140
  store i32 %144, i32* %146, align 4, !noalias !68
  %147 = add nuw nsw i32 %iq.i.i.i.i.0138, 1
  %exitcond143.not = icmp eq i32 %147, 6
  br i1 %exitcond143.not, label %148, label %139, !llvm.loop !71

148:                                              ; preds = %139
  %149 = and i32 %132, -2147483648
  %150 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %145, i32* %150, align 4, !noalias !68
  %151 = and i32 %133, 31
  %152 = sub nsw i32 6, %138
  %153 = sext i32 %152 to i64
  %154 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %153
  %155 = load i32, i32* %154, align 4
  %156 = sub nsw i32 5, %138
  %157 = sext i32 %156 to i64
  %158 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %157
  %159 = load i32, i32* %158, align 4
  %.not117 = icmp eq i32 %151, 0
  br i1 %.not117, label %__internal_trig_reduction_slowpath.exit.i.i.i, label %160

160:                                              ; preds = %148
  %161 = sub nsw i32 4, %138
  %162 = sub nuw nsw i32 32, %151
  %163 = shl i32 %155, %151
  %164 = lshr i32 %159, %162
  %165 = add i32 %164, %163
  %166 = shl i32 %159, %151
  %167 = sext i32 %161 to i64
  %168 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %167
  %169 = load i32, i32* %168, align 4
  %170 = lshr i32 %169, %162
  %171 = add i32 %170, %166
  br label %__internal_trig_reduction_slowpath.exit.i.i.i

__internal_trig_reduction_slowpath.exit.i.i.i:    ; preds = %160, %148
  %hi.i.i.i.i.1 = phi i32 [ %165, %160 ], [ %155, %148 ]
  %lo.i.i.i.i.0 = phi i32 [ %171, %160 ], [ %159, %148 ]
  %172 = lshr i32 %hi.i.i.i.i.1, 30
  %173 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.1, i32 %lo.i.i.i.i.0, i32 noundef 2)
  %174 = shl i32 %lo.i.i.i.i.0, 2
  %175 = lshr i32 %173, 31
  %176 = add nuw nsw i32 %175, %172
  %.not118 = icmp eq i32 %149, 0
  %177 = sub nsw i32 0, %176
  %spec.select = select i1 %.not118, i32 %176, i32 %177
  %.not119 = icmp sgt i32 %173, -1
  %178 = xor i32 %149, -2147483648
  %s.i.i.i.i.0 = select i1 %.not119, i32 %149, i32 %178
  %not..not119 = xor i1 %.not119, true
  %179 = sext i1 %not..not119 to i32
  %hi.i.i.i.i.2 = xor i32 %173, %179
  %lo.i.i.i.i.1 = xor i32 %174, %179
  %180 = zext i32 %hi.i.i.i.i.2 to i64
  %181 = shl nuw i64 %180, 32
  %182 = zext i32 %lo.i.i.i.i.1 to i64
  %183 = or i64 %181, %182
  %184 = sitofp i64 %183 to double
  %185 = fmul double %184, 0x3BF921FB54442D19
  %186 = fptrunc double %185 to float
  %.not121 = icmp eq i32 %s.i.i.i.i.0, 0
  %187 = fneg float %186
  %r.i.i.i.i.0 = select i1 %.not121, float %186, float %187
  br label %__internal_trig_reduction_kernel.exit.i.i

__internal_trig_reduction_kernel.exit.i.i:        ; preds = %__nv_fmul_rn.exit.i.i.i, %__internal_trig_reduction_slowpath.exit.i.i.i, %__nv_rsqrtf.exit.i
  %i.i.i.1 = phi i32 [ %.036, %__nv_rsqrtf.exit.i ], [ 0, %__nv_fmul_rn.exit.i.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %t.i.i.i.1 = phi float [ %.039, %__nv_rsqrtf.exit.i ], [ %.043, %__nv_fmul_rn.exit.i.i.i ], [ %r.i.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %188 = and i32 %i.i.i.1, 3
  %189 = uitofp i32 %188 to float
  %190 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not86 = icmp eq i32 %190, 0
  %191 = call float @llvm.nvvm.fma.rn.ftz.f(float %189, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %192 = call float @llvm.fma.f32(float %189, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.047 = select i1 %.not86, float %192, float %191
  %193 = fadd float %t.i.i.i.1, %.047
  %194 = fmul float %193, 0x3FE45F3060000000
  %195 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not87 = icmp eq i32 %195, 0
  %196 = call i32 @llvm.nvvm.f2i.rn.ftz(float %194) #139
  %197 = call i32 @llvm.nvvm.f2i.rn(float %194) #139
  %.048 = select i1 %.not87, i32 %197, i32 %196
  %198 = sitofp i32 %.048 to float
  %199 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not88 = icmp eq i32 %199, 0
  %200 = call float @llvm.nvvm.fma.rn.ftz.f(float %198, float noundef 0xBFF921FB40000000, float %193) #139
  %201 = call float @llvm.fma.f32(float %198, float noundef 0xBFF921FB40000000, float %193)
  %.049 = select i1 %.not88, float %201, float %200
  %202 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not89 = icmp eq i32 %202, 0
  %203 = call float @llvm.nvvm.fma.rn.ftz.f(float %198, float noundef 0xBE74442D00000000, float %.049) #139
  %204 = call float @llvm.fma.f32(float %198, float noundef 0xBE74442D00000000, float %.049)
  %.050 = select i1 %.not89, float %204, float %203
  %205 = add i32 %.048, 1
  %206 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not90 = icmp eq i32 %206, 0
  %207 = call float @llvm.nvvm.mul.rn.ftz.f(float %.050, float %.050) #139
  %208 = fmul float %.050, %.050
  %.051 = select i1 %.not90, float %208, float %207
  %209 = and i32 %205, 1
  %.not91 = icmp eq i32 %209, 0
  %210 = select i1 %.not91, float %.050, float 1.000000e+00
  %211 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not92 = icmp eq i32 %211, 0
  %212 = call float @llvm.nvvm.fma.rn.ftz.f(float %.051, float %210, float noundef 0.000000e+00) #139
  %213 = call float @llvm.fma.f32(float %.051, float %210, float noundef 0.000000e+00)
  %.052 = select i1 %.not92, float %213, float %212
  br i1 %.not91, label %__internal_fmad.exit2.i.i.i, label %__internal_fmad.exit1.i.i.i

__internal_fmad.exit1.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i
  %214 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not115 = icmp eq i32 %214, 0
  %215 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.051, float noundef 0xBF56C0FDA0000000) #139
  %216 = call float @llvm.fma.f32(float %.051, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.053 = select i1 %.not115, float %216, float %215
  br label %__internal_fmad.exit2.i.i.i

__internal_fmad.exit2.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i, %__internal_fmad.exit1.i.i.i
  %217 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i ]
  %218 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i ]
  %219 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i ], [ %.053, %__internal_fmad.exit1.i.i.i ]
  %220 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not95 = icmp eq i32 %220, 0
  %221 = call float @llvm.nvvm.fma.rn.ftz.f(float %219, float %.051, float noundef %218) #139
  %222 = call float @llvm.fma.f32(float %219, float %.051, float noundef %218)
  %.054 = select i1 %.not95, float %222, float %221
  %223 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not97 = icmp eq i32 %223, 0
  %224 = call float @llvm.nvvm.fma.rn.ftz.f(float %.054, float %.051, float noundef %217) #139
  %225 = call float @llvm.fma.f32(float %.054, float %.051, float noundef %217)
  %.055 = select i1 %.not97, float %225, float %224
  %226 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not98 = icmp eq i32 %226, 0
  %227 = call float @llvm.nvvm.fma.rn.ftz.f(float %.055, float %.052, float %210) #139
  %228 = call float @llvm.fma.f32(float %.055, float %.052, float %210)
  %.056 = select i1 %.not98, float %228, float %227
  %229 = and i32 %205, 2
  %.not99 = icmp eq i32 %229, 0
  br i1 %.not99, label %__internal_cos_offset_f.exit.i, label %__internal_fmad.exit5.i.i.i

__internal_fmad.exit5.i.i.i:                      ; preds = %__internal_fmad.exit2.i.i.i
  %230 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not114 = icmp eq i32 %230, 0
  %231 = call float @llvm.nvvm.fma.rn.ftz.f(float %.056, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %232 = call float @llvm.fma.f32(float %.056, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.057 = select i1 %.not114, float %232, float %231
  br label %__internal_cos_offset_f.exit.i

__internal_cos_offset_f.exit.i:                   ; preds = %__internal_fmad.exit5.i.i.i, %__internal_fmad.exit2.i.i.i
  %z.i.i.i.0 = phi float [ %.057, %__internal_fmad.exit5.i.i.i ], [ %.056, %__internal_fmad.exit2.i.i.i ]
  %233 = fmul float %107, %z.i.i.i.0
  br label %234

234:                                              ; preds = %__internal_cos_offset_f.exit.i, %__nv_isinff.exit.i, %__internal_fmad.exit.i
  %r.i.1 = phi float [ %70, %__internal_fmad.exit.i ], [ %233, %__internal_cos_offset_f.exit.i ], [ 0.000000e+00, %__nv_isinff.exit.i ]
  %235 = fcmp olt float %.01, 0.000000e+00
  %236 = fneg float %r.i.1
  %r.i.2 = select i1 %235, float %236, float %r.i.1
  %237 = fcmp olt float %.08, 0x39B4484C00000000
  br i1 %237, label %238, label %.__nv_j1f.exit_crit_edge

.__nv_j1f.exit_crit_edge:                         ; preds = %234
  %.pre = bitcast float %.01 to i32
  br label %__nv_j1f.exit

238:                                              ; preds = %234
  %239 = bitcast float %.01 to i32
  %240 = and i32 %239, -2147483648
  %241 = bitcast float %r.i.2 to i32
  %242 = and i32 %241, 2147483647
  %243 = or i32 %242, %240
  %244 = bitcast i32 %243 to float
  br label %__nv_j1f.exit

__nv_j1f.exit:                                    ; preds = %.__nv_j1f.exit_crit_edge, %238
  %.pre-phi = phi i32 [ %.pre, %.__nv_j1f.exit_crit_edge ], [ %239, %238 ]
  %r.i.3 = phi float [ %r.i.2, %.__nv_j1f.exit_crit_edge ], [ %244, %238 ]
  %245 = add i32 %.pre-phi, -1059760811
  %246 = and i32 %245, -8388608
  %247 = sub i32 %.pre-phi, %246
  %248 = bitcast i32 %247 to float
  %249 = sitofp i32 %246 to float
  %250 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not100 = icmp eq i32 %250, 0
  %251 = call float @llvm.nvvm.fma.rn.ftz.f(float %249, float noundef 0x3E80000000000000, float noundef 0.000000e+00) #139
  %252 = call float @llvm.fma.f32(float %249, float noundef 0x3E80000000000000, float noundef 0.000000e+00)
  %.064 = select i1 %.not100, float %252, float %251
  %253 = fadd float %248, -1.000000e+00
  %254 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not101 = icmp eq i32 %254, 0
  %255 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %253, float noundef 0x3FC2073EC0000000) #139
  %256 = call float @llvm.fma.f32(float %253, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.065 = select i1 %.not101, float %256, float %255
  %257 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not102 = icmp eq i32 %257, 0
  %258 = call float @llvm.nvvm.fma.rn.ftz.f(float %.065, float %253, float noundef 0xBFBF19B980000000) #139
  %259 = call float @llvm.fma.f32(float %.065, float %253, float noundef 0xBFBF19B980000000)
  %.066 = select i1 %.not102, float %259, float %258
  %260 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not103 = icmp eq i32 %260, 0
  %261 = call float @llvm.nvvm.fma.rn.ftz.f(float %.066, float %253, float noundef 0x3FC1E52AA0000000) #139
  %262 = call float @llvm.fma.f32(float %.066, float %253, float noundef 0x3FC1E52AA0000000)
  %.067 = select i1 %.not103, float %262, float %261
  %263 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not104 = icmp eq i32 %263, 0
  %264 = call float @llvm.nvvm.fma.rn.ftz.f(float %.067, float %253, float noundef 0xBFC55B1720000000) #139
  %265 = call float @llvm.fma.f32(float %.067, float %253, float noundef 0xBFC55B1720000000)
  %.069 = select i1 %.not104, float %265, float %264
  %266 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not105 = icmp eq i32 %266, 0
  %267 = call float @llvm.nvvm.fma.rn.ftz.f(float %.069, float %253, float noundef 0x3FC99DA160000000) #139
  %268 = call float @llvm.fma.f32(float %.069, float %253, float noundef 0x3FC99DA160000000)
  %.070 = select i1 %.not105, float %268, float %267
  %269 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not106 = icmp eq i32 %269, 0
  %270 = call float @llvm.nvvm.fma.rn.ftz.f(float %.070, float %253, float noundef 0xBFCFFFE440000000) #139
  %271 = call float @llvm.fma.f32(float %.070, float %253, float noundef 0xBFCFFFE440000000)
  %.072 = select i1 %.not106, float %271, float %270
  %272 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not107 = icmp eq i32 %272, 0
  %273 = call float @llvm.nvvm.fma.rn.ftz.f(float %.072, float %253, float noundef 0x3FD5554F00000000) #139
  %274 = call float @llvm.fma.f32(float %.072, float %253, float noundef 0x3FD5554F00000000)
  %.074 = select i1 %.not107, float %274, float %273
  %275 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not108 = icmp eq i32 %275, 0
  %276 = call float @llvm.nvvm.fma.rn.ftz.f(float %.074, float %253, float noundef -5.000000e-01) #139
  %277 = call float @llvm.fma.f32(float %.074, float %253, float noundef -5.000000e-01)
  %.062 = select i1 %.not108, float %277, float %276
  %278 = fmul float %253, %.062
  %279 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not109 = icmp eq i32 %279, 0
  %280 = call float @llvm.nvvm.fma.rn.ftz.f(float %278, float %253, float %253) #139
  %281 = call float @llvm.fma.f32(float %278, float %253, float %253)
  %.061 = select i1 %.not109, float %281, float %280
  %282 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not110 = icmp eq i32 %282, 0
  %283 = call float @llvm.nvvm.fma.rn.ftz.f(float %.064, float noundef 0x3FE62E4300000000, float %.061) #139
  %284 = call float @llvm.fma.f32(float %.064, float noundef 0x3FE62E4300000000, float %.061)
  %.060 = select i1 %.not110, float %284, float %283
  %285 = icmp ugt i32 %.pre-phi, 2139095039
  br i1 %285, label %__nv_fmaf_rn.exit.i, label %__internal_accurate_logf.exit

__nv_fmaf_rn.exit.i:                              ; preds = %__nv_j1f.exit
  %286 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not113 = icmp eq i32 %286, 0
  %287 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %288 = call float @llvm.fma.f32(float %.01, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.059 = select i1 %.not113, float %288, float %287
  br label %__internal_accurate_logf.exit

__internal_accurate_logf.exit:                    ; preds = %__nv_fmaf_rn.exit.i, %__nv_j1f.exit
  %r.i5.0 = phi float [ %.059, %__nv_fmaf_rn.exit.i ], [ %.060, %__nv_j1f.exit ]
  %289 = fcmp oeq float %.01, 0.000000e+00
  %r.i5.1 = select i1 %289, float 0xFFF0000000000000, float %r.i5.0
  %290 = fdiv float -1.000000e+00, %.01
  %291 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not111 = icmp eq i32 %291, 0
  %292 = call float @llvm.nvvm.fma.rn.ftz.f(float %r.i.3, float %r.i5.1, float %290) #139
  %293 = call float @llvm.fma.f32(float %r.i.3, float %r.i5.1, float %290)
  %.075 = select i1 %.not111, float %293, float %292
  %294 = fmul float %.075, 0x3FE45F3060000000
  %295 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not112 = icmp eq i32 %295, 0
  %296 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %.01, float %294) #139
  %297 = call float @llvm.fma.f32(float %.07, float %.01, float %294)
  %.076 = select i1 %.not112, float %297, float %296
  br label %551

298:                                              ; preds = %6
  %299 = fcmp ugt float %.01, 0x400E81DDC0000000
  br i1 %299, label %330, label %__internal_fmad.exit11

__internal_fmad.exit11:                           ; preds = %298
  %300 = fadd float %.01, 0xC00193BEE0000000
  %301 = fadd float %300, 0x3E76401B80000000
  %302 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not56 = icmp eq i32 %302, 0
  %303 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBEF13C53A0000000, float %301, float noundef 0x3F20C95100000000) #139
  %304 = call float @llvm.fma.f32(float %301, float noundef 0xBEF13C53A0000000, float noundef 0x3F20C95100000000)
  %.078 = select i1 %.not56, float %304, float %303
  %305 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not57 = icmp eq i32 %305, 0
  %306 = call float @llvm.nvvm.fma.rn.ftz.f(float %.078, float %301, float noundef 0xBF3E1561A0000000) #139
  %307 = call float @llvm.fma.f32(float %.078, float %301, float noundef 0xBF3E1561A0000000)
  %.079 = select i1 %.not57, float %307, float %306
  %308 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not58 = icmp eq i32 %308, 0
  %309 = call float @llvm.nvvm.fma.rn.ftz.f(float %.079, float %301, float noundef 0x3F51EC2040000000) #139
  %310 = call float @llvm.fma.f32(float %.079, float %301, float noundef 0x3F51EC2040000000)
  %.080 = select i1 %.not58, float %310, float %309
  %311 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not59 = icmp eq i32 %311, 0
  %312 = call float @llvm.nvvm.fma.rn.ftz.f(float %.080, float %301, float noundef 0xBF658E08A0000000) #139
  %313 = call float @llvm.fma.f32(float %.080, float %301, float noundef 0xBF658E08A0000000)
  %.082 = select i1 %.not59, float %313, float %312
  %314 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not60 = icmp eq i32 %314, 0
  %315 = call float @llvm.nvvm.fma.rn.ftz.f(float %.082, float %301, float noundef 0x3F7E6BBEE0000000) #139
  %316 = call float @llvm.fma.f32(float %.082, float %301, float noundef 0x3F7E6BBEE0000000)
  %.083 = select i1 %.not60, float %316, float %315
  %317 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not61 = icmp eq i32 %317, 0
  %318 = call float @llvm.nvvm.fma.rn.ftz.f(float %.083, float %301, float noundef 0xBF73A12F80000000) #139
  %319 = call float @llvm.fma.f32(float %.083, float %301, float noundef 0xBF73A12F80000000)
  %.084 = select i1 %.not61, float %319, float %318
  %320 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not62 = icmp eq i32 %320, 0
  %321 = call float @llvm.nvvm.fma.rn.ftz.f(float %.084, float %301, float noundef 0xBFA0D2D140000000) #139
  %322 = call float @llvm.fma.f32(float %.084, float %301, float noundef 0xBFA0D2D140000000)
  %.086 = select i1 %.not62, float %322, float %321
  %323 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not63 = icmp eq i32 %323, 0
  %324 = call float @llvm.nvvm.fma.rn.ftz.f(float %.086, float %301, float noundef 0xBFBE56FBE0000000) #139
  %325 = call float @llvm.fma.f32(float %.086, float %301, float noundef 0xBFBE56FBE0000000)
  %.087 = select i1 %.not63, float %325, float %324
  %326 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not64 = icmp eq i32 %326, 0
  %327 = call float @llvm.nvvm.fma.rn.ftz.f(float %.087, float %301, float noundef 0x3FE0AA4840000000) #139
  %328 = call float @llvm.fma.f32(float %.087, float %301, float noundef 0x3FE0AA4840000000)
  %.088 = select i1 %.not64, float %328, float %327
  %329 = fmul float %301, %.088
  br label %551

330:                                              ; preds = %298
  %331 = fcmp ugt float %.01, 0x401C0D26E0000000
  br i1 %331, label %362, label %__internal_fmad.exit29

__internal_fmad.exit29:                           ; preds = %330
  %332 = fadd float %.01, 0xC015B7FE40000000
  %333 = fadd float %332, 0xBE8D0F6060000000
  %334 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %334, 0
  %335 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E57CAFA00000000, float %333, float noundef 0xBEA11482C0000000) #139
  %336 = call float @llvm.fma.f32(float %333, float noundef 0x3E57CAFA00000000, float noundef 0xBEA11482C0000000)
  %.089 = select i1 %.not47, float %336, float %335
  %337 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %337, 0
  %338 = call float @llvm.nvvm.fma.rn.ftz.f(float %.089, float %333, float noundef 0xBEC7F14280000000) #139
  %339 = call float @llvm.fma.f32(float %.089, float %333, float noundef 0xBEC7F14280000000)
  %.090 = select i1 %.not48, float %339, float %338
  %340 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %340, 0
  %341 = call float @llvm.nvvm.fma.rn.ftz.f(float %.090, float %333, float noundef 0x3F084FC040000000) #139
  %342 = call float @llvm.fma.f32(float %.090, float %333, float noundef 0x3F084FC040000000)
  %.091 = select i1 %.not49, float %342, float %341
  %343 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %343, 0
  %344 = call float @llvm.nvvm.fma.rn.ftz.f(float %.091, float %333, float noundef 0x3F23376380000000) #139
  %345 = call float @llvm.fma.f32(float %.091, float %333, float noundef 0x3F23376380000000)
  %.093 = select i1 %.not50, float %345, float %344
  %346 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not51 = icmp eq i32 %346, 0
  %347 = call float @llvm.nvvm.fma.rn.ftz.f(float %.093, float %333, float noundef 0xBF61BE3FA0000000) #139
  %348 = call float @llvm.fma.f32(float %.093, float %333, float noundef 0xBF61BE3FA0000000)
  %.095 = select i1 %.not51, float %348, float %347
  %349 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not52 = icmp eq i32 %349, 0
  %350 = call float @llvm.nvvm.fma.rn.ftz.f(float %.095, float %333, float noundef 0xBF710A3120000000) #139
  %351 = call float @llvm.fma.f32(float %.095, float %333, float noundef 0xBF710A3120000000)
  %.096 = select i1 %.not52, float %351, float %350
  %352 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not53 = icmp eq i32 %352, 0
  %353 = call float @llvm.nvvm.fma.rn.ftz.f(float %.096, float %333, float noundef 0x3FAA15D820000000) #139
  %354 = call float @llvm.fma.f32(float %.096, float %333, float noundef 0x3FAA15D820000000)
  %.097 = select i1 %.not53, float %354, float %353
  %355 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not54 = icmp eq i32 %355, 0
  %356 = call float @llvm.nvvm.fma.rn.ftz.f(float %.097, float %333, float noundef 0x3FA00B9F80000000) #139
  %357 = call float @llvm.fma.f32(float %.097, float %333, float noundef 0x3FA00B9F80000000)
  %.098 = select i1 %.not54, float %357, float %356
  %358 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not55 = icmp eq i32 %358, 0
  %359 = call float @llvm.nvvm.fma.rn.ftz.f(float %.098, float %333, float noundef 0xBFD5C7C560000000) #139
  %360 = call float @llvm.fma.f32(float %.098, float %333, float noundef 0xBFD5C7C560000000)
  %.099 = select i1 %.not55, float %360, float %359
  %361 = fmul float %333, %.099
  br label %551

362:                                              ; preds = %330
  %363 = fcmp ugt float %.01, 0x4024585C60000000
  br i1 %363, label %__nv_isinff.exit, label %__internal_fmad.exit47

__internal_fmad.exit47:                           ; preds = %362
  %364 = fadd float %.01, 0xC0213127A0000000
  %365 = fadd float %364, 0xBE9CC2D360000000
  %366 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not39 = icmp eq i32 %366, 0
  %367 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EA19E7060000000, float %365, float noundef 0x3EC05205C0000000) #139
  %368 = call float @llvm.fma.f32(float %365, float noundef 0x3EA19E7060000000, float noundef 0x3EC05205C0000000)
  %.0100 = select i1 %.not39, float %368, float %367
  %369 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not40 = icmp eq i32 %369, 0
  %370 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0100, float %365, float noundef 0xBF06EBEE20000000) #139
  %371 = call float @llvm.fma.f32(float %.0100, float %365, float noundef 0xBF06EBEE20000000)
  %.0102 = select i1 %.not40, float %371, float %370
  %372 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not41 = icmp eq i32 %372, 0
  %373 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0102, float %365, float noundef 0xBF1B3F5500000000) #139
  %374 = call float @llvm.fma.f32(float %.0102, float %365, float noundef 0xBF1B3F5500000000)
  %.0103 = select i1 %.not41, float %374, float %373
  %375 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %375, 0
  %376 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0103, float %365, float noundef 0x3F607A3340000000) #139
  %377 = call float @llvm.fma.f32(float %.0103, float %365, float noundef 0x3F607A3340000000)
  %.0104 = select i1 %.not42, float %377, float %376
  %378 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %378, 0
  %379 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0104, float %365, float noundef 0x3F63CE6DA0000000) #139
  %380 = call float @llvm.fma.f32(float %.0104, float %365, float noundef 0x3F63CE6DA0000000)
  %.0105 = select i1 %.not43, float %380, float %379
  %381 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %381, 0
  %382 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0105, float %365, float noundef 0xBFA6395CA0000000) #139
  %383 = call float @llvm.fma.f32(float %.0105, float %365, float noundef 0xBFA6395CA0000000)
  %.0106 = select i1 %.not44, float %383, float %382
  %384 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %384, 0
  %385 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0106, float %365, float noundef 0xBF902B36C0000000) #139
  %386 = call float @llvm.fma.f32(float %.0106, float %365, float noundef 0xBF902B36C0000000)
  %.0107 = select i1 %.not45, float %386, float %385
  %387 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %387, 0
  %388 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0107, float %365, float noundef 0x3FD15F9940000000) #139
  %389 = call float @llvm.fma.f32(float %.0107, float %365, float noundef 0x3FD15F9940000000)
  %.0109 = select i1 %.not46, float %389, float %388
  %390 = fmul float %365, %.0109
  br label %551

__nv_isinff.exit:                                 ; preds = %362
  %391 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %391, 0
  %392 = call float @llvm.nvvm.fabs.ftz.f(float %.01) #139
  %393 = call float @llvm.fabs.f32(float %.01)
  %.0110 = select i1 %.not2, float %393, float %392
  %394 = fcmp oeq float %.0110, 0x7FF0000000000000
  br i1 %394, label %551, label %__internal_fmad.exit66

__internal_fmad.exit66:                           ; preds = %__nv_isinff.exit
  %395 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.01) #141, !srcloc !73
  %396 = fmul float %395, %395
  %397 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %397, 0
  %398 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FE4CFEC00000000, float %396, float noundef 0xBFC8957200000000) #139
  %399 = call float @llvm.fma.f32(float %396, float noundef 0x3FE4CFEC00000000, float noundef 0xBFC8957200000000)
  %.0111 = select i1 %.not3, float %399, float %398
  %400 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %400, 0
  %401 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0111, float %396, float noundef 0x3FC7FFD7E0000000) #139
  %402 = call float @llvm.fma.f32(float %.0111, float %396, float noundef 0x3FC7FFD7E0000000)
  %.0112 = select i1 %.not4, float %402, float %401
  %403 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %403, 0
  %404 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0112, float %396, float noundef 1.000000e+00) #139
  %405 = call float @llvm.fma.f32(float %.0112, float %396, float noundef 1.000000e+00)
  %.0113 = select i1 %.not5, float %405, float %404
  %406 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %406, 0
  %407 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFFC9C3560000000, float %396, float noundef 0x3FD76E7560000000) #139
  %408 = call float @llvm.fma.f32(float %396, float noundef 0xBFFC9C3560000000, float noundef 0x3FD76E7560000000)
  %.0114 = select i1 %.not6, float %408, float %407
  %409 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %409, 0
  %410 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0114, float %396, float noundef 0xBFC4FF6DC0000000) #139
  %411 = call float @llvm.fma.f32(float %.0114, float %396, float noundef 0xBFC4FF6DC0000000)
  %.0108 = select i1 %.not7, float %411, float %410
  %412 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %412, 0
  %413 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0108, float %396, float noundef 0x3FD7FFFFE0000000) #139
  %414 = call float @llvm.fma.f32(float %.0108, float %396, float noundef 0x3FD7FFFFE0000000)
  %.0101 = select i1 %.not8, float %414, float %413
  %415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %415, 0
  %416 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0101, float %395, float %.01) #139
  %417 = call float @llvm.fma.f32(float %.0101, float %395, float %.01)
  %.094 = select i1 %.not9, float %417, float %416
  %418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %418, 0
  br i1 %.not10, label %421, label %419

419:                                              ; preds = %__internal_fmad.exit66
  %420 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.01) #139
  br label %__nv_rsqrtf.exit

421:                                              ; preds = %__internal_fmad.exit66
  %422 = call float @llvm.nvvm.rsqrt.approx.f(float %.01) #139
  br label %__nv_rsqrtf.exit

__nv_rsqrtf.exit:                                 ; preds = %421, %419
  %.092 = phi float [ %420, %419 ], [ %422, %421 ]
  %423 = fmul float %.092, 0x3FE9884540000000
  %424 = fmul float %.0113, %423
  %425 = fmul float %.094, 0x3FE45F3060000000
  %426 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %426, 0
  %427 = call i32 @llvm.nvvm.f2i.rn.ftz(float %425) #139
  %428 = call i32 @llvm.nvvm.f2i.rn(float %425) #139
  %.085 = select i1 %.not11, i32 %428, i32 %427
  %429 = sitofp i32 %.085 to float
  %430 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %430, 0
  %431 = call float @llvm.nvvm.fma.rn.ftz.f(float %429, float noundef 0xBFF921FB40000000, float %.094) #139
  %432 = call float @llvm.fma.f32(float %429, float noundef 0xBFF921FB40000000, float %.094)
  %.081 = select i1 %.not12, float %432, float %431
  %433 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %433, 0
  %434 = call float @llvm.nvvm.fma.rn.ftz.f(float %429, float noundef 0xBE74442D00000000, float %.081) #139
  %435 = call float @llvm.fma.f32(float %429, float noundef 0xBE74442D00000000, float %.081)
  %.077 = select i1 %.not13, float %435, float %434
  %436 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %436, 0
  %437 = call float @llvm.nvvm.fma.rn.ftz.f(float %429, float noundef 0xBCF84698A0000000, float %.077) #139
  %438 = call float @llvm.fma.f32(float %429, float noundef 0xBCF84698A0000000, float %.077)
  %.073 = select i1 %.not14, float %438, float %437
  %439 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %439, 0
  %440 = call float @llvm.nvvm.fabs.ftz.f(float %.094) #139
  %441 = call float @llvm.fabs.f32(float %.094)
  %.071 = select i1 %.not15, float %441, float %440
  %442 = fcmp ult float %.071, 1.056150e+05
  br i1 %442, label %__internal_trig_reduction_kernel.exit.i, label %__nv_isinff.exit.i.i

__nv_isinff.exit.i.i:                             ; preds = %__nv_rsqrtf.exit
  %443 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %443, 0
  %.068 = select i1 %.not32, float %441, float %440
  %444 = fcmp oeq float %.068, 0x7FF0000000000000
  br i1 %444, label %__nv_fmul_rn.exit.i.i, label %448

__nv_fmul_rn.exit.i.i:                            ; preds = %__nv_isinff.exit.i.i
  %445 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not38 = icmp eq i32 %445, 0
  %446 = call float @llvm.nvvm.mul.rn.ftz.f(float %.094, float noundef 0.000000e+00) #139
  %447 = fmul float %.094, 0.000000e+00
  %.063 = select i1 %.not38, float %447, float %446
  br label %__internal_trig_reduction_kernel.exit.i

448:                                              ; preds = %__nv_isinff.exit.i.i
  %449 = bitcast float %.094 to i32
  %450 = lshr i32 %449, 23
  %451 = and i32 %450, 255
  %452 = add nsw i32 %451, -128
  %453 = shl i32 %449, 8
  %454 = or i32 %453, -2147483648
  %455 = lshr i32 %452, 5
  br label %456

456:                                              ; preds = %448, %456
  %hi.i.i.i.0136 = phi i32 [ 0, %448 ], [ %462, %456 ]
  %iq.i.i.i.0135 = phi i32 [ 0, %448 ], [ %464, %456 ]
  %457 = zext i32 %iq.i.i.i.0135 to i64
  %458 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %457
  %459 = load i32, i32 addrspace(1)* %458, align 4
  %460 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %459, i32 %454, i32 %hi.i.i.i.0136) #141, !srcloc !67
  %461 = extractvalue { i32, i32 } %460, 0
  %462 = extractvalue { i32, i32 } %460, 1
  %463 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %457
  store i32 %461, i32* %463, align 4, !noalias !68
  %464 = add nuw nsw i32 %iq.i.i.i.0135, 1
  %exitcond.not = icmp eq i32 %464, 6
  br i1 %exitcond.not, label %465, label %456, !llvm.loop !71

465:                                              ; preds = %456
  %466 = and i32 %449, -2147483648
  %467 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %462, i32* %467, align 4, !noalias !68
  %468 = and i32 %450, 31
  %469 = sub nsw i32 6, %455
  %470 = sext i32 %469 to i64
  %471 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %470
  %472 = load i32, i32* %471, align 4
  %473 = sub nsw i32 5, %455
  %474 = sext i32 %473 to i64
  %475 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %474
  %476 = load i32, i32* %475, align 4
  %.not33 = icmp eq i32 %468, 0
  br i1 %.not33, label %__internal_trig_reduction_slowpath.exit.i.i, label %477

477:                                              ; preds = %465
  %478 = sub nsw i32 4, %455
  %479 = sub nuw nsw i32 32, %468
  %480 = shl i32 %472, %468
  %481 = lshr i32 %476, %479
  %482 = add i32 %481, %480
  %483 = shl i32 %476, %468
  %484 = sext i32 %478 to i64
  %485 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %484
  %486 = load i32, i32* %485, align 4
  %487 = lshr i32 %486, %479
  %488 = add i32 %487, %483
  br label %__internal_trig_reduction_slowpath.exit.i.i

__internal_trig_reduction_slowpath.exit.i.i:      ; preds = %477, %465
  %lo.i.i.i.0 = phi i32 [ %488, %477 ], [ %476, %465 ]
  %hi.i.i.i.1 = phi i32 [ %482, %477 ], [ %472, %465 ]
  %489 = lshr i32 %hi.i.i.i.1, 30
  %490 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.1, i32 %lo.i.i.i.0, i32 noundef 2)
  %491 = shl i32 %lo.i.i.i.0, 2
  %492 = lshr i32 %490, 31
  %493 = add nuw nsw i32 %492, %489
  %.not34 = icmp eq i32 %466, 0
  %494 = sub nsw i32 0, %493
  %spec.select1 = select i1 %.not34, i32 %493, i32 %494
  %.not35 = icmp sgt i32 %490, -1
  %495 = xor i32 %466, -2147483648
  %not..not35 = xor i1 %.not35, true
  %496 = sext i1 %not..not35 to i32
  %lo.i.i.i.1 = xor i32 %491, %496
  %hi.i.i.i.2 = xor i32 %490, %496
  %s.i.i.i.0 = select i1 %.not35, i32 %466, i32 %495
  %497 = zext i32 %hi.i.i.i.2 to i64
  %498 = shl nuw i64 %497, 32
  %499 = zext i32 %lo.i.i.i.1 to i64
  %500 = or i64 %498, %499
  %501 = sitofp i64 %500 to double
  %502 = fmul double %501, 0x3BF921FB54442D19
  %503 = fptrunc double %502 to float
  %.not37 = icmp eq i32 %s.i.i.i.0, 0
  %504 = fneg float %503
  %r.i.i.i.0 = select i1 %.not37, float %503, float %504
  br label %__internal_trig_reduction_kernel.exit.i

__internal_trig_reduction_kernel.exit.i:          ; preds = %__nv_fmul_rn.exit.i.i, %__internal_trig_reduction_slowpath.exit.i.i, %__nv_rsqrtf.exit
  %i.i81.1 = phi i32 [ %.085, %__nv_rsqrtf.exit ], [ 0, %__nv_fmul_rn.exit.i.i ], [ %spec.select1, %__internal_trig_reduction_slowpath.exit.i.i ]
  %t.i.i.1 = phi float [ %.073, %__nv_rsqrtf.exit ], [ %.063, %__nv_fmul_rn.exit.i.i ], [ %r.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i ]
  %505 = and i32 %i.i81.1, 3
  %506 = uitofp i32 %505 to float
  %507 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %507, 0
  %508 = call float @llvm.nvvm.fma.rn.ftz.f(float %506, float noundef 0x3FF921FB60000000, float noundef 0xC00F6A7A20000000) #139
  %509 = call float @llvm.fma.f32(float %506, float noundef 0x3FF921FB60000000, float noundef 0xC00F6A7A20000000)
  %.046 = select i1 %.not16, float %509, float %508
  %510 = fadd float %t.i.i.1, %.046
  %511 = fmul float %510, 0x3FE45F3060000000
  %512 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %512, 0
  %513 = call i32 @llvm.nvvm.f2i.rn.ftz(float %511) #139
  %514 = call i32 @llvm.nvvm.f2i.rn(float %511) #139
  %.045 = select i1 %.not17, i32 %514, i32 %513
  %515 = sitofp i32 %.045 to float
  %516 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %516, 0
  %517 = call float @llvm.nvvm.fma.rn.ftz.f(float %515, float noundef 0xBFF921FB40000000, float %510) #139
  %518 = call float @llvm.fma.f32(float %515, float noundef 0xBFF921FB40000000, float %510)
  %.044 = select i1 %.not18, float %518, float %517
  %519 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %519, 0
  %520 = call float @llvm.nvvm.fma.rn.ftz.f(float %515, float noundef 0xBE74442D00000000, float %.044) #139
  %521 = call float @llvm.fma.f32(float %515, float noundef 0xBE74442D00000000, float %.044)
  %.041 = select i1 %.not19, float %521, float %520
  %522 = add i32 %.045, 1
  %523 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %523, 0
  %524 = call float @llvm.nvvm.mul.rn.ftz.f(float %.041, float %.041) #139
  %525 = fmul float %.041, %.041
  %.035 = select i1 %.not20, float %525, float %524
  %526 = and i32 %522, 1
  %.not21 = icmp eq i32 %526, 0
  %527 = select i1 %.not21, float %.041, float 1.000000e+00
  %528 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %528, 0
  %529 = call float @llvm.nvvm.fma.rn.ftz.f(float %.035, float %527, float noundef 0.000000e+00) #139
  %530 = call float @llvm.fma.f32(float %.035, float %527, float noundef 0.000000e+00)
  %.030 = select i1 %.not22, float %530, float %529
  br i1 %.not21, label %__internal_fmad.exit2.i.i, label %__internal_fmad.exit1.i.i

__internal_fmad.exit1.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i
  %531 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %531, 0
  %532 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.035, float noundef 0xBF56C0FDA0000000) #139
  %533 = call float @llvm.fma.f32(float %.035, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.023 = select i1 %.not31, float %533, float %532
  br label %__internal_fmad.exit2.i.i

__internal_fmad.exit2.i.i:                        ; preds = %__internal_trig_reduction_kernel.exit.i, %__internal_fmad.exit1.i.i
  %534 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i ]
  %535 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i ]
  %536 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i ], [ %.023, %__internal_fmad.exit1.i.i ]
  %537 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %537, 0
  %538 = call float @llvm.nvvm.fma.rn.ftz.f(float %536, float %.035, float noundef %535) #139
  %539 = call float @llvm.fma.f32(float %536, float %.035, float noundef %535)
  %.016 = select i1 %.not25, float %539, float %538
  %540 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %540, 0
  %541 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %.035, float noundef %534) #139
  %542 = call float @llvm.fma.f32(float %.016, float %.035, float noundef %534)
  %.09 = select i1 %.not27, float %542, float %541
  %543 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %543, 0
  %544 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %.030, float %527) #139
  %545 = call float @llvm.fma.f32(float %.09, float %.030, float %527)
  %.05 = select i1 %.not28, float %545, float %544
  %546 = and i32 %522, 2
  %.not29 = icmp eq i32 %546, 0
  br i1 %.not29, label %__internal_cos_offset_f.exit, label %__internal_fmad.exit5.i.i

__internal_fmad.exit5.i.i:                        ; preds = %__internal_fmad.exit2.i.i
  %547 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %547, 0
  %548 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %549 = call float @llvm.fma.f32(float %.05, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0 = select i1 %.not30, float %549, float %548
  br label %__internal_cos_offset_f.exit

__internal_cos_offset_f.exit:                     ; preds = %__internal_fmad.exit5.i.i, %__internal_fmad.exit2.i.i
  %z.i.i79.0 = phi float [ %.0, %__internal_fmad.exit5.i.i ], [ %.05, %__internal_fmad.exit2.i.i ]
  %550 = fmul float %424, %z.i.i79.0
  br label %551

551:                                              ; preds = %__internal_accurate_logf.exit, %__internal_fmad.exit29, %__internal_cos_offset_f.exit, %__nv_isinff.exit, %__internal_fmad.exit47, %__internal_fmad.exit11, %4
  %r.5 = phi float [ %5, %4 ], [ %.076, %__internal_accurate_logf.exit ], [ %329, %__internal_fmad.exit11 ], [ %361, %__internal_fmad.exit29 ], [ %390, %__internal_fmad.exit47 ], [ %550, %__internal_cos_offset_f.exit ], [ 0.000000e+00, %__nv_isinff.exit ]
  %552 = fcmp olt float %a, 0.000000e+00
  %r.6 = select i1 %552, float 0x7FFFFFFFE0000000, float %r.5
  ret float %r.6
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_ynf(i32 %n, float %a) local_unnamed_addr #47 {
__nv_fabsf.exit.i:
  %result.i.i.i.i.i = alloca [7 x i32], align 4
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.03 = select i1 %.not, float %2, float %1
  %3 = fcmp olt float %.03, 0x3810000000000000
  br i1 %3, label %4, label %6

4:                                                ; preds = %__nv_fabsf.exit.i
  %5 = fdiv float 0xBFE45F3060000000, %.03
  br label %__nv_y1f.exit

6:                                                ; preds = %__nv_fabsf.exit.i
  %7 = fcmp ugt float %.03, 0x3FFB2D5880000000
  br i1 %7, label %298, label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %6
  %8 = fmul float %.03, %.03
  %9 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not483 = icmp eq i32 %9, 0
  %10 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E428C5980000000, float %8, float noundef 0xBEB0A4FB40000000) #139
  %11 = call float @llvm.fma.f32(float %8, float noundef 0x3E428C5980000000, float noundef 0xBEB0A4FB40000000)
  %.04 = select i1 %.not483, float %11, float %10
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not484 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %8, float noundef 0x3F12C7D2A0000000) #139
  %14 = call float @llvm.fma.f32(float %.04, float %8, float noundef 0x3F12C7D2A0000000)
  %.05 = select i1 %.not484, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not485 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %8, float noundef 0xBF6835B960000000) #139
  %17 = call float @llvm.fma.f32(float %.05, float %8, float noundef 0xBF6835B960000000)
  %.07 = select i1 %.not485, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not486 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %8, float noundef 0x3FABD39760000000) #139
  %20 = call float @llvm.fma.f32(float %.07, float %8, float noundef 0x3FABD39760000000)
  %.09 = select i1 %.not486, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not487 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %8, float noundef 0xBFC9186620000000) #139
  %23 = call float @llvm.fma.f32(float %.09, float %8, float noundef 0xBFC9186620000000)
  %.010 = select i1 %.not487, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not488 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fabs.ftz.f(float %.03) #139
  %26 = call float @llvm.fabs.f32(float %.03)
  %.011 = select i1 %.not488, float %26, float %25
  %27 = fcmp ugt float %.011, 8.000000e+00
  br i1 %27, label %__nv_isinff.exit.i.i, label %__internal_fmad.exit.i.i

__internal_fmad.exit.i.i:                         ; preds = %__internal_fmad.exit.i
  %28 = fadd float %.011, 0xC00EA75580000000
  %29 = fadd float %28, 0x3E74A121E0000000
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not541 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3D35E68C60000000, float %29, float noundef 0x3D7037E840000000) #139
  %32 = call float @llvm.fma.f32(float %29, float noundef 0x3D35E68C60000000, float noundef 0x3D7037E840000000)
  %.012 = select i1 %.not541, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not542 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %29, float noundef 0xBDBC43D820000000) #139
  %35 = call float @llvm.fma.f32(float %.012, float %29, float noundef 0xBDBC43D820000000)
  %.014 = select i1 %.not542, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not543 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %29, float noundef 0xBDEBBBDFE0000000) #139
  %38 = call float @llvm.fma.f32(float %.014, float %29, float noundef 0xBDEBBBDFE0000000)
  %.015 = select i1 %.not543, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not544 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %29, float noundef 0x3E336193A0000000) #139
  %41 = call float @llvm.fma.f32(float %.015, float %29, float noundef 0x3E336193A0000000)
  %.016 = select i1 %.not544, float %41, float %40
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not545 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %29, float noundef 0x3E5D022E60000000) #139
  %44 = call float @llvm.fma.f32(float %.016, float %29, float noundef 0x3E5D022E60000000)
  %.017 = select i1 %.not545, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not546 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %29, float noundef 0xBEA1F1B900000000) #139
  %47 = call float @llvm.fma.f32(float %.017, float %29, float noundef 0xBEA1F1B900000000)
  %.018 = select i1 %.not546, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not547 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %29, float noundef 0xBEC3CCA7A0000000) #139
  %50 = call float @llvm.fma.f32(float %.018, float %29, float noundef 0xBEC3CCA7A0000000)
  %.019 = select i1 %.not547, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not548 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %29, float noundef 0x3F059B38A0000000) #139
  %53 = call float @llvm.fma.f32(float %.019, float %29, float noundef 0x3F059B38A0000000)
  %.021 = select i1 %.not548, float %53, float %52
  %54 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not549 = icmp eq i32 %54, 0
  %55 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %29, float noundef 0x3F1F3D6200000000) #139
  %56 = call float @llvm.fma.f32(float %.021, float %29, float noundef 0x3F1F3D6200000000)
  %.022 = select i1 %.not549, float %56, float %55
  %57 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not550 = icmp eq i32 %57, 0
  %58 = call float @llvm.nvvm.fma.rn.ftz.f(float %.022, float %29, float noundef 0xBF5D9D7380000000) #139
  %59 = call float @llvm.fma.f32(float %.022, float %29, float noundef 0xBF5D9D7380000000)
  %.023 = select i1 %.not550, float %59, float %58
  %60 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not551 = icmp eq i32 %60, 0
  %61 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float %29, float noundef 0xBF64EDFFA0000000) #139
  %62 = call float @llvm.fma.f32(float %.023, float %29, float noundef 0xBF64EDFFA0000000)
  %.024 = select i1 %.not551, float %62, float %61
  %63 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not552 = icmp eq i32 %63, 0
  %64 = call float @llvm.nvvm.fma.rn.ftz.f(float %.024, float %29, float noundef 0x3FA0E73260000000) #139
  %65 = call float @llvm.fma.f32(float %.024, float %29, float noundef 0x3FA0E73260000000)
  %.025 = select i1 %.not552, float %65, float %64
  %66 = fadd float %.011, 0xC01C0FF600000000
  %67 = fadd float %66, 0x3E88971B60000000
  %68 = fmul float %67, %.025
  %69 = fmul float %29, %68
  %70 = fmul float %.011, %69
  br label %234

__nv_isinff.exit.i.i:                             ; preds = %__internal_fmad.exit.i
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not489 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fabs.ftz.f(float %.011) #139
  %73 = call float @llvm.fabs.f32(float %.011)
  %.026 = select i1 %.not489, float %73, float %72
  %74 = fcmp oeq float %.026, 0x7FF0000000000000
  br i1 %74, label %234, label %__internal_fmad.exit12.i.i

__internal_fmad.exit12.i.i:                       ; preds = %__nv_isinff.exit.i.i
  %75 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.011) #141, !srcloc !73
  %76 = fmul float %75, %75
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not490 = icmp eq i32 %77, 0
  %78 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC0105966E0000000, float %76, float noundef 0x3FE7FEFD20000000) #139
  %79 = call float @llvm.fma.f32(float %76, float noundef 0xC0105966E0000000, float noundef 0x3FE7FEFD20000000)
  %.028 = select i1 %.not490, float %79, float %78
  %80 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not491 = icmp eq i32 %80, 0
  %81 = call float @llvm.nvvm.fma.rn.ftz.f(float %.028, float %76, float noundef 0xBFC8B175C0000000) #139
  %82 = call float @llvm.fma.f32(float %.028, float %76, float noundef 0xBFC8B175C0000000)
  %.029 = select i1 %.not491, float %82, float %81
  %83 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not492 = icmp eq i32 %83, 0
  %84 = call float @llvm.nvvm.fma.rn.ftz.f(float %.029, float %76, float noundef 0x3FC7FFF160000000) #139
  %85 = call float @llvm.fma.f32(float %.029, float %76, float noundef 0x3FC7FFF160000000)
  %.030 = select i1 %.not492, float %85, float %84
  %86 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not493 = icmp eq i32 %86, 0
  %87 = call float @llvm.nvvm.fma.rn.ftz.f(float %.030, float %76, float noundef 1.000000e+00) #139
  %88 = call float @llvm.fma.f32(float %.030, float %76, float noundef 1.000000e+00)
  %.031 = select i1 %.not493, float %88, float %87
  %89 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not494 = icmp eq i32 %89, 0
  %90 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFF9477440000000, float %76, float noundef 0x3FD72295A0000000) #139
  %91 = call float @llvm.fma.f32(float %76, float noundef 0xBFF9477440000000, float noundef 0x3FD72295A0000000)
  %.032 = select i1 %.not494, float %91, float %90
  %92 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not495 = icmp eq i32 %92, 0
  %93 = call float @llvm.nvvm.fma.rn.ftz.f(float %.032, float %76, float noundef 0xBFC4FE5D80000000) #139
  %94 = call float @llvm.fma.f32(float %.032, float %76, float noundef 0xBFC4FE5D80000000)
  %.033 = select i1 %.not495, float %94, float %93
  %95 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not496 = icmp eq i32 %95, 0
  %96 = call float @llvm.nvvm.fma.rn.ftz.f(float %.033, float %76, float noundef 0x3FD7FFFFA0000000) #139
  %97 = call float @llvm.fma.f32(float %.033, float %76, float noundef 0x3FD7FFFFA0000000)
  %.035 = select i1 %.not496, float %97, float %96
  %98 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not497 = icmp eq i32 %98, 0
  %99 = call float @llvm.nvvm.fma.rn.ftz.f(float %.035, float %75, float %.011) #139
  %100 = call float @llvm.fma.f32(float %.035, float %75, float %.011)
  %.036 = select i1 %.not497, float %100, float %99
  %101 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not498 = icmp eq i32 %101, 0
  br i1 %.not498, label %104, label %102

102:                                              ; preds = %__internal_fmad.exit12.i.i
  %103 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.011) #139
  br label %__nv_rsqrtf.exit.i.i

104:                                              ; preds = %__internal_fmad.exit12.i.i
  %105 = call float @llvm.nvvm.rsqrt.approx.f(float %.011) #139
  br label %__nv_rsqrtf.exit.i.i

__nv_rsqrtf.exit.i.i:                             ; preds = %104, %102
  %.037 = phi float [ %103, %102 ], [ %105, %104 ]
  %106 = fmul float %.037, 0x3FE9884540000000
  %107 = fmul float %.031, %106
  %108 = fmul float %.036, 0x3FE45F3060000000
  %109 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not499 = icmp eq i32 %109, 0
  %110 = call i32 @llvm.nvvm.f2i.rn.ftz(float %108) #139
  %111 = call i32 @llvm.nvvm.f2i.rn(float %108) #139
  %.039 = select i1 %.not499, i32 %111, i32 %110
  %112 = sitofp i32 %.039 to float
  %113 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not500 = icmp eq i32 %113, 0
  %114 = call float @llvm.nvvm.fma.rn.ftz.f(float %112, float noundef 0xBFF921FB40000000, float %.036) #139
  %115 = call float @llvm.fma.f32(float %112, float noundef 0xBFF921FB40000000, float %.036)
  %.040 = select i1 %.not500, float %115, float %114
  %116 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not501 = icmp eq i32 %116, 0
  %117 = call float @llvm.nvvm.fma.rn.ftz.f(float %112, float noundef 0xBE74442D00000000, float %.040) #139
  %118 = call float @llvm.fma.f32(float %112, float noundef 0xBE74442D00000000, float %.040)
  %.041 = select i1 %.not501, float %118, float %117
  %119 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not502 = icmp eq i32 %119, 0
  %120 = call float @llvm.nvvm.fma.rn.ftz.f(float %112, float noundef 0xBCF84698A0000000, float %.041) #139
  %121 = call float @llvm.fma.f32(float %112, float noundef 0xBCF84698A0000000, float %.041)
  %.042 = select i1 %.not502, float %121, float %120
  %122 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not503 = icmp eq i32 %122, 0
  %123 = call float @llvm.nvvm.fabs.ftz.f(float %.036) #139
  %124 = call float @llvm.fabs.f32(float %.036)
  %.044 = select i1 %.not503, float %124, float %123
  %125 = fcmp ult float %.044, 1.056150e+05
  br i1 %125, label %__internal_trig_reduction_kernel.exit.i.i.i, label %__nv_isinff.exit.i.i.i.i

__nv_isinff.exit.i.i.i.i:                         ; preds = %__nv_rsqrtf.exit.i.i
  %126 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not534 = icmp eq i32 %126, 0
  %.045 = select i1 %.not534, float %124, float %123
  %127 = fcmp oeq float %.045, 0x7FF0000000000000
  br i1 %127, label %__nv_fmul_rn.exit.i.i.i.i, label %131

__nv_fmul_rn.exit.i.i.i.i:                        ; preds = %__nv_isinff.exit.i.i.i.i
  %128 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not540 = icmp eq i32 %128, 0
  %129 = call float @llvm.nvvm.mul.rn.ftz.f(float %.036, float noundef 0.000000e+00) #139
  %130 = fmul float %.036, 0.000000e+00
  %.046 = select i1 %.not540, float %130, float %129
  br label %__internal_trig_reduction_kernel.exit.i.i.i

131:                                              ; preds = %__nv_isinff.exit.i.i.i.i
  %132 = bitcast float %.036 to i32
  %133 = lshr i32 %132, 23
  %134 = and i32 %133, 255
  %135 = add nsw i32 %134, -128
  %136 = shl i32 %132, 8
  %137 = or i32 %136, -2147483648
  %138 = lshr i32 %135, 5
  br label %139

139:                                              ; preds = %131, %139
  %iq.i.i.i.i.i.0571 = phi i32 [ 0, %131 ], [ %147, %139 ]
  %hi.i.i.i.i.i.0570 = phi i32 [ 0, %131 ], [ %145, %139 ]
  %140 = zext i32 %iq.i.i.i.i.i.0571 to i64
  %141 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %140
  %142 = load i32, i32 addrspace(1)* %141, align 4
  %143 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %142, i32 %137, i32 %hi.i.i.i.i.i.0570) #141, !srcloc !67
  %144 = extractvalue { i32, i32 } %143, 0
  %145 = extractvalue { i32, i32 } %143, 1
  %146 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %140
  store i32 %144, i32* %146, align 4, !noalias !68
  %147 = add nuw nsw i32 %iq.i.i.i.i.i.0571, 1
  %exitcond601.not = icmp eq i32 %147, 6
  br i1 %exitcond601.not, label %148, label %139, !llvm.loop !71

148:                                              ; preds = %139
  %149 = and i32 %132, -2147483648
  %150 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %145, i32* %150, align 4, !noalias !68
  %151 = and i32 %133, 31
  %152 = sub nsw i32 6, %138
  %153 = sext i32 %152 to i64
  %154 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %153
  %155 = load i32, i32* %154, align 4
  %156 = sub nsw i32 5, %138
  %157 = sext i32 %156 to i64
  %158 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %157
  %159 = load i32, i32* %158, align 4
  %.not535 = icmp eq i32 %151, 0
  br i1 %.not535, label %__internal_trig_reduction_slowpath.exit.i.i.i.i, label %160

160:                                              ; preds = %148
  %161 = sub nsw i32 4, %138
  %162 = sub nuw nsw i32 32, %151
  %163 = shl i32 %155, %151
  %164 = lshr i32 %159, %162
  %165 = add i32 %164, %163
  %166 = shl i32 %159, %151
  %167 = sext i32 %161 to i64
  %168 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %167
  %169 = load i32, i32* %168, align 4
  %170 = lshr i32 %169, %162
  %171 = add i32 %170, %166
  br label %__internal_trig_reduction_slowpath.exit.i.i.i.i

__internal_trig_reduction_slowpath.exit.i.i.i.i:  ; preds = %160, %148
  %hi.i.i.i.i.i.1 = phi i32 [ %165, %160 ], [ %155, %148 ]
  %lo.i.i.i.i.i.0 = phi i32 [ %171, %160 ], [ %159, %148 ]
  %172 = lshr i32 %hi.i.i.i.i.i.1, 30
  %173 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.i.1, i32 %lo.i.i.i.i.i.0, i32 noundef 2)
  %174 = shl i32 %lo.i.i.i.i.i.0, 2
  %175 = lshr i32 %173, 31
  %176 = add nuw nsw i32 %175, %172
  %.not536 = icmp eq i32 %149, 0
  %177 = sub nsw i32 0, %176
  %spec.select = select i1 %.not536, i32 %176, i32 %177
  %.not537 = icmp sgt i32 %173, -1
  %178 = xor i32 %149, -2147483648
  %s.i.i.i.i.i.0 = select i1 %.not537, i32 %149, i32 %178
  %not..not537 = xor i1 %.not537, true
  %179 = sext i1 %not..not537 to i32
  %hi.i.i.i.i.i.2 = xor i32 %173, %179
  %lo.i.i.i.i.i.1 = xor i32 %174, %179
  %180 = zext i32 %hi.i.i.i.i.i.2 to i64
  %181 = shl nuw i64 %180, 32
  %182 = zext i32 %lo.i.i.i.i.i.1 to i64
  %183 = or i64 %181, %182
  %184 = sitofp i64 %183 to double
  %185 = fmul double %184, 0x3BF921FB54442D19
  %186 = fptrunc double %185 to float
  %.not539 = icmp eq i32 %s.i.i.i.i.i.0, 0
  %187 = fneg float %186
  %r.i.i.i.i.i.0 = select i1 %.not539, float %186, float %187
  br label %__internal_trig_reduction_kernel.exit.i.i.i

__internal_trig_reduction_kernel.exit.i.i.i:      ; preds = %__nv_fmul_rn.exit.i.i.i.i, %__internal_trig_reduction_slowpath.exit.i.i.i.i, %__nv_rsqrtf.exit.i.i
  %i.i.i.i.1 = phi i32 [ %.039, %__nv_rsqrtf.exit.i.i ], [ 0, %__nv_fmul_rn.exit.i.i.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i.i.i ]
  %t.i.i.i.i.1 = phi float [ %.042, %__nv_rsqrtf.exit.i.i ], [ %.046, %__nv_fmul_rn.exit.i.i.i.i ], [ %r.i.i.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i.i.i ]
  %188 = and i32 %i.i.i.i.1, 3
  %189 = uitofp i32 %188 to float
  %190 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not504 = icmp eq i32 %190, 0
  %191 = call float @llvm.nvvm.fma.rn.ftz.f(float %189, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %192 = call float @llvm.fma.f32(float %189, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.048 = select i1 %.not504, float %192, float %191
  %193 = fadd float %t.i.i.i.i.1, %.048
  %194 = fmul float %193, 0x3FE45F3060000000
  %195 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not505 = icmp eq i32 %195, 0
  %196 = call i32 @llvm.nvvm.f2i.rn.ftz(float %194) #139
  %197 = call i32 @llvm.nvvm.f2i.rn(float %194) #139
  %.049 = select i1 %.not505, i32 %197, i32 %196
  %198 = sitofp i32 %.049 to float
  %199 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not506 = icmp eq i32 %199, 0
  %200 = call float @llvm.nvvm.fma.rn.ftz.f(float %198, float noundef 0xBFF921FB40000000, float %193) #139
  %201 = call float @llvm.fma.f32(float %198, float noundef 0xBFF921FB40000000, float %193)
  %.050 = select i1 %.not506, float %201, float %200
  %202 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not507 = icmp eq i32 %202, 0
  %203 = call float @llvm.nvvm.fma.rn.ftz.f(float %198, float noundef 0xBE74442D00000000, float %.050) #139
  %204 = call float @llvm.fma.f32(float %198, float noundef 0xBE74442D00000000, float %.050)
  %.052 = select i1 %.not507, float %204, float %203
  %205 = add i32 %.049, 1
  %206 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not508 = icmp eq i32 %206, 0
  %207 = call float @llvm.nvvm.mul.rn.ftz.f(float %.052, float %.052) #139
  %208 = fmul float %.052, %.052
  %.053 = select i1 %.not508, float %208, float %207
  %209 = and i32 %205, 1
  %.not509 = icmp eq i32 %209, 0
  %210 = select i1 %.not509, float %.052, float 1.000000e+00
  %211 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not510 = icmp eq i32 %211, 0
  %212 = call float @llvm.nvvm.fma.rn.ftz.f(float %.053, float %210, float noundef 0.000000e+00) #139
  %213 = call float @llvm.fma.f32(float %.053, float %210, float noundef 0.000000e+00)
  %.055 = select i1 %.not510, float %213, float %212
  br i1 %.not509, label %__internal_fmad.exit2.i.i.i.i, label %__internal_fmad.exit1.i.i.i.i

__internal_fmad.exit1.i.i.i.i:                    ; preds = %__internal_trig_reduction_kernel.exit.i.i.i
  %214 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not533 = icmp eq i32 %214, 0
  %215 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.053, float noundef 0xBF56C0FDA0000000) #139
  %216 = call float @llvm.fma.f32(float %.053, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.056 = select i1 %.not533, float %216, float %215
  br label %__internal_fmad.exit2.i.i.i.i

__internal_fmad.exit2.i.i.i.i:                    ; preds = %__internal_trig_reduction_kernel.exit.i.i.i, %__internal_fmad.exit1.i.i.i.i
  %217 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i.i ]
  %218 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i.i ]
  %219 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i.i ], [ %.056, %__internal_fmad.exit1.i.i.i.i ]
  %220 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not513 = icmp eq i32 %220, 0
  %221 = call float @llvm.nvvm.fma.rn.ftz.f(float %219, float %.053, float noundef %218) #139
  %222 = call float @llvm.fma.f32(float %219, float %.053, float noundef %218)
  %.057 = select i1 %.not513, float %222, float %221
  %223 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not515 = icmp eq i32 %223, 0
  %224 = call float @llvm.nvvm.fma.rn.ftz.f(float %.057, float %.053, float noundef %217) #139
  %225 = call float @llvm.fma.f32(float %.057, float %.053, float noundef %217)
  %.059 = select i1 %.not515, float %225, float %224
  %226 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not516 = icmp eq i32 %226, 0
  %227 = call float @llvm.nvvm.fma.rn.ftz.f(float %.059, float %.055, float %210) #139
  %228 = call float @llvm.fma.f32(float %.059, float %.055, float %210)
  %.061 = select i1 %.not516, float %228, float %227
  %229 = and i32 %205, 2
  %.not517 = icmp eq i32 %229, 0
  br i1 %.not517, label %__internal_cos_offset_f.exit.i.i, label %__internal_fmad.exit5.i.i.i.i

__internal_fmad.exit5.i.i.i.i:                    ; preds = %__internal_fmad.exit2.i.i.i.i
  %230 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not532 = icmp eq i32 %230, 0
  %231 = call float @llvm.nvvm.fma.rn.ftz.f(float %.061, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %232 = call float @llvm.fma.f32(float %.061, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.062 = select i1 %.not532, float %232, float %231
  br label %__internal_cos_offset_f.exit.i.i

__internal_cos_offset_f.exit.i.i:                 ; preds = %__internal_fmad.exit5.i.i.i.i, %__internal_fmad.exit2.i.i.i.i
  %z.i.i.i.i.0 = phi float [ %.062, %__internal_fmad.exit5.i.i.i.i ], [ %.061, %__internal_fmad.exit2.i.i.i.i ]
  %233 = fmul float %107, %z.i.i.i.i.0
  br label %234

234:                                              ; preds = %__internal_cos_offset_f.exit.i.i, %__nv_isinff.exit.i.i, %__internal_fmad.exit.i.i
  %r.i.i.1 = phi float [ %70, %__internal_fmad.exit.i.i ], [ %233, %__internal_cos_offset_f.exit.i.i ], [ 0.000000e+00, %__nv_isinff.exit.i.i ]
  %235 = fcmp olt float %.03, 0.000000e+00
  %236 = fneg float %r.i.i.1
  %r.i.i.2 = select i1 %235, float %236, float %r.i.i.1
  %237 = fcmp olt float %.011, 0x39B4484C00000000
  br i1 %237, label %238, label %.__nv_j1f.exit.i_crit_edge

.__nv_j1f.exit.i_crit_edge:                       ; preds = %234
  %.pre602 = bitcast float %.03 to i32
  br label %__nv_j1f.exit.i

238:                                              ; preds = %234
  %239 = bitcast float %.03 to i32
  %240 = and i32 %239, -2147483648
  %241 = bitcast float %r.i.i.2 to i32
  %242 = and i32 %241, 2147483647
  %243 = or i32 %242, %240
  %244 = bitcast i32 %243 to float
  br label %__nv_j1f.exit.i

__nv_j1f.exit.i:                                  ; preds = %.__nv_j1f.exit.i_crit_edge, %238
  %.pre-phi603 = phi i32 [ %.pre602, %.__nv_j1f.exit.i_crit_edge ], [ %239, %238 ]
  %r.i.i.3 = phi float [ %r.i.i.2, %.__nv_j1f.exit.i_crit_edge ], [ %244, %238 ]
  %245 = add i32 %.pre-phi603, -1059760811
  %246 = and i32 %245, -8388608
  %247 = sub i32 %.pre-phi603, %246
  %248 = bitcast i32 %247 to float
  %249 = sitofp i32 %246 to float
  %250 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not518 = icmp eq i32 %250, 0
  %251 = call float @llvm.nvvm.fma.rn.ftz.f(float %249, float noundef 0x3E80000000000000, float noundef 0.000000e+00) #139
  %252 = call float @llvm.fma.f32(float %249, float noundef 0x3E80000000000000, float noundef 0.000000e+00)
  %.068 = select i1 %.not518, float %252, float %251
  %253 = fadd float %248, -1.000000e+00
  %254 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not519 = icmp eq i32 %254, 0
  %255 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %253, float noundef 0x3FC2073EC0000000) #139
  %256 = call float @llvm.fma.f32(float %253, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.069 = select i1 %.not519, float %256, float %255
  %257 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not520 = icmp eq i32 %257, 0
  %258 = call float @llvm.nvvm.fma.rn.ftz.f(float %.069, float %253, float noundef 0xBFBF19B980000000) #139
  %259 = call float @llvm.fma.f32(float %.069, float %253, float noundef 0xBFBF19B980000000)
  %.070 = select i1 %.not520, float %259, float %258
  %260 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not521 = icmp eq i32 %260, 0
  %261 = call float @llvm.nvvm.fma.rn.ftz.f(float %.070, float %253, float noundef 0x3FC1E52AA0000000) #139
  %262 = call float @llvm.fma.f32(float %.070, float %253, float noundef 0x3FC1E52AA0000000)
  %.071 = select i1 %.not521, float %262, float %261
  %263 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not522 = icmp eq i32 %263, 0
  %264 = call float @llvm.nvvm.fma.rn.ftz.f(float %.071, float %253, float noundef 0xBFC55B1720000000) #139
  %265 = call float @llvm.fma.f32(float %.071, float %253, float noundef 0xBFC55B1720000000)
  %.072 = select i1 %.not522, float %265, float %264
  %266 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not523 = icmp eq i32 %266, 0
  %267 = call float @llvm.nvvm.fma.rn.ftz.f(float %.072, float %253, float noundef 0x3FC99DA160000000) #139
  %268 = call float @llvm.fma.f32(float %.072, float %253, float noundef 0x3FC99DA160000000)
  %.073 = select i1 %.not523, float %268, float %267
  %269 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not524 = icmp eq i32 %269, 0
  %270 = call float @llvm.nvvm.fma.rn.ftz.f(float %.073, float %253, float noundef 0xBFCFFFE440000000) #139
  %271 = call float @llvm.fma.f32(float %.073, float %253, float noundef 0xBFCFFFE440000000)
  %.074 = select i1 %.not524, float %271, float %270
  %272 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not525 = icmp eq i32 %272, 0
  %273 = call float @llvm.nvvm.fma.rn.ftz.f(float %.074, float %253, float noundef 0x3FD5554F00000000) #139
  %274 = call float @llvm.fma.f32(float %.074, float %253, float noundef 0x3FD5554F00000000)
  %.075 = select i1 %.not525, float %274, float %273
  %275 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not526 = icmp eq i32 %275, 0
  %276 = call float @llvm.nvvm.fma.rn.ftz.f(float %.075, float %253, float noundef -5.000000e-01) #139
  %277 = call float @llvm.fma.f32(float %.075, float %253, float noundef -5.000000e-01)
  %.067 = select i1 %.not526, float %277, float %276
  %278 = fmul float %253, %.067
  %279 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not527 = icmp eq i32 %279, 0
  %280 = call float @llvm.nvvm.fma.rn.ftz.f(float %278, float %253, float %253) #139
  %281 = call float @llvm.fma.f32(float %278, float %253, float %253)
  %.066 = select i1 %.not527, float %281, float %280
  %282 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not528 = icmp eq i32 %282, 0
  %283 = call float @llvm.nvvm.fma.rn.ftz.f(float %.068, float noundef 0x3FE62E4300000000, float %.066) #139
  %284 = call float @llvm.fma.f32(float %.068, float noundef 0x3FE62E4300000000, float %.066)
  %.065 = select i1 %.not528, float %284, float %283
  %285 = icmp ugt i32 %.pre-phi603, 2139095039
  br i1 %285, label %__nv_fmaf_rn.exit.i.i, label %__internal_accurate_logf.exit.i

__nv_fmaf_rn.exit.i.i:                            ; preds = %__nv_j1f.exit.i
  %286 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not531 = icmp eq i32 %286, 0
  %287 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %288 = call float @llvm.fma.f32(float %.03, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.064 = select i1 %.not531, float %288, float %287
  br label %__internal_accurate_logf.exit.i

__internal_accurate_logf.exit.i:                  ; preds = %__nv_fmaf_rn.exit.i.i, %__nv_j1f.exit.i
  %r.i5.i.0 = phi float [ %.064, %__nv_fmaf_rn.exit.i.i ], [ %.065, %__nv_j1f.exit.i ]
  %289 = fcmp oeq float %.03, 0.000000e+00
  %r.i5.i.1 = select i1 %289, float 0xFFF0000000000000, float %r.i5.i.0
  %290 = fdiv float -1.000000e+00, %.03
  %291 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not529 = icmp eq i32 %291, 0
  %292 = call float @llvm.nvvm.fma.rn.ftz.f(float %r.i.i.3, float %r.i5.i.1, float %290) #139
  %293 = call float @llvm.fma.f32(float %r.i.i.3, float %r.i5.i.1, float %290)
  %.076 = select i1 %.not529, float %293, float %292
  %294 = fmul float %.076, 0x3FE45F3060000000
  %295 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not530 = icmp eq i32 %295, 0
  %296 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %.03, float %294) #139
  %297 = call float @llvm.fma.f32(float %.010, float %.03, float %294)
  %.077 = select i1 %.not530, float %297, float %296
  br label %__nv_y1f.exit

298:                                              ; preds = %6
  %299 = fcmp ugt float %.03, 0x400E81DDC0000000
  br i1 %299, label %330, label %__internal_fmad.exit11.i

__internal_fmad.exit11.i:                         ; preds = %298
  %300 = fadd float %.03, 0xC00193BEE0000000
  %301 = fadd float %300, 0x3E76401B80000000
  %302 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not474 = icmp eq i32 %302, 0
  %303 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBEF13C53A0000000, float %301, float noundef 0x3F20C95100000000) #139
  %304 = call float @llvm.fma.f32(float %301, float noundef 0xBEF13C53A0000000, float noundef 0x3F20C95100000000)
  %.078 = select i1 %.not474, float %304, float %303
  %305 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not475 = icmp eq i32 %305, 0
  %306 = call float @llvm.nvvm.fma.rn.ftz.f(float %.078, float %301, float noundef 0xBF3E1561A0000000) #139
  %307 = call float @llvm.fma.f32(float %.078, float %301, float noundef 0xBF3E1561A0000000)
  %.079 = select i1 %.not475, float %307, float %306
  %308 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not476 = icmp eq i32 %308, 0
  %309 = call float @llvm.nvvm.fma.rn.ftz.f(float %.079, float %301, float noundef 0x3F51EC2040000000) #139
  %310 = call float @llvm.fma.f32(float %.079, float %301, float noundef 0x3F51EC2040000000)
  %.080 = select i1 %.not476, float %310, float %309
  %311 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not477 = icmp eq i32 %311, 0
  %312 = call float @llvm.nvvm.fma.rn.ftz.f(float %.080, float %301, float noundef 0xBF658E08A0000000) #139
  %313 = call float @llvm.fma.f32(float %.080, float %301, float noundef 0xBF658E08A0000000)
  %.082 = select i1 %.not477, float %313, float %312
  %314 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not478 = icmp eq i32 %314, 0
  %315 = call float @llvm.nvvm.fma.rn.ftz.f(float %.082, float %301, float noundef 0x3F7E6BBEE0000000) #139
  %316 = call float @llvm.fma.f32(float %.082, float %301, float noundef 0x3F7E6BBEE0000000)
  %.083 = select i1 %.not478, float %316, float %315
  %317 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not479 = icmp eq i32 %317, 0
  %318 = call float @llvm.nvvm.fma.rn.ftz.f(float %.083, float %301, float noundef 0xBF73A12F80000000) #139
  %319 = call float @llvm.fma.f32(float %.083, float %301, float noundef 0xBF73A12F80000000)
  %.084 = select i1 %.not479, float %319, float %318
  %320 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not480 = icmp eq i32 %320, 0
  %321 = call float @llvm.nvvm.fma.rn.ftz.f(float %.084, float %301, float noundef 0xBFA0D2D140000000) #139
  %322 = call float @llvm.fma.f32(float %.084, float %301, float noundef 0xBFA0D2D140000000)
  %.085 = select i1 %.not480, float %322, float %321
  %323 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not481 = icmp eq i32 %323, 0
  %324 = call float @llvm.nvvm.fma.rn.ftz.f(float %.085, float %301, float noundef 0xBFBE56FBE0000000) #139
  %325 = call float @llvm.fma.f32(float %.085, float %301, float noundef 0xBFBE56FBE0000000)
  %.087 = select i1 %.not481, float %325, float %324
  %326 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not482 = icmp eq i32 %326, 0
  %327 = call float @llvm.nvvm.fma.rn.ftz.f(float %.087, float %301, float noundef 0x3FE0AA4840000000) #139
  %328 = call float @llvm.fma.f32(float %.087, float %301, float noundef 0x3FE0AA4840000000)
  %.088 = select i1 %.not482, float %328, float %327
  %329 = fmul float %301, %.088
  br label %__nv_y1f.exit

330:                                              ; preds = %298
  %331 = fcmp ugt float %.03, 0x401C0D26E0000000
  br i1 %331, label %362, label %__internal_fmad.exit29.i

__internal_fmad.exit29.i:                         ; preds = %330
  %332 = fadd float %.03, 0xC015B7FE40000000
  %333 = fadd float %332, 0xBE8D0F6060000000
  %334 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not465 = icmp eq i32 %334, 0
  %335 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E57CAFA00000000, float %333, float noundef 0xBEA11482C0000000) #139
  %336 = call float @llvm.fma.f32(float %333, float noundef 0x3E57CAFA00000000, float noundef 0xBEA11482C0000000)
  %.090 = select i1 %.not465, float %336, float %335
  %337 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not466 = icmp eq i32 %337, 0
  %338 = call float @llvm.nvvm.fma.rn.ftz.f(float %.090, float %333, float noundef 0xBEC7F14280000000) #139
  %339 = call float @llvm.fma.f32(float %.090, float %333, float noundef 0xBEC7F14280000000)
  %.092 = select i1 %.not466, float %339, float %338
  %340 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not467 = icmp eq i32 %340, 0
  %341 = call float @llvm.nvvm.fma.rn.ftz.f(float %.092, float %333, float noundef 0x3F084FC040000000) #139
  %342 = call float @llvm.fma.f32(float %.092, float %333, float noundef 0x3F084FC040000000)
  %.093 = select i1 %.not467, float %342, float %341
  %343 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not468 = icmp eq i32 %343, 0
  %344 = call float @llvm.nvvm.fma.rn.ftz.f(float %.093, float %333, float noundef 0x3F23376380000000) #139
  %345 = call float @llvm.fma.f32(float %.093, float %333, float noundef 0x3F23376380000000)
  %.094 = select i1 %.not468, float %345, float %344
  %346 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not469 = icmp eq i32 %346, 0
  %347 = call float @llvm.nvvm.fma.rn.ftz.f(float %.094, float %333, float noundef 0xBF61BE3FA0000000) #139
  %348 = call float @llvm.fma.f32(float %.094, float %333, float noundef 0xBF61BE3FA0000000)
  %.096 = select i1 %.not469, float %348, float %347
  %349 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not470 = icmp eq i32 %349, 0
  %350 = call float @llvm.nvvm.fma.rn.ftz.f(float %.096, float %333, float noundef 0xBF710A3120000000) #139
  %351 = call float @llvm.fma.f32(float %.096, float %333, float noundef 0xBF710A3120000000)
  %.097 = select i1 %.not470, float %351, float %350
  %352 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not471 = icmp eq i32 %352, 0
  %353 = call float @llvm.nvvm.fma.rn.ftz.f(float %.097, float %333, float noundef 0x3FAA15D820000000) #139
  %354 = call float @llvm.fma.f32(float %.097, float %333, float noundef 0x3FAA15D820000000)
  %.098 = select i1 %.not471, float %354, float %353
  %355 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not472 = icmp eq i32 %355, 0
  %356 = call float @llvm.nvvm.fma.rn.ftz.f(float %.098, float %333, float noundef 0x3FA00B9F80000000) #139
  %357 = call float @llvm.fma.f32(float %.098, float %333, float noundef 0x3FA00B9F80000000)
  %.0100 = select i1 %.not472, float %357, float %356
  %358 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not473 = icmp eq i32 %358, 0
  %359 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0100, float %333, float noundef 0xBFD5C7C560000000) #139
  %360 = call float @llvm.fma.f32(float %.0100, float %333, float noundef 0xBFD5C7C560000000)
  %.0101 = select i1 %.not473, float %360, float %359
  %361 = fmul float %333, %.0101
  br label %__nv_y1f.exit

362:                                              ; preds = %330
  %363 = fcmp ugt float %.03, 0x4024585C60000000
  br i1 %363, label %__nv_isinff.exit.i, label %__internal_fmad.exit47.i

__internal_fmad.exit47.i:                         ; preds = %362
  %364 = fadd float %.03, 0xC0213127A0000000
  %365 = fadd float %364, 0xBE9CC2D360000000
  %366 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not457 = icmp eq i32 %366, 0
  %367 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EA19E7060000000, float %365, float noundef 0x3EC05205C0000000) #139
  %368 = call float @llvm.fma.f32(float %365, float noundef 0x3EA19E7060000000, float noundef 0x3EC05205C0000000)
  %.0102 = select i1 %.not457, float %368, float %367
  %369 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not458 = icmp eq i32 %369, 0
  %370 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0102, float %365, float noundef 0xBF06EBEE20000000) #139
  %371 = call float @llvm.fma.f32(float %.0102, float %365, float noundef 0xBF06EBEE20000000)
  %.0104 = select i1 %.not458, float %371, float %370
  %372 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not459 = icmp eq i32 %372, 0
  %373 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0104, float %365, float noundef 0xBF1B3F5500000000) #139
  %374 = call float @llvm.fma.f32(float %.0104, float %365, float noundef 0xBF1B3F5500000000)
  %.0105 = select i1 %.not459, float %374, float %373
  %375 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not460 = icmp eq i32 %375, 0
  %376 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0105, float %365, float noundef 0x3F607A3340000000) #139
  %377 = call float @llvm.fma.f32(float %.0105, float %365, float noundef 0x3F607A3340000000)
  %.0106 = select i1 %.not460, float %377, float %376
  %378 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not461 = icmp eq i32 %378, 0
  %379 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0106, float %365, float noundef 0x3F63CE6DA0000000) #139
  %380 = call float @llvm.fma.f32(float %.0106, float %365, float noundef 0x3F63CE6DA0000000)
  %.0107 = select i1 %.not461, float %380, float %379
  %381 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not462 = icmp eq i32 %381, 0
  %382 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0107, float %365, float noundef 0xBFA6395CA0000000) #139
  %383 = call float @llvm.fma.f32(float %.0107, float %365, float noundef 0xBFA6395CA0000000)
  %.0108 = select i1 %.not462, float %383, float %382
  %384 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not463 = icmp eq i32 %384, 0
  %385 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0108, float %365, float noundef 0xBF902B36C0000000) #139
  %386 = call float @llvm.fma.f32(float %.0108, float %365, float noundef 0xBF902B36C0000000)
  %.0109 = select i1 %.not463, float %386, float %385
  %387 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not464 = icmp eq i32 %387, 0
  %388 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0109, float %365, float noundef 0x3FD15F9940000000) #139
  %389 = call float @llvm.fma.f32(float %.0109, float %365, float noundef 0x3FD15F9940000000)
  %.0111 = select i1 %.not464, float %389, float %388
  %390 = fmul float %365, %.0111
  br label %__nv_y1f.exit

__nv_isinff.exit.i:                               ; preds = %362
  %391 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %391, 0
  %392 = call float @llvm.nvvm.fabs.ftz.f(float %.03) #139
  %393 = call float @llvm.fabs.f32(float %.03)
  %.0113 = select i1 %.not8, float %393, float %392
  %394 = fcmp oeq float %.0113, 0x7FF0000000000000
  br i1 %394, label %__nv_y1f.exit, label %__internal_fmad.exit66.i

__internal_fmad.exit66.i:                         ; preds = %__nv_isinff.exit.i
  %395 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.03) #141, !srcloc !73
  %396 = fmul float %395, %395
  %397 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %397, 0
  %398 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FE4CFEC00000000, float %396, float noundef 0xBFC8957200000000) #139
  %399 = call float @llvm.fma.f32(float %396, float noundef 0x3FE4CFEC00000000, float noundef 0xBFC8957200000000)
  %.0114 = select i1 %.not9, float %399, float %398
  %400 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %400, 0
  %401 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0114, float %396, float noundef 0x3FC7FFD7E0000000) #139
  %402 = call float @llvm.fma.f32(float %.0114, float %396, float noundef 0x3FC7FFD7E0000000)
  %.0115 = select i1 %.not10, float %402, float %401
  %403 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %403, 0
  %404 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0115, float %396, float noundef 1.000000e+00) #139
  %405 = call float @llvm.fma.f32(float %.0115, float %396, float noundef 1.000000e+00)
  %.0116 = select i1 %.not11, float %405, float %404
  %406 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %406, 0
  %407 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFFC9C3560000000, float %396, float noundef 0x3FD76E7560000000) #139
  %408 = call float @llvm.fma.f32(float %396, float noundef 0xBFFC9C3560000000, float noundef 0x3FD76E7560000000)
  %.0117 = select i1 %.not12, float %408, float %407
  %409 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %409, 0
  %410 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0117, float %396, float noundef 0xBFC4FF6DC0000000) #139
  %411 = call float @llvm.fma.f32(float %.0117, float %396, float noundef 0xBFC4FF6DC0000000)
  %.0118 = select i1 %.not13, float %411, float %410
  %412 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %412, 0
  %413 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0118, float %396, float noundef 0x3FD7FFFFE0000000) #139
  %414 = call float @llvm.fma.f32(float %.0118, float %396, float noundef 0x3FD7FFFFE0000000)
  %.0120 = select i1 %.not14, float %414, float %413
  %415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %415, 0
  %416 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0120, float %395, float %.03) #139
  %417 = call float @llvm.fma.f32(float %.0120, float %395, float %.03)
  %.0121 = select i1 %.not15, float %417, float %416
  %418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %418, 0
  br i1 %.not16, label %421, label %419

419:                                              ; preds = %__internal_fmad.exit66.i
  %420 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.03) #139
  br label %__nv_rsqrtf.exit.i

421:                                              ; preds = %__internal_fmad.exit66.i
  %422 = call float @llvm.nvvm.rsqrt.approx.f(float %.03) #139
  br label %__nv_rsqrtf.exit.i

__nv_rsqrtf.exit.i:                               ; preds = %421, %419
  %.0122 = phi float [ %420, %419 ], [ %422, %421 ]
  %423 = fmul float %.0122, 0x3FE9884540000000
  %424 = fmul float %.0116, %423
  %425 = fmul float %.0121, 0x3FE45F3060000000
  %426 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %426, 0
  %427 = call i32 @llvm.nvvm.f2i.rn.ftz(float %425) #139
  %428 = call i32 @llvm.nvvm.f2i.rn(float %425) #139
  %.0124 = select i1 %.not17, i32 %428, i32 %427
  %429 = sitofp i32 %.0124 to float
  %430 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %430, 0
  %431 = call float @llvm.nvvm.fma.rn.ftz.f(float %429, float noundef 0xBFF921FB40000000, float %.0121) #139
  %432 = call float @llvm.fma.f32(float %429, float noundef 0xBFF921FB40000000, float %.0121)
  %.0125 = select i1 %.not18, float %432, float %431
  %433 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %433, 0
  %434 = call float @llvm.nvvm.fma.rn.ftz.f(float %429, float noundef 0xBE74442D00000000, float %.0125) #139
  %435 = call float @llvm.fma.f32(float %429, float noundef 0xBE74442D00000000, float %.0125)
  %.0126 = select i1 %.not19, float %435, float %434
  %436 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %436, 0
  %437 = call float @llvm.nvvm.fma.rn.ftz.f(float %429, float noundef 0xBCF84698A0000000, float %.0126) #139
  %438 = call float @llvm.fma.f32(float %429, float noundef 0xBCF84698A0000000, float %.0126)
  %.0127 = select i1 %.not20, float %438, float %437
  %439 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %439, 0
  %440 = call float @llvm.nvvm.fabs.ftz.f(float %.0121) #139
  %441 = call float @llvm.fabs.f32(float %.0121)
  %.0129 = select i1 %.not21, float %441, float %440
  %442 = fcmp ult float %.0129, 1.056150e+05
  br i1 %442, label %__internal_trig_reduction_kernel.exit.i.i, label %__nv_isinff.exit.i.i.i

__nv_isinff.exit.i.i.i:                           ; preds = %__nv_rsqrtf.exit.i
  %443 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not450 = icmp eq i32 %443, 0
  %.0130 = select i1 %.not450, float %441, float %440
  %444 = fcmp oeq float %.0130, 0x7FF0000000000000
  br i1 %444, label %__nv_fmul_rn.exit.i.i.i, label %448

__nv_fmul_rn.exit.i.i.i:                          ; preds = %__nv_isinff.exit.i.i.i
  %445 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not456 = icmp eq i32 %445, 0
  %446 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0121, float noundef 0.000000e+00) #139
  %447 = fmul float %.0121, 0.000000e+00
  %.0131 = select i1 %.not456, float %447, float %446
  br label %__internal_trig_reduction_kernel.exit.i.i

448:                                              ; preds = %__nv_isinff.exit.i.i.i
  %449 = bitcast float %.0121 to i32
  %450 = lshr i32 %449, 23
  %451 = and i32 %450, 255
  %452 = add nsw i32 %451, -128
  %453 = shl i32 %449, 8
  %454 = or i32 %453, -2147483648
  %455 = lshr i32 %452, 5
  br label %456

456:                                              ; preds = %448, %456
  %iq.i.i.i.i.0569 = phi i32 [ 0, %448 ], [ %464, %456 ]
  %hi.i.i.i.i.0568 = phi i32 [ 0, %448 ], [ %462, %456 ]
  %457 = zext i32 %iq.i.i.i.i.0569 to i64
  %458 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %457
  %459 = load i32, i32 addrspace(1)* %458, align 4
  %460 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %459, i32 %454, i32 %hi.i.i.i.i.0568) #141, !srcloc !67
  %461 = extractvalue { i32, i32 } %460, 0
  %462 = extractvalue { i32, i32 } %460, 1
  %463 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %457
  store i32 %461, i32* %463, align 4, !noalias !68
  %464 = add nuw nsw i32 %iq.i.i.i.i.0569, 1
  %exitcond600.not = icmp eq i32 %464, 6
  br i1 %exitcond600.not, label %465, label %456, !llvm.loop !71

465:                                              ; preds = %456
  %466 = and i32 %449, -2147483648
  %467 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %462, i32* %467, align 4, !noalias !68
  %468 = and i32 %450, 31
  %469 = sub nsw i32 6, %455
  %470 = sext i32 %469 to i64
  %471 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %470
  %472 = load i32, i32* %471, align 4
  %473 = sub nsw i32 5, %455
  %474 = sext i32 %473 to i64
  %475 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %474
  %476 = load i32, i32* %475, align 4
  %.not451 = icmp eq i32 %468, 0
  br i1 %.not451, label %__internal_trig_reduction_slowpath.exit.i.i.i, label %477

477:                                              ; preds = %465
  %478 = sub nsw i32 4, %455
  %479 = sub nuw nsw i32 32, %468
  %480 = shl i32 %472, %468
  %481 = lshr i32 %476, %479
  %482 = add i32 %481, %480
  %483 = shl i32 %476, %468
  %484 = sext i32 %478 to i64
  %485 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %484
  %486 = load i32, i32* %485, align 4
  %487 = lshr i32 %486, %479
  %488 = add i32 %487, %483
  br label %__internal_trig_reduction_slowpath.exit.i.i.i

__internal_trig_reduction_slowpath.exit.i.i.i:    ; preds = %477, %465
  %hi.i.i.i.i.1 = phi i32 [ %482, %477 ], [ %472, %465 ]
  %lo.i.i.i.i.0 = phi i32 [ %488, %477 ], [ %476, %465 ]
  %489 = lshr i32 %hi.i.i.i.i.1, 30
  %490 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.1, i32 %lo.i.i.i.i.0, i32 noundef 2)
  %491 = shl i32 %lo.i.i.i.i.0, 2
  %492 = lshr i32 %490, 31
  %493 = add nuw nsw i32 %492, %489
  %.not452 = icmp eq i32 %466, 0
  %494 = sub nsw i32 0, %493
  %spec.select1 = select i1 %.not452, i32 %493, i32 %494
  %.not453 = icmp sgt i32 %490, -1
  %495 = xor i32 %466, -2147483648
  %s.i.i.i.i.0 = select i1 %.not453, i32 %466, i32 %495
  %not..not453 = xor i1 %.not453, true
  %496 = sext i1 %not..not453 to i32
  %hi.i.i.i.i.2 = xor i32 %490, %496
  %lo.i.i.i.i.1 = xor i32 %491, %496
  %497 = zext i32 %hi.i.i.i.i.2 to i64
  %498 = shl nuw i64 %497, 32
  %499 = zext i32 %lo.i.i.i.i.1 to i64
  %500 = or i64 %498, %499
  %501 = sitofp i64 %500 to double
  %502 = fmul double %501, 0x3BF921FB54442D19
  %503 = fptrunc double %502 to float
  %.not455 = icmp eq i32 %s.i.i.i.i.0, 0
  %504 = fneg float %503
  %r.i.i.i.i.0 = select i1 %.not455, float %503, float %504
  br label %__internal_trig_reduction_kernel.exit.i.i

__internal_trig_reduction_kernel.exit.i.i:        ; preds = %__nv_fmul_rn.exit.i.i.i, %__internal_trig_reduction_slowpath.exit.i.i.i, %__nv_rsqrtf.exit.i
  %i.i81.i.1 = phi i32 [ %.0124, %__nv_rsqrtf.exit.i ], [ 0, %__nv_fmul_rn.exit.i.i.i ], [ %spec.select1, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %t.i.i.i.1 = phi float [ %.0127, %__nv_rsqrtf.exit.i ], [ %.0131, %__nv_fmul_rn.exit.i.i.i ], [ %r.i.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %505 = and i32 %i.i81.i.1, 3
  %506 = uitofp i32 %505 to float
  %507 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %507, 0
  %508 = call float @llvm.nvvm.fma.rn.ftz.f(float %506, float noundef 0x3FF921FB60000000, float noundef 0xC00F6A7A20000000) #139
  %509 = call float @llvm.fma.f32(float %506, float noundef 0x3FF921FB60000000, float noundef 0xC00F6A7A20000000)
  %.0133 = select i1 %.not22, float %509, float %508
  %510 = fadd float %t.i.i.i.1, %.0133
  %511 = fmul float %510, 0x3FE45F3060000000
  %512 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %512, 0
  %513 = call i32 @llvm.nvvm.f2i.rn.ftz(float %511) #139
  %514 = call i32 @llvm.nvvm.f2i.rn(float %511) #139
  %.0135 = select i1 %.not23, i32 %514, i32 %513
  %515 = sitofp i32 %.0135 to float
  %516 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %516, 0
  %517 = call float @llvm.nvvm.fma.rn.ftz.f(float %515, float noundef 0xBFF921FB40000000, float %510) #139
  %518 = call float @llvm.fma.f32(float %515, float noundef 0xBFF921FB40000000, float %510)
  %.0136 = select i1 %.not24, float %518, float %517
  %519 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %519, 0
  %520 = call float @llvm.nvvm.fma.rn.ftz.f(float %515, float noundef 0xBE74442D00000000, float %.0136) #139
  %521 = call float @llvm.fma.f32(float %515, float noundef 0xBE74442D00000000, float %.0136)
  %.0137 = select i1 %.not25, float %521, float %520
  %522 = add i32 %.0135, 1
  %523 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not26 = icmp eq i32 %523, 0
  %524 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0137, float %.0137) #139
  %525 = fmul float %.0137, %.0137
  %.0138 = select i1 %.not26, float %525, float %524
  %526 = and i32 %522, 1
  %.not27 = icmp eq i32 %526, 0
  %527 = select i1 %.not27, float %.0137, float 1.000000e+00
  %528 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %528, 0
  %529 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0138, float %527, float noundef 0.000000e+00) #139
  %530 = call float @llvm.fma.f32(float %.0138, float %527, float noundef 0.000000e+00)
  %.0139 = select i1 %.not28, float %530, float %529
  br i1 %.not27, label %__internal_fmad.exit2.i.i.i, label %__internal_fmad.exit1.i.i.i

__internal_fmad.exit1.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i
  %531 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not449 = icmp eq i32 %531, 0
  %532 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0138, float noundef 0xBF56C0FDA0000000) #139
  %533 = call float @llvm.fma.f32(float %.0138, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0141 = select i1 %.not449, float %533, float %532
  br label %__internal_fmad.exit2.i.i.i

__internal_fmad.exit2.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i, %__internal_fmad.exit1.i.i.i
  %534 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i ]
  %535 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i ]
  %536 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i ], [ %.0141, %__internal_fmad.exit1.i.i.i ]
  %537 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %537, 0
  %538 = call float @llvm.nvvm.fma.rn.ftz.f(float %536, float %.0138, float noundef %535) #139
  %539 = call float @llvm.fma.f32(float %536, float %.0138, float noundef %535)
  %.0142 = select i1 %.not31, float %539, float %538
  %540 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not33 = icmp eq i32 %540, 0
  %541 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0142, float %.0138, float noundef %534) #139
  %542 = call float @llvm.fma.f32(float %.0142, float %.0138, float noundef %534)
  %.0143 = select i1 %.not33, float %542, float %541
  %543 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not34 = icmp eq i32 %543, 0
  %544 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0143, float %.0139, float %527) #139
  %545 = call float @llvm.fma.f32(float %.0143, float %.0139, float %527)
  %.0144 = select i1 %.not34, float %545, float %544
  %546 = and i32 %522, 2
  %.not35 = icmp eq i32 %546, 0
  br i1 %.not35, label %__internal_cos_offset_f.exit.i, label %__internal_fmad.exit5.i.i.i

__internal_fmad.exit5.i.i.i:                      ; preds = %__internal_fmad.exit2.i.i.i
  %547 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not448 = icmp eq i32 %547, 0
  %548 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0144, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %549 = call float @llvm.fma.f32(float %.0144, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0145 = select i1 %.not448, float %549, float %548
  br label %__internal_cos_offset_f.exit.i

__internal_cos_offset_f.exit.i:                   ; preds = %__internal_fmad.exit5.i.i.i, %__internal_fmad.exit2.i.i.i
  %z.i.i79.i.0 = phi float [ %.0145, %__internal_fmad.exit5.i.i.i ], [ %.0144, %__internal_fmad.exit2.i.i.i ]
  %550 = fmul float %424, %z.i.i79.i.0
  br label %__nv_y1f.exit

__nv_y1f.exit:                                    ; preds = %__internal_accurate_logf.exit.i, %__internal_fmad.exit29.i, %__internal_cos_offset_f.exit.i, %__nv_isinff.exit.i, %__internal_fmad.exit47.i, %__internal_fmad.exit11.i, %4
  %r.i.5 = phi float [ %5, %4 ], [ %.077, %__internal_accurate_logf.exit.i ], [ %329, %__internal_fmad.exit11.i ], [ %361, %__internal_fmad.exit29.i ], [ %390, %__internal_fmad.exit47.i ], [ %550, %__internal_cos_offset_f.exit.i ], [ 0.000000e+00, %__nv_isinff.exit.i ]
  %551 = fcmp olt float %a, 0.000000e+00
  %r.i.6 = select i1 %551, float 0x7FFFFFFFE0000000, float %r.i.5
  %552 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not36 = icmp eq i32 %552, 0
  %.0147 = select i1 %.not36, float %2, float %1
  %553 = fcmp ugt float %.0147, 0x3FDC982EC0000000
  br i1 %553, label %837, label %__internal_fmad.exit.i83

__internal_fmad.exit.i83:                         ; preds = %__nv_y1f.exit
  %554 = fmul float %.0147, %.0147
  %555 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not378 = icmp eq i32 %555, 0
  %556 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E7B7CB580000000, float %554, float noundef 0xBEE3E936C0000000) #139
  %557 = call float @llvm.fma.f32(float %554, float noundef 0x3E7B7CB580000000, float noundef 0xBEE3E936C0000000)
  %.0148 = select i1 %.not378, float %557, float %556
  %558 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not379 = icmp eq i32 %558, 0
  %559 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0148, float %554, float noundef 0x3F41A62000000000) #139
  %560 = call float @llvm.fma.f32(float %.0148, float %554, float noundef 0x3F41A62000000000)
  %.0150 = select i1 %.not379, float %560, float %559
  %561 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not380 = icmp eq i32 %561, 0
  %562 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0150, float %554, float noundef 0xBF9075B1C0000000) #139
  %563 = call float @llvm.fma.f32(float %.0150, float %554, float noundef 0xBF9075B1C0000000)
  %.0151 = select i1 %.not380, float %563, float %562
  %564 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not381 = icmp eq i32 %564, 0
  %565 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0151, float %554, float noundef 0x3FC6BBCB40000000) #139
  %566 = call float @llvm.fma.f32(float %.0151, float %554, float noundef 0x3FC6BBCB40000000)
  %.0152 = select i1 %.not381, float %566, float %565
  %567 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not382 = icmp eq i32 %567, 0
  %568 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0152, float %554, float noundef 0xBFB2E4D6A0000000) #139
  %569 = call float @llvm.fma.f32(float %.0152, float %554, float noundef 0xBFB2E4D6A0000000)
  %.0153 = select i1 %.not382, float %569, float %568
  %570 = fcmp olt float %.0147, 0x3810000000000000
  %571 = fmul float %.0147, 0x4160000000000000
  %.0155 = select i1 %570, float %571, float %.0147
  %i.i.i72.0 = select i1 %570, float -2.300000e+01, float 0.000000e+00
  %572 = bitcast float %.0155 to i32
  %573 = add i32 %572, -1059760811
  %574 = and i32 %573, -8388608
  %575 = sub i32 %572, %574
  %576 = bitcast i32 %575 to float
  %577 = sitofp i32 %574 to float
  %578 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not383 = icmp eq i32 %578, 0
  %579 = call float @llvm.nvvm.fma.rn.ftz.f(float %577, float noundef 0x3E80000000000000, float noundef %i.i.i72.0) #139
  %580 = call float @llvm.fma.f32(float %577, float noundef 0x3E80000000000000, float noundef %i.i.i72.0)
  %.0161 = select i1 %.not383, float %580, float %579
  %581 = fadd float %576, -1.000000e+00
  %582 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not384 = icmp eq i32 %582, 0
  %583 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %581, float noundef 0x3FC2073EC0000000) #139
  %584 = call float @llvm.fma.f32(float %581, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.0162 = select i1 %.not384, float %584, float %583
  %585 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not385 = icmp eq i32 %585, 0
  %586 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0162, float %581, float noundef 0xBFBF19B980000000) #139
  %587 = call float @llvm.fma.f32(float %.0162, float %581, float noundef 0xBFBF19B980000000)
  %.0163 = select i1 %.not385, float %587, float %586
  %588 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not386 = icmp eq i32 %588, 0
  %589 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0163, float %581, float noundef 0x3FC1E52AA0000000) #139
  %590 = call float @llvm.fma.f32(float %.0163, float %581, float noundef 0x3FC1E52AA0000000)
  %.0164 = select i1 %.not386, float %590, float %589
  %591 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not387 = icmp eq i32 %591, 0
  %592 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0164, float %581, float noundef 0xBFC55B1720000000) #139
  %593 = call float @llvm.fma.f32(float %.0164, float %581, float noundef 0xBFC55B1720000000)
  %.0165 = select i1 %.not387, float %593, float %592
  %594 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not388 = icmp eq i32 %594, 0
  %595 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0165, float %581, float noundef 0x3FC99DA160000000) #139
  %596 = call float @llvm.fma.f32(float %.0165, float %581, float noundef 0x3FC99DA160000000)
  %.0166 = select i1 %.not388, float %596, float %595
  %597 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not389 = icmp eq i32 %597, 0
  %598 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0166, float %581, float noundef 0xBFCFFFE440000000) #139
  %599 = call float @llvm.fma.f32(float %.0166, float %581, float noundef 0xBFCFFFE440000000)
  %.0168 = select i1 %.not389, float %599, float %598
  %600 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not390 = icmp eq i32 %600, 0
  %601 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0168, float %581, float noundef 0x3FD5554F00000000) #139
  %602 = call float @llvm.fma.f32(float %.0168, float %581, float noundef 0x3FD5554F00000000)
  %.0169 = select i1 %.not390, float %602, float %601
  %603 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not391 = icmp eq i32 %603, 0
  %604 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0169, float %581, float noundef -5.000000e-01) #139
  %605 = call float @llvm.fma.f32(float %.0169, float %581, float noundef -5.000000e-01)
  %.0159 = select i1 %.not391, float %605, float %604
  %606 = fmul float %581, %.0159
  %607 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not392 = icmp eq i32 %607, 0
  %608 = call float @llvm.nvvm.fma.rn.ftz.f(float %606, float %581, float %581) #139
  %609 = call float @llvm.fma.f32(float %606, float %581, float %581)
  %.0158 = select i1 %.not392, float %609, float %608
  %610 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not393 = icmp eq i32 %610, 0
  %611 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0161, float noundef 0x3FE62E4300000000, float %.0158) #139
  %612 = call float @llvm.fma.f32(float %.0161, float noundef 0x3FE62E4300000000, float %.0158)
  %.0157 = select i1 %.not393, float %612, float %611
  %613 = icmp ugt i32 %572, 2139095039
  br i1 %613, label %__nv_fmaf_rn.exit.i.i99, label %__internal_accurate_logf.exit.i100

__nv_fmaf_rn.exit.i.i99:                          ; preds = %__internal_fmad.exit.i83
  %614 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not447 = icmp eq i32 %614, 0
  %615 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0155, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %616 = call float @llvm.fma.f32(float %.0155, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.0156 = select i1 %.not447, float %616, float %615
  br label %__internal_accurate_logf.exit.i100

__internal_accurate_logf.exit.i100:               ; preds = %__nv_fmaf_rn.exit.i.i99, %__internal_fmad.exit.i83
  %r.i.i71.0 = phi float [ %.0156, %__nv_fmaf_rn.exit.i.i99 ], [ %.0157, %__internal_fmad.exit.i83 ]
  %617 = fcmp oeq float %.0155, 0.000000e+00
  %r.i.i71.0.op = fmul float %r.i.i71.0, 0x3FE45F3060000000
  %618 = select i1 %617, float 0xFFF0000000000000, float %r.i.i71.0.op
  %619 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not394 = icmp eq i32 %619, 0
  %620 = call float @llvm.nvvm.fabs.ftz.f(float %.0147) #139
  %621 = call float @llvm.fabs.f32(float %.0147)
  %.0170 = select i1 %.not394, float %621, float %620
  %622 = fcmp ugt float %.0170, 8.000000e+00
  br i1 %622, label %__nv_isinff.exit.i.i115, label %__internal_fmad.exit.i.i102

__internal_fmad.exit.i.i102:                      ; preds = %__internal_accurate_logf.exit.i100
  %623 = fadd float %.0170, 0xC0033D1520000000
  %624 = fadd float %623, 0xBE7D2E3660000000
  %625 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not434 = icmp eq i32 %625, 0
  %626 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBCD6771CE0000000, float %624, float noundef 0xBD35953660000000) #139
  %627 = call float @llvm.fma.f32(float %624, float noundef 0xBCD6771CE0000000, float noundef 0xBD35953660000000)
  %.0172 = select i1 %.not434, float %627, float %626
  %628 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not435 = icmp eq i32 %628, 0
  %629 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0172, float %624, float noundef 0x3D87E1C300000000) #139
  %630 = call float @llvm.fma.f32(float %.0172, float %624, float noundef 0x3D87E1C300000000)
  %.0173 = select i1 %.not435, float %630, float %629
  %631 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not436 = icmp eq i32 %631, 0
  %632 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0173, float %624, float noundef 0xBD9A82F020000000) #139
  %633 = call float @llvm.fma.f32(float %.0173, float %624, float noundef 0xBD9A82F020000000)
  %.0174 = select i1 %.not436, float %633, float %632
  %634 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not437 = icmp eq i32 %634, 0
  %635 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0174, float %624, float noundef 0xBDFD21E700000000) #139
  %636 = call float @llvm.fma.f32(float %.0174, float %624, float noundef 0xBDFD21E700000000)
  %.0175 = select i1 %.not437, float %636, float %635
  %637 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not438 = icmp eq i32 %637, 0
  %638 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0175, float %624, float noundef 0x3E04060B60000000) #139
  %639 = call float @llvm.fma.f32(float %.0175, float %624, float noundef 0x3E04060B60000000)
  %.0176 = select i1 %.not438, float %639, float %638
  %640 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not439 = icmp eq i32 %640, 0
  %641 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0176, float %624, float noundef 0x3E6F2E2860000000) #139
  %642 = call float @llvm.fma.f32(float %.0176, float %624, float noundef 0x3E6F2E2860000000)
  %.0177 = select i1 %.not439, float %642, float %641
  %643 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not440 = icmp eq i32 %643, 0
  %644 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0177, float %624, float noundef 0x3E1EEDF0A0000000) #139
  %645 = call float @llvm.fma.f32(float %.0177, float %624, float noundef 0x3E1EEDF0A0000000)
  %.0179 = select i1 %.not440, float %645, float %644
  %646 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not441 = icmp eq i32 %646, 0
  %647 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0179, float %624, float noundef 0xBED6DBF8C0000000) #139
  %648 = call float @llvm.fma.f32(float %.0179, float %624, float noundef 0xBED6DBF8C0000000)
  %.0180 = select i1 %.not441, float %648, float %647
  %649 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not442 = icmp eq i32 %649, 0
  %650 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0180, float %624, float noundef 0xBEDECCB920000000) #139
  %651 = call float @llvm.fma.f32(float %.0180, float %624, float noundef 0xBEDECCB920000000)
  %.0181 = select i1 %.not442, float %651, float %650
  %652 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not443 = icmp eq i32 %652, 0
  %653 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0181, float %624, float noundef 0x3F33C5BD60000000) #139
  %654 = call float @llvm.fma.f32(float %.0181, float %624, float noundef 0x3F33C5BD60000000)
  %.0182 = select i1 %.not443, float %654, float %653
  %655 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not444 = icmp eq i32 %655, 0
  %656 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0182, float %624, float noundef 0x3F495C6680000000) #139
  %657 = call float @llvm.fma.f32(float %.0182, float %624, float noundef 0x3F495C6680000000)
  %.0183 = select i1 %.not444, float %657, float %656
  %658 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not445 = icmp eq i32 %658, 0
  %659 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0183, float %624, float noundef 0xBF7DD54360000000) #139
  %660 = call float @llvm.fma.f32(float %.0183, float %624, float noundef 0xBF7DD54360000000)
  %.0184 = select i1 %.not445, float %660, float %659
  %661 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not446 = icmp eq i32 %661, 0
  %662 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0184, float %624, float noundef 0xBF9B4EE8E0000000) #139
  %663 = call float @llvm.fma.f32(float %.0184, float %624, float noundef 0xBF9B4EE8E0000000)
  %.0186 = select i1 %.not446, float %663, float %662
  %664 = fmul float %624, %.0186
  %665 = fadd float %.0170, 0xC016148F60000000
  %666 = fadd float %665, 0x3E734F46E0000000
  %667 = fmul float %666, %664
  %668 = fadd float %.0170, 0xC0214EB560000000
  %669 = fadd float %668, 0xBE9999BDA0000000
  %670 = fmul float %669, %667
  br label %__nv_j0f.exit.i

__nv_isinff.exit.i.i115:                          ; preds = %__internal_accurate_logf.exit.i100
  %671 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not395 = icmp eq i32 %671, 0
  %672 = call float @llvm.nvvm.fabs.ftz.f(float %.0170) #139
  %673 = call float @llvm.fabs.f32(float %.0170)
  %.0187 = select i1 %.not395, float %673, float %672
  %674 = fcmp oeq float %.0187, 0x7FF0000000000000
  br i1 %674, label %__nv_j0f.exit.i, label %__internal_fmad.exit13.i.i116

__internal_fmad.exit13.i.i116:                    ; preds = %__nv_isinff.exit.i.i115
  %675 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0170) #141, !srcloc !73
  %676 = fmul float %675, %675
  %677 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not396 = icmp eq i32 %677, 0
  %678 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400ADFD260000000, float %676, float noundef 0xBFE076F840000000) #139
  %679 = call float @llvm.fma.f32(float %676, float noundef 0x400ADFD260000000, float noundef 0xBFE076F840000000)
  %.0188 = select i1 %.not396, float %679, float %678
  %680 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not397 = icmp eq i32 %680, 0
  %681 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0188, float %676, float noundef 0x3FBA767E60000000) #139
  %682 = call float @llvm.fma.f32(float %.0188, float %676, float noundef 0x3FBA767E60000000)
  %.0189 = select i1 %.not397, float %682, float %681
  %683 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not398 = icmp eq i32 %683, 0
  %684 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0189, float %676, float noundef 0xBFAFFFF6C0000000) #139
  %685 = call float @llvm.fma.f32(float %.0189, float %676, float noundef 0xBFAFFFF6C0000000)
  %.0190 = select i1 %.not398, float %685, float %684
  %686 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not399 = icmp eq i32 %686, 0
  %687 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0190, float %676, float noundef 1.000000e+00) #139
  %688 = call float @llvm.fma.f32(float %.0190, float %676, float noundef 1.000000e+00)
  %.0191 = select i1 %.not399, float %688, float %687
  %689 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not400 = icmp eq i32 %689, 0
  %690 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF23C0120000000, float %676, float noundef 0xBFCA4825A0000000) #139
  %691 = call float @llvm.fma.f32(float %676, float noundef 0x3FF23C0120000000, float noundef 0xBFCA4825A0000000)
  %.0193 = select i1 %.not400, float %691, float %690
  %692 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not401 = icmp eq i32 %692, 0
  %693 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0193, float %676, float noundef 0x3FB0A9DA20000000) #139
  %694 = call float @llvm.fma.f32(float %.0193, float %676, float noundef 0x3FB0A9DA20000000)
  %.0194 = select i1 %.not401, float %694, float %693
  %695 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not402 = icmp eq i32 %695, 0
  %696 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0194, float %676, float noundef 0xBFBFFFFFE0000000) #139
  %697 = call float @llvm.fma.f32(float %.0194, float %676, float noundef 0xBFBFFFFFE0000000)
  %.0195 = select i1 %.not402, float %697, float %696
  %698 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not403 = icmp eq i32 %698, 0
  %699 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0195, float %675, float %.0170) #139
  %700 = call float @llvm.fma.f32(float %.0195, float %675, float %.0170)
  %.0196 = select i1 %.not403, float %700, float %699
  %701 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not404 = icmp eq i32 %701, 0
  br i1 %.not404, label %704, label %702

702:                                              ; preds = %__internal_fmad.exit13.i.i116
  %703 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0170) #139
  br label %__nv_rsqrtf.exit.i.i123

704:                                              ; preds = %__internal_fmad.exit13.i.i116
  %705 = call float @llvm.nvvm.rsqrt.approx.f(float %.0170) #139
  br label %__nv_rsqrtf.exit.i.i123

__nv_rsqrtf.exit.i.i123:                          ; preds = %704, %702
  %.0197 = phi float [ %703, %702 ], [ %705, %704 ]
  %706 = fmul float %.0197, 0x3FE9884540000000
  %707 = fmul float %.0191, %706
  %708 = fmul float %.0196, 0x3FE45F3060000000
  %709 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not405 = icmp eq i32 %709, 0
  %710 = call i32 @llvm.nvvm.f2i.rn.ftz(float %708) #139
  %711 = call i32 @llvm.nvvm.f2i.rn(float %708) #139
  %.0199 = select i1 %.not405, i32 %711, i32 %710
  %712 = sitofp i32 %.0199 to float
  %713 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not406 = icmp eq i32 %713, 0
  %714 = call float @llvm.nvvm.fma.rn.ftz.f(float %712, float noundef 0xBFF921FB40000000, float %.0196) #139
  %715 = call float @llvm.fma.f32(float %712, float noundef 0xBFF921FB40000000, float %.0196)
  %.0200 = select i1 %.not406, float %715, float %714
  %716 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not407 = icmp eq i32 %716, 0
  %717 = call float @llvm.nvvm.fma.rn.ftz.f(float %712, float noundef 0xBE74442D00000000, float %.0200) #139
  %718 = call float @llvm.fma.f32(float %712, float noundef 0xBE74442D00000000, float %.0200)
  %.0202 = select i1 %.not407, float %718, float %717
  %719 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not408 = icmp eq i32 %719, 0
  %720 = call float @llvm.nvvm.fma.rn.ftz.f(float %712, float noundef 0xBCF84698A0000000, float %.0202) #139
  %721 = call float @llvm.fma.f32(float %712, float noundef 0xBCF84698A0000000, float %.0202)
  %.0203 = select i1 %.not408, float %721, float %720
  %722 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not409 = icmp eq i32 %722, 0
  %723 = call float @llvm.nvvm.fabs.ftz.f(float %.0196) #139
  %724 = call float @llvm.fabs.f32(float %.0196)
  %.0204 = select i1 %.not409, float %724, float %723
  %725 = fcmp ult float %.0204, 1.056150e+05
  br i1 %725, label %__internal_trig_reduction_kernel.exit.i.i.i132, label %__nv_isinff.exit.i.i.i.i129

__nv_isinff.exit.i.i.i.i129:                      ; preds = %__nv_rsqrtf.exit.i.i123
  %726 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not427 = icmp eq i32 %726, 0
  %.0205 = select i1 %.not427, float %724, float %723
  %727 = fcmp oeq float %.0205, 0x7FF0000000000000
  br i1 %727, label %__nv_fmul_rn.exit.i.i.i.i130, label %731

__nv_fmul_rn.exit.i.i.i.i130:                     ; preds = %__nv_isinff.exit.i.i.i.i129
  %728 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not433 = icmp eq i32 %728, 0
  %729 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0196, float noundef 0.000000e+00) #139
  %730 = fmul float %.0196, 0.000000e+00
  %.0206 = select i1 %.not433, float %730, float %729
  br label %__internal_trig_reduction_kernel.exit.i.i.i132

731:                                              ; preds = %__nv_isinff.exit.i.i.i.i129
  %732 = bitcast float %.0196 to i32
  %733 = lshr i32 %732, 23
  %734 = and i32 %733, 255
  %735 = add nsw i32 %734, -128
  %736 = shl i32 %732, 8
  %737 = or i32 %736, -2147483648
  %738 = lshr i32 %735, 5
  br label %739

739:                                              ; preds = %731, %739
  %iq.i.i.i.i.i41.0567 = phi i32 [ 0, %731 ], [ %747, %739 ]
  %hi.i.i.i.i.i36.0566 = phi i32 [ 0, %731 ], [ %745, %739 ]
  %740 = zext i32 %iq.i.i.i.i.i41.0567 to i64
  %741 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %740
  %742 = load i32, i32 addrspace(1)* %741, align 4
  %743 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %742, i32 %737, i32 %hi.i.i.i.i.i36.0566) #141, !srcloc !67
  %744 = extractvalue { i32, i32 } %743, 0
  %745 = extractvalue { i32, i32 } %743, 1
  %746 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %740
  store i32 %744, i32* %746, align 4, !noalias !68
  %747 = add nuw nsw i32 %iq.i.i.i.i.i41.0567, 1
  %exitcond599.not = icmp eq i32 %747, 6
  br i1 %exitcond599.not, label %748, label %739, !llvm.loop !71

748:                                              ; preds = %739
  %749 = and i32 %732, -2147483648
  %750 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %745, i32* %750, align 4, !noalias !68
  %751 = and i32 %733, 31
  %752 = sub nsw i32 6, %738
  %753 = sext i32 %752 to i64
  %754 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %753
  %755 = load i32, i32* %754, align 4
  %756 = sub nsw i32 5, %738
  %757 = sext i32 %756 to i64
  %758 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %757
  %759 = load i32, i32* %758, align 4
  %.not428 = icmp eq i32 %751, 0
  br i1 %.not428, label %__internal_trig_reduction_slowpath.exit.i.i.i.i131, label %760

760:                                              ; preds = %748
  %761 = sub nsw i32 4, %738
  %762 = sub nuw nsw i32 32, %751
  %763 = shl i32 %755, %751
  %764 = lshr i32 %759, %762
  %765 = add i32 %764, %763
  %766 = shl i32 %759, %751
  %767 = sext i32 %761 to i64
  %768 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %767
  %769 = load i32, i32* %768, align 4
  %770 = lshr i32 %769, %762
  %771 = add i32 %770, %766
  br label %__internal_trig_reduction_slowpath.exit.i.i.i.i131

__internal_trig_reduction_slowpath.exit.i.i.i.i131: ; preds = %760, %748
  %hi.i.i.i.i.i36.1 = phi i32 [ %765, %760 ], [ %755, %748 ]
  %lo.i.i.i.i.i37.0 = phi i32 [ %771, %760 ], [ %759, %748 ]
  %772 = lshr i32 %hi.i.i.i.i.i36.1, 30
  %773 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.i36.1, i32 %lo.i.i.i.i.i37.0, i32 noundef 2)
  %774 = shl i32 %lo.i.i.i.i.i37.0, 2
  %775 = lshr i32 %773, 31
  %776 = add nuw nsw i32 %775, %772
  %.not429 = icmp eq i32 %749, 0
  %777 = sub nsw i32 0, %776
  %spec.select2 = select i1 %.not429, i32 %776, i32 %777
  %.not430 = icmp sgt i32 %773, -1
  %778 = xor i32 %749, -2147483648
  %s.i.i.i.i.i34.0 = select i1 %.not430, i32 %749, i32 %778
  %not..not430 = xor i1 %.not430, true
  %779 = sext i1 %not..not430 to i32
  %hi.i.i.i.i.i36.2 = xor i32 %773, %779
  %lo.i.i.i.i.i37.1 = xor i32 %774, %779
  %780 = zext i32 %hi.i.i.i.i.i36.2 to i64
  %781 = shl nuw i64 %780, 32
  %782 = zext i32 %lo.i.i.i.i.i37.1 to i64
  %783 = or i64 %781, %782
  %784 = sitofp i64 %783 to double
  %785 = fmul double %784, 0x3BF921FB54442D19
  %786 = fptrunc double %785 to float
  %.not432 = icmp eq i32 %s.i.i.i.i.i34.0, 0
  %787 = fneg float %786
  %r.i.i.i.i.i45.0 = select i1 %.not432, float %786, float %787
  br label %__internal_trig_reduction_kernel.exit.i.i.i132

__internal_trig_reduction_kernel.exit.i.i.i132:   ; preds = %__nv_fmul_rn.exit.i.i.i.i130, %__internal_trig_reduction_slowpath.exit.i.i.i.i131, %__nv_rsqrtf.exit.i.i123
  %i.i.i.i51.1 = phi i32 [ %.0199, %__nv_rsqrtf.exit.i.i123 ], [ 0, %__nv_fmul_rn.exit.i.i.i.i130 ], [ %spec.select2, %__internal_trig_reduction_slowpath.exit.i.i.i.i131 ]
  %t.i.i.i.i48.1 = phi float [ %.0203, %__nv_rsqrtf.exit.i.i123 ], [ %.0206, %__nv_fmul_rn.exit.i.i.i.i130 ], [ %r.i.i.i.i.i45.0, %__internal_trig_reduction_slowpath.exit.i.i.i.i131 ]
  %788 = and i32 %i.i.i.i51.1, 3
  %789 = uitofp i32 %788 to float
  %790 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not410 = icmp eq i32 %790, 0
  %791 = call float @llvm.nvvm.fma.rn.ftz.f(float %789, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000) #139
  %792 = call float @llvm.fma.f32(float %789, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000)
  %.0209 = select i1 %.not410, float %792, float %791
  %793 = fadd float %t.i.i.i.i48.1, %.0209
  %794 = fmul float %793, 0x3FE45F3060000000
  %795 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not411 = icmp eq i32 %795, 0
  %796 = call i32 @llvm.nvvm.f2i.rn.ftz(float %794) #139
  %797 = call i32 @llvm.nvvm.f2i.rn(float %794) #139
  %.0210 = select i1 %.not411, i32 %797, i32 %796
  %798 = sitofp i32 %.0210 to float
  %799 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not412 = icmp eq i32 %799, 0
  %800 = call float @llvm.nvvm.fma.rn.ftz.f(float %798, float noundef 0xBFF921FB40000000, float %793) #139
  %801 = call float @llvm.fma.f32(float %798, float noundef 0xBFF921FB40000000, float %793)
  %.0211 = select i1 %.not412, float %801, float %800
  %802 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not413 = icmp eq i32 %802, 0
  %803 = call float @llvm.nvvm.fma.rn.ftz.f(float %798, float noundef 0xBE74442D00000000, float %.0211) #139
  %804 = call float @llvm.fma.f32(float %798, float noundef 0xBE74442D00000000, float %.0211)
  %.0212 = select i1 %.not413, float %804, float %803
  %805 = add i32 %.0210, 1
  %806 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not414 = icmp eq i32 %806, 0
  %807 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0212, float %.0212) #139
  %808 = fmul float %.0212, %.0212
  %.0214 = select i1 %.not414, float %808, float %807
  %809 = and i32 %805, 1
  %.not415 = icmp eq i32 %809, 0
  %810 = select i1 %.not415, float %.0212, float 1.000000e+00
  %811 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not416 = icmp eq i32 %811, 0
  %812 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0214, float %810, float noundef 0.000000e+00) #139
  %813 = call float @llvm.fma.f32(float %.0214, float %810, float noundef 0.000000e+00)
  %.0215 = select i1 %.not416, float %813, float %812
  br i1 %.not415, label %__internal_fmad.exit2.i.i.i.i140, label %__internal_fmad.exit1.i.i.i.i139

__internal_fmad.exit1.i.i.i.i139:                 ; preds = %__internal_trig_reduction_kernel.exit.i.i.i132
  %814 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not426 = icmp eq i32 %814, 0
  %815 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0214, float noundef 0xBF56C0FDA0000000) #139
  %816 = call float @llvm.fma.f32(float %.0214, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0216 = select i1 %.not426, float %816, float %815
  br label %__internal_fmad.exit2.i.i.i.i140

__internal_fmad.exit2.i.i.i.i140:                 ; preds = %__internal_trig_reduction_kernel.exit.i.i.i132, %__internal_fmad.exit1.i.i.i.i139
  %817 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i.i132 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i.i139 ]
  %818 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i.i132 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i.i139 ]
  %819 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i.i132 ], [ %.0216, %__internal_fmad.exit1.i.i.i.i139 ]
  %820 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not419 = icmp eq i32 %820, 0
  %821 = call float @llvm.nvvm.fma.rn.ftz.f(float %819, float %.0214, float noundef %818) #139
  %822 = call float @llvm.fma.f32(float %819, float %.0214, float noundef %818)
  %.0217 = select i1 %.not419, float %822, float %821
  %823 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not421 = icmp eq i32 %823, 0
  %824 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0217, float %.0214, float noundef %817) #139
  %825 = call float @llvm.fma.f32(float %.0217, float %.0214, float noundef %817)
  %.0218 = select i1 %.not421, float %825, float %824
  %826 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not422 = icmp eq i32 %826, 0
  %827 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0218, float %.0215, float %810) #139
  %828 = call float @llvm.fma.f32(float %.0218, float %.0215, float %810)
  %.0219 = select i1 %.not422, float %828, float %827
  %829 = and i32 %805, 2
  %.not423 = icmp eq i32 %829, 0
  br i1 %.not423, label %__internal_cos_offset_f.exit.i.i144, label %__internal_fmad.exit5.i.i.i.i143

__internal_fmad.exit5.i.i.i.i143:                 ; preds = %__internal_fmad.exit2.i.i.i.i140
  %830 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not425 = icmp eq i32 %830, 0
  %831 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0219, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %832 = call float @llvm.fma.f32(float %.0219, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0221 = select i1 %.not425, float %832, float %831
  br label %__internal_cos_offset_f.exit.i.i144

__internal_cos_offset_f.exit.i.i144:              ; preds = %__internal_fmad.exit5.i.i.i.i143, %__internal_fmad.exit2.i.i.i.i140
  %z.i.i.i.i27.0 = phi float [ %.0221, %__internal_fmad.exit5.i.i.i.i143 ], [ %.0219, %__internal_fmad.exit2.i.i.i.i140 ]
  %833 = fmul float %707, %z.i.i.i.i27.0
  br label %__nv_j0f.exit.i

__nv_j0f.exit.i:                                  ; preds = %__internal_cos_offset_f.exit.i.i144, %__nv_isinff.exit.i.i115, %__internal_fmad.exit.i.i102
  %r.i5.i57.1 = phi float [ %670, %__internal_fmad.exit.i.i102 ], [ %833, %__internal_cos_offset_f.exit.i.i144 ], [ 0.000000e+00, %__nv_isinff.exit.i.i115 ]
  %834 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not424 = icmp eq i32 %834, 0
  %835 = call float @llvm.nvvm.fma.rn.ftz.f(float %618, float %r.i5.i57.1, float %.0153) #139
  %836 = call float @llvm.fma.f32(float %618, float %r.i5.i57.1, float %.0153)
  %.0222 = select i1 %.not424, float %836, float %835
  br label %__nv_y0f.exit

837:                                              ; preds = %__nv_y1f.exit
  %838 = fcmp ugt float %.0147, 0x3FFECF5F00000000
  br i1 %838, label %878, label %__internal_fmad.exit9.i146

__internal_fmad.exit9.i146:                       ; preds = %837
  %839 = fadd float %.0147, 0xBFEC982EC0000000
  %840 = fadd float %839, 0x3E4CAFA060000000
  %841 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not366 = icmp eq i32 %841, 0
  %842 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F9BB91660000000, float %840, float noundef 0xBFC04AE980000000) #139
  %843 = call float @llvm.fma.f32(float %840, float noundef 0x3F9BB91660000000, float noundef 0xBFC04AE980000000)
  %.0223 = select i1 %.not366, float %843, float %842
  %844 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not367 = icmp eq i32 %844, 0
  %845 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0223, float %840, float noundef 0x3FCFE59920000000) #139
  %846 = call float @llvm.fma.f32(float %.0223, float %840, float noundef 0x3FCFE59920000000)
  %.0224 = select i1 %.not367, float %846, float %845
  %847 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not368 = icmp eq i32 %847, 0
  %848 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0224, float %840, float noundef 0xBFD17E5360000000) #139
  %849 = call float @llvm.fma.f32(float %.0224, float %840, float noundef 0xBFD17E5360000000)
  %.0225 = select i1 %.not368, float %849, float %848
  %850 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not369 = icmp eq i32 %850, 0
  %851 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0225, float %840, float noundef 0x3FCB79D260000000) #139
  %852 = call float @llvm.fma.f32(float %.0225, float %840, float noundef 0x3FCB79D260000000)
  %.0226 = select i1 %.not369, float %852, float %851
  %853 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not370 = icmp eq i32 %853, 0
  %854 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0226, float %840, float noundef 0xBFC7189FE0000000) #139
  %855 = call float @llvm.fma.f32(float %.0226, float %840, float noundef 0xBFC7189FE0000000)
  %.0228 = select i1 %.not370, float %855, float %854
  %856 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not371 = icmp eq i32 %856, 0
  %857 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0228, float %840, float noundef 0x3FC84EE9A0000000) #139
  %858 = call float @llvm.fma.f32(float %.0228, float %840, float noundef 0x3FC84EE9A0000000)
  %.0229 = select i1 %.not371, float %858, float %857
  %859 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not372 = icmp eq i32 %859, 0
  %860 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0229, float %840, float noundef 0xBFCA4B9640000000) #139
  %861 = call float @llvm.fma.f32(float %.0229, float %840, float noundef 0xBFCA4B9640000000)
  %.0230 = select i1 %.not372, float %861, float %860
  %862 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not373 = icmp eq i32 %862, 0
  %863 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0230, float %840, float noundef 0x3FCC1E87A0000000) #139
  %864 = call float @llvm.fma.f32(float %.0230, float %840, float noundef 0x3FCC1E87A0000000)
  %.0231 = select i1 %.not373, float %864, float %863
  %865 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not374 = icmp eq i32 %865, 0
  %866 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0231, float %840, float noundef 0xBFCCF228A0000000) #139
  %867 = call float @llvm.fma.f32(float %.0231, float %840, float noundef 0xBFCCF228A0000000)
  %.0232 = select i1 %.not374, float %867, float %866
  %868 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not375 = icmp eq i32 %868, 0
  %869 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0232, float %840, float noundef 0x3FCC3A4940000000) #139
  %870 = call float @llvm.fma.f32(float %.0232, float %840, float noundef 0x3FCC3A4940000000)
  %.0233 = select i1 %.not375, float %870, float %869
  %871 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not376 = icmp eq i32 %871, 0
  %872 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0233, float %840, float noundef 0xBFDF7E35A0000000) #139
  %873 = call float @llvm.fma.f32(float %.0233, float %840, float noundef 0xBFDF7E35A0000000)
  %.0235 = select i1 %.not376, float %873, float %872
  %874 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not377 = icmp eq i32 %874, 0
  %875 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0235, float %840, float noundef 0x3FEC243760000000) #139
  %876 = call float @llvm.fma.f32(float %.0235, float %840, float noundef 0x3FEC243760000000)
  %.0236 = select i1 %.not377, float %876, float %875
  %877 = fmul float %840, %.0236
  br label %__nv_y0f.exit

878:                                              ; preds = %837
  %879 = fcmp ugt float %.0147, 0x40161663C0000000
  br i1 %879, label %910, label %__internal_fmad.exit33.i158

__internal_fmad.exit33.i158:                      ; preds = %878
  %880 = fadd float %.0147, 0xC00FA95340000000
  %881 = fadd float %880, 0xBE7B30AD40000000
  %882 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not357 = icmp eq i32 %882, 0
  %883 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBE893BA7E0000000, float %881, float noundef 0xBE8BC4C0E0000000) #139
  %884 = call float @llvm.fma.f32(float %881, float noundef 0xBE893BA7E0000000, float noundef 0xBE8BC4C0E0000000)
  %.0237 = select i1 %.not357, float %884, float %883
  %885 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not358 = icmp eq i32 %885, 0
  %886 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0237, float %881, float noundef 0xBED0AE0C80000000) #139
  %887 = call float @llvm.fma.f32(float %.0237, float %881, float noundef 0xBED0AE0C80000000)
  %.0238 = select i1 %.not358, float %887, float %886
  %888 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not359 = icmp eq i32 %888, 0
  %889 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0238, float %881, float noundef 0x3F0AA8C200000000) #139
  %890 = call float @llvm.fma.f32(float %.0238, float %881, float noundef 0x3F0AA8C200000000)
  %.0239 = select i1 %.not359, float %890, float %889
  %891 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not360 = icmp eq i32 %891, 0
  %892 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0239, float %881, float noundef 0x3F2959DAE0000000) #139
  %893 = call float @llvm.fma.f32(float %.0239, float %881, float noundef 0x3F2959DAE0000000)
  %.0240 = select i1 %.not360, float %893, float %892
  %894 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not361 = icmp eq i32 %894, 0
  %895 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0240, float %881, float noundef 0xBF61E34180000000) #139
  %896 = call float @llvm.fma.f32(float %.0240, float %881, float noundef 0xBF61E34180000000)
  %.0242 = select i1 %.not361, float %896, float %895
  %897 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not362 = icmp eq i32 %897, 0
  %898 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0242, float %881, float noundef 0xBF7C0FE5C0000000) #139
  %899 = call float @llvm.fma.f32(float %.0242, float %881, float noundef 0xBF7C0FE5C0000000)
  %.0243 = select i1 %.not362, float %899, float %898
  %900 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not363 = icmp eq i32 %900, 0
  %901 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0243, float %881, float noundef 0x3FADF6D6A0000000) #139
  %902 = call float @llvm.fma.f32(float %.0243, float %881, float noundef 0x3FADF6D6A0000000)
  %.0244 = select i1 %.not363, float %902, float %901
  %903 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not364 = icmp eq i32 %903, 0
  %904 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0244, float %881, float noundef 0x3FAA09BE20000000) #139
  %905 = call float @llvm.fma.f32(float %.0244, float %881, float noundef 0x3FAA09BE20000000)
  %.0245 = select i1 %.not364, float %905, float %904
  %906 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not365 = icmp eq i32 %906, 0
  %907 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0245, float %881, float noundef 0xBFD9C34260000000) #139
  %908 = call float @llvm.fma.f32(float %.0245, float %881, float noundef 0xBFD9C34260000000)
  %.0246 = select i1 %.not365, float %908, float %907
  %909 = fmul float %881, %.0246
  br label %__nv_y0f.exit

910:                                              ; preds = %878
  %911 = fcmp ugt float %.0147, 0x40214EF300000000
  br i1 %911, label %__nv_isinff.exit.i173, label %__internal_fmad.exit51.i167

__internal_fmad.exit51.i167:                      ; preds = %910
  %912 = fadd float %.0147, 0xC01C581DC0000000
  %913 = fadd float %912, 0xBE739C8400000000
  %914 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not349 = icmp eq i32 %914, 0
  %915 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EA219D7C0000000, float %913, float noundef 0x3EC53B4D80000000) #139
  %916 = call float @llvm.fma.f32(float %913, float noundef 0x3EA219D7C0000000, float noundef 0x3EC53B4D80000000)
  %.0247 = select i1 %.not349, float %916, float %915
  %917 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not350 = icmp eq i32 %917, 0
  %918 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0247, float %913, float noundef 0xBF080A9800000000) #139
  %919 = call float @llvm.fma.f32(float %.0247, float %913, float noundef 0xBF080A9800000000)
  %.0249 = select i1 %.not350, float %919, float %918
  %920 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not351 = icmp eq i32 %920, 0
  %921 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0249, float %913, float noundef 0xBF22631560000000) #139
  %922 = call float @llvm.fma.f32(float %.0249, float %913, float noundef 0xBF22631560000000)
  %.0250 = select i1 %.not351, float %922, float %921
  %923 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not352 = icmp eq i32 %923, 0
  %924 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0250, float %913, float noundef 0x3F61D32420000000) #139
  %925 = call float @llvm.fma.f32(float %.0250, float %913, float noundef 0x3F61D32420000000)
  %.0251 = select i1 %.not352, float %925, float %924
  %926 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not353 = icmp eq i32 %926, 0
  %927 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0251, float %913, float noundef 0x3F6B2E9AA0000000) #139
  %928 = call float @llvm.fma.f32(float %.0251, float %913, float noundef 0x3F6B2E9AA0000000)
  %.0252 = select i1 %.not353, float %928, float %927
  %929 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not354 = icmp eq i32 %929, 0
  %930 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0252, float %913, float noundef 0xBFA8969AE0000000) #139
  %931 = call float @llvm.fma.f32(float %.0252, float %913, float noundef 0xBFA8969AE0000000)
  %.0253 = select i1 %.not354, float %931, float %930
  %932 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not355 = icmp eq i32 %932, 0
  %933 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0253, float %913, float noundef 0xBF95AEF320000000) #139
  %934 = call float @llvm.fma.f32(float %.0253, float %913, float noundef 0xBF95AEF320000000)
  %.0254 = select i1 %.not355, float %934, float %933
  %935 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not356 = icmp eq i32 %935, 0
  %936 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0254, float %913, float noundef 0x3FD334CCA0000000) #139
  %937 = call float @llvm.fma.f32(float %.0254, float %913, float noundef 0x3FD334CCA0000000)
  %.0256 = select i1 %.not356, float %937, float %936
  %938 = fmul float %913, %.0256
  br label %__nv_y0f.exit

__nv_isinff.exit.i173:                            ; preds = %910
  %939 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not37 = icmp eq i32 %939, 0
  %940 = call float @llvm.nvvm.fabs.ftz.f(float %.0147) #139
  %941 = call float @llvm.fabs.f32(float %.0147)
  %.0257 = select i1 %.not37, float %941, float %940
  %942 = fcmp oeq float %.0257, 0x7FF0000000000000
  br i1 %942, label %__nv_y0f.exit, label %__internal_fmad.exit70.i174

__internal_fmad.exit70.i174:                      ; preds = %__nv_isinff.exit.i173
  %943 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0147) #141, !srcloc !73
  %944 = fmul float %943, %943
  %945 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not38 = icmp eq i32 %945, 0
  %946 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFD98D3E60000000, float %944, float noundef 0x3FBA1ABE00000000) #139
  %947 = call float @llvm.fma.f32(float %944, float noundef 0xBFD98D3E60000000, float noundef 0x3FBA1ABE00000000)
  %.0258 = select i1 %.not38, float %947, float %946
  %948 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not39 = icmp eq i32 %948, 0
  %949 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0258, float %944, float noundef 0xBFAFFF0AA0000000) #139
  %950 = call float @llvm.fma.f32(float %.0258, float %944, float noundef 0xBFAFFF0AA0000000)
  %.0259 = select i1 %.not39, float %950, float %949
  %951 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not40 = icmp eq i32 %951, 0
  %952 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0259, float %944, float noundef 1.000000e+00) #139
  %953 = call float @llvm.fma.f32(float %.0259, float %944, float noundef 1.000000e+00)
  %.0260 = select i1 %.not40, float %953, float %952
  %954 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not41 = icmp eq i32 %954, 0
  %955 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF199AC20000000, float %944, float noundef 0xBFCA1A6380000000) #139
  %956 = call float @llvm.fma.f32(float %944, float noundef 0x3FF199AC20000000, float noundef 0xBFCA1A6380000000)
  %.0261 = select i1 %.not41, float %956, float %955
  %957 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %957, 0
  %958 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0261, float %944, float noundef 0x3FB0A8F060000000) #139
  %959 = call float @llvm.fma.f32(float %.0261, float %944, float noundef 0x3FB0A8F060000000)
  %.0263 = select i1 %.not42, float %959, float %958
  %960 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %960, 0
  %961 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0263, float %944, float noundef 0xBFBFFFFF60000000) #139
  %962 = call float @llvm.fma.f32(float %.0263, float %944, float noundef 0xBFBFFFFF60000000)
  %.0264 = select i1 %.not43, float %962, float %961
  %963 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %963, 0
  %964 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0264, float %943, float %.0147) #139
  %965 = call float @llvm.fma.f32(float %.0264, float %943, float %.0147)
  %.0265 = select i1 %.not44, float %965, float %964
  %966 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %966, 0
  br i1 %.not45, label %969, label %967

967:                                              ; preds = %__internal_fmad.exit70.i174
  %968 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0147) #139
  br label %__nv_rsqrtf.exit.i179

969:                                              ; preds = %__internal_fmad.exit70.i174
  %970 = call float @llvm.nvvm.rsqrt.approx.f(float %.0147) #139
  br label %__nv_rsqrtf.exit.i179

__nv_rsqrtf.exit.i179:                            ; preds = %969, %967
  %.0266 = phi float [ %968, %967 ], [ %970, %969 ]
  %971 = fmul float %.0266, 0x3FE9884540000000
  %972 = fmul float %.0260, %971
  %973 = fmul float %.0265, 0x3FE45F3060000000
  %974 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %974, 0
  %975 = call i32 @llvm.nvvm.f2i.rn.ftz(float %973) #139
  %976 = call i32 @llvm.nvvm.f2i.rn(float %973) #139
  %.0268 = select i1 %.not46, i32 %976, i32 %975
  %977 = sitofp i32 %.0268 to float
  %978 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %978, 0
  %979 = call float @llvm.nvvm.fma.rn.ftz.f(float %977, float noundef 0xBFF921FB40000000, float %.0265) #139
  %980 = call float @llvm.fma.f32(float %977, float noundef 0xBFF921FB40000000, float %.0265)
  %.0269 = select i1 %.not47, float %980, float %979
  %981 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %981, 0
  %982 = call float @llvm.nvvm.fma.rn.ftz.f(float %977, float noundef 0xBE74442D00000000, float %.0269) #139
  %983 = call float @llvm.fma.f32(float %977, float noundef 0xBE74442D00000000, float %.0269)
  %.0270 = select i1 %.not48, float %983, float %982
  %984 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %984, 0
  %985 = call float @llvm.nvvm.fma.rn.ftz.f(float %977, float noundef 0xBCF84698A0000000, float %.0270) #139
  %986 = call float @llvm.fma.f32(float %977, float noundef 0xBCF84698A0000000, float %.0270)
  %.0272 = select i1 %.not49, float %986, float %985
  %987 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %987, 0
  %988 = call float @llvm.nvvm.fabs.ftz.f(float %.0265) #139
  %989 = call float @llvm.fabs.f32(float %.0265)
  %.0273 = select i1 %.not50, float %989, float %988
  %990 = fcmp ult float %.0273, 1.056150e+05
  br i1 %990, label %__internal_trig_reduction_kernel.exit.i.i185, label %__nv_isinff.exit.i.i.i182

__nv_isinff.exit.i.i.i182:                        ; preds = %__nv_rsqrtf.exit.i179
  %991 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not342 = icmp eq i32 %991, 0
  %.0274 = select i1 %.not342, float %989, float %988
  %992 = fcmp oeq float %.0274, 0x7FF0000000000000
  br i1 %992, label %__nv_fmul_rn.exit.i.i.i183, label %996

__nv_fmul_rn.exit.i.i.i183:                       ; preds = %__nv_isinff.exit.i.i.i182
  %993 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not348 = icmp eq i32 %993, 0
  %994 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0265, float noundef 0.000000e+00) #139
  %995 = fmul float %.0265, 0.000000e+00
  %.0275 = select i1 %.not348, float %995, float %994
  br label %__internal_trig_reduction_kernel.exit.i.i185

996:                                              ; preds = %__nv_isinff.exit.i.i.i182
  %997 = bitcast float %.0265 to i32
  %998 = lshr i32 %997, 23
  %999 = and i32 %998, 255
  %1000 = add nsw i32 %999, -128
  %1001 = shl i32 %997, 8
  %1002 = or i32 %1001, -2147483648
  %1003 = lshr i32 %1000, 5
  br label %1004

1004:                                             ; preds = %996, %1004
  %iq.i.i.i.i15.0565 = phi i32 [ 0, %996 ], [ %1012, %1004 ]
  %hi.i.i.i.i10.0564 = phi i32 [ 0, %996 ], [ %1010, %1004 ]
  %1005 = zext i32 %iq.i.i.i.i15.0565 to i64
  %1006 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %1005
  %1007 = load i32, i32 addrspace(1)* %1006, align 4
  %1008 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %1007, i32 %1002, i32 %hi.i.i.i.i10.0564) #141, !srcloc !67
  %1009 = extractvalue { i32, i32 } %1008, 0
  %1010 = extractvalue { i32, i32 } %1008, 1
  %1011 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1005
  store i32 %1009, i32* %1011, align 4, !noalias !68
  %1012 = add nuw nsw i32 %iq.i.i.i.i15.0565, 1
  %exitcond598.not = icmp eq i32 %1012, 6
  br i1 %exitcond598.not, label %1013, label %1004, !llvm.loop !71

1013:                                             ; preds = %1004
  %1014 = and i32 %997, -2147483648
  %1015 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %1010, i32* %1015, align 4, !noalias !68
  %1016 = and i32 %998, 31
  %1017 = sub nsw i32 6, %1003
  %1018 = sext i32 %1017 to i64
  %1019 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1018
  %1020 = load i32, i32* %1019, align 4
  %1021 = sub nsw i32 5, %1003
  %1022 = sext i32 %1021 to i64
  %1023 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1022
  %1024 = load i32, i32* %1023, align 4
  %.not343 = icmp eq i32 %1016, 0
  br i1 %.not343, label %__internal_trig_reduction_slowpath.exit.i.i.i184, label %1025

1025:                                             ; preds = %1013
  %1026 = sub nsw i32 4, %1003
  %1027 = sub nuw nsw i32 32, %1016
  %1028 = shl i32 %1020, %1016
  %1029 = lshr i32 %1024, %1027
  %1030 = add i32 %1029, %1028
  %1031 = shl i32 %1024, %1016
  %1032 = sext i32 %1026 to i64
  %1033 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1032
  %1034 = load i32, i32* %1033, align 4
  %1035 = lshr i32 %1034, %1027
  %1036 = add i32 %1035, %1031
  br label %__internal_trig_reduction_slowpath.exit.i.i.i184

__internal_trig_reduction_slowpath.exit.i.i.i184: ; preds = %1025, %1013
  %hi.i.i.i.i10.1 = phi i32 [ %1030, %1025 ], [ %1020, %1013 ]
  %lo.i.i.i.i11.0 = phi i32 [ %1036, %1025 ], [ %1024, %1013 ]
  %1037 = lshr i32 %hi.i.i.i.i10.1, 30
  %1038 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i10.1, i32 %lo.i.i.i.i11.0, i32 noundef 2)
  %1039 = shl i32 %lo.i.i.i.i11.0, 2
  %1040 = lshr i32 %1038, 31
  %1041 = add nuw nsw i32 %1040, %1037
  %.not344 = icmp eq i32 %1014, 0
  %1042 = sub nsw i32 0, %1041
  %spec.select3 = select i1 %.not344, i32 %1041, i32 %1042
  %.not345 = icmp sgt i32 %1038, -1
  %1043 = xor i32 %1014, -2147483648
  %s.i.i.i.i9.0 = select i1 %.not345, i32 %1014, i32 %1043
  %not..not345 = xor i1 %.not345, true
  %1044 = sext i1 %not..not345 to i32
  %hi.i.i.i.i10.2 = xor i32 %1038, %1044
  %lo.i.i.i.i11.1 = xor i32 %1039, %1044
  %1045 = zext i32 %hi.i.i.i.i10.2 to i64
  %1046 = shl nuw i64 %1045, 32
  %1047 = zext i32 %lo.i.i.i.i11.1 to i64
  %1048 = or i64 %1046, %1047
  %1049 = sitofp i64 %1048 to double
  %1050 = fmul double %1049, 0x3BF921FB54442D19
  %1051 = fptrunc double %1050 to float
  %.not347 = icmp eq i32 %s.i.i.i.i9.0, 0
  %1052 = fneg float %1051
  %r.i.i.i.i19.0 = select i1 %.not347, float %1051, float %1052
  br label %__internal_trig_reduction_kernel.exit.i.i185

__internal_trig_reduction_kernel.exit.i.i185:     ; preds = %__nv_fmul_rn.exit.i.i.i183, %__internal_trig_reduction_slowpath.exit.i.i.i184, %__nv_rsqrtf.exit.i179
  %i.i85.i.1 = phi i32 [ %.0268, %__nv_rsqrtf.exit.i179 ], [ 0, %__nv_fmul_rn.exit.i.i.i183 ], [ %spec.select3, %__internal_trig_reduction_slowpath.exit.i.i.i184 ]
  %t.i.i.i22.1 = phi float [ %.0272, %__nv_rsqrtf.exit.i179 ], [ %.0275, %__nv_fmul_rn.exit.i.i.i183 ], [ %r.i.i.i.i19.0, %__internal_trig_reduction_slowpath.exit.i.i.i184 ]
  %1053 = and i32 %i.i85.i.1, 3
  %1054 = uitofp i32 %1053 to float
  %1055 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not51 = icmp eq i32 %1055, 0
  %1056 = call float @llvm.nvvm.fma.rn.ftz.f(float %1054, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %1057 = call float @llvm.fma.f32(float %1054, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.0278 = select i1 %.not51, float %1057, float %1056
  %1058 = fadd float %t.i.i.i22.1, %.0278
  %1059 = fmul float %1058, 0x3FE45F3060000000
  %1060 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not52 = icmp eq i32 %1060, 0
  %1061 = call i32 @llvm.nvvm.f2i.rn.ftz(float %1059) #139
  %1062 = call i32 @llvm.nvvm.f2i.rn(float %1059) #139
  %.0279 = select i1 %.not52, i32 %1062, i32 %1061
  %1063 = sitofp i32 %.0279 to float
  %1064 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not53 = icmp eq i32 %1064, 0
  %1065 = call float @llvm.nvvm.fma.rn.ftz.f(float %1063, float noundef 0xBFF921FB40000000, float %1058) #139
  %1066 = call float @llvm.fma.f32(float %1063, float noundef 0xBFF921FB40000000, float %1058)
  %.0280 = select i1 %.not53, float %1066, float %1065
  %1067 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not54 = icmp eq i32 %1067, 0
  %1068 = call float @llvm.nvvm.fma.rn.ftz.f(float %1063, float noundef 0xBE74442D00000000, float %.0280) #139
  %1069 = call float @llvm.fma.f32(float %1063, float noundef 0xBE74442D00000000, float %.0280)
  %.0281 = select i1 %.not54, float %1069, float %1068
  %1070 = add i32 %.0279, 1
  %1071 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not55 = icmp eq i32 %1071, 0
  %1072 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0281, float %.0281) #139
  %1073 = fmul float %.0281, %.0281
  %.0282 = select i1 %.not55, float %1073, float %1072
  %1074 = and i32 %1070, 1
  %.not56 = icmp eq i32 %1074, 0
  %1075 = select i1 %.not56, float %.0281, float 1.000000e+00
  %1076 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not57 = icmp eq i32 %1076, 0
  %1077 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0282, float %1075, float noundef 0.000000e+00) #139
  %1078 = call float @llvm.fma.f32(float %.0282, float %1075, float noundef 0.000000e+00)
  %.0284 = select i1 %.not57, float %1078, float %1077
  br i1 %.not56, label %__internal_fmad.exit2.i.i.i190, label %__internal_fmad.exit1.i.i.i189

__internal_fmad.exit1.i.i.i189:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i185
  %1079 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not341 = icmp eq i32 %1079, 0
  %1080 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0282, float noundef 0xBF56C0FDA0000000) #139
  %1081 = call float @llvm.fma.f32(float %.0282, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0285 = select i1 %.not341, float %1081, float %1080
  br label %__internal_fmad.exit2.i.i.i190

__internal_fmad.exit2.i.i.i190:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i185, %__internal_fmad.exit1.i.i.i189
  %1082 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i185 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i189 ]
  %1083 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i185 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i189 ]
  %1084 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i185 ], [ %.0285, %__internal_fmad.exit1.i.i.i189 ]
  %1085 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not60 = icmp eq i32 %1085, 0
  %1086 = call float @llvm.nvvm.fma.rn.ftz.f(float %1084, float %.0282, float noundef %1083) #139
  %1087 = call float @llvm.fma.f32(float %1084, float %.0282, float noundef %1083)
  %.0286 = select i1 %.not60, float %1087, float %1086
  %1088 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not62 = icmp eq i32 %1088, 0
  %1089 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0286, float %.0282, float noundef %1082) #139
  %1090 = call float @llvm.fma.f32(float %.0286, float %.0282, float noundef %1082)
  %.0287 = select i1 %.not62, float %1090, float %1089
  %1091 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not63 = icmp eq i32 %1091, 0
  %1092 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0287, float %.0284, float %1075) #139
  %1093 = call float @llvm.fma.f32(float %.0287, float %.0284, float %1075)
  %.0288 = select i1 %.not63, float %1093, float %1092
  %1094 = and i32 %1070, 2
  %.not64 = icmp eq i32 %1094, 0
  br i1 %.not64, label %__internal_cos_offset_f.exit.i194, label %__internal_fmad.exit5.i.i.i193

__internal_fmad.exit5.i.i.i193:                   ; preds = %__internal_fmad.exit2.i.i.i190
  %1095 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not340 = icmp eq i32 %1095, 0
  %1096 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0288, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %1097 = call float @llvm.fma.f32(float %.0288, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0289 = select i1 %.not340, float %1097, float %1096
  br label %__internal_cos_offset_f.exit.i194

__internal_cos_offset_f.exit.i194:                ; preds = %__internal_fmad.exit5.i.i.i193, %__internal_fmad.exit2.i.i.i190
  %z.i.i83.i.0 = phi float [ %.0289, %__internal_fmad.exit5.i.i.i193 ], [ %.0288, %__internal_fmad.exit2.i.i.i190 ]
  %1098 = fmul float %972, %z.i.i83.i.0
  br label %__nv_y0f.exit

__nv_y0f.exit:                                    ; preds = %__internal_fmad.exit9.i146, %__internal_fmad.exit51.i167, %__nv_isinff.exit.i173, %__internal_cos_offset_f.exit.i194, %__internal_fmad.exit33.i158, %__nv_j0f.exit.i
  %r.i75.4 = phi float [ %.0222, %__nv_j0f.exit.i ], [ %877, %__internal_fmad.exit9.i146 ], [ %909, %__internal_fmad.exit33.i158 ], [ %938, %__internal_fmad.exit51.i167 ], [ %1098, %__internal_cos_offset_f.exit.i194 ], [ 0.000000e+00, %__nv_isinff.exit.i173 ]
  %r.i75.5 = select i1 %551, float 0x7FFFFFFFE0000000, float %r.i75.4
  switch i32 %n, label %2195 [
    i32 0, label %__nv_fabsf.exit.i281
    i32 1, label %__nv_fabsf.exit.i496
  ]

__nv_fabsf.exit.i281:                             ; preds = %__nv_y0f.exit
  %1099 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not199 = icmp eq i32 %1099, 0
  %.0291 = select i1 %.not199, float %2, float %1
  %1100 = fcmp ugt float %.0291, 0x3FDC982EC0000000
  br i1 %1100, label %1384, label %__internal_fmad.exit.i282

__internal_fmad.exit.i282:                        ; preds = %__nv_fabsf.exit.i281
  %1101 = fmul float %.0291, %.0291
  %1102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not266 = icmp eq i32 %1102, 0
  %1103 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E7B7CB580000000, float %1101, float noundef 0xBEE3E936C0000000) #139
  %1104 = call float @llvm.fma.f32(float %1101, float noundef 0x3E7B7CB580000000, float noundef 0xBEE3E936C0000000)
  %.0293 = select i1 %.not266, float %1104, float %1103
  %1105 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not267 = icmp eq i32 %1105, 0
  %1106 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0293, float %1101, float noundef 0x3F41A62000000000) #139
  %1107 = call float @llvm.fma.f32(float %.0293, float %1101, float noundef 0x3F41A62000000000)
  %.0294 = select i1 %.not267, float %1107, float %1106
  %1108 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not268 = icmp eq i32 %1108, 0
  %1109 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0294, float %1101, float noundef 0xBF9075B1C0000000) #139
  %1110 = call float @llvm.fma.f32(float %.0294, float %1101, float noundef 0xBF9075B1C0000000)
  %.0295 = select i1 %.not268, float %1110, float %1109
  %1111 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not269 = icmp eq i32 %1111, 0
  %1112 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0295, float %1101, float noundef 0x3FC6BBCB40000000) #139
  %1113 = call float @llvm.fma.f32(float %.0295, float %1101, float noundef 0x3FC6BBCB40000000)
  %.0296 = select i1 %.not269, float %1113, float %1112
  %1114 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not270 = icmp eq i32 %1114, 0
  %1115 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0296, float %1101, float noundef 0xBFB2E4D6A0000000) #139
  %1116 = call float @llvm.fma.f32(float %.0296, float %1101, float noundef 0xBFB2E4D6A0000000)
  %.0297 = select i1 %.not270, float %1116, float %1115
  %1117 = fcmp olt float %.0291, 0x3810000000000000
  %1118 = fmul float %.0291, 0x4160000000000000
  %.0299 = select i1 %1117, float %1118, float %.0291
  %i.i.i271.0 = select i1 %1117, float -2.300000e+01, float 0.000000e+00
  %1119 = bitcast float %.0299 to i32
  %1120 = add i32 %1119, -1059760811
  %1121 = and i32 %1120, -8388608
  %1122 = sub i32 %1119, %1121
  %1123 = bitcast i32 %1122 to float
  %1124 = sitofp i32 %1121 to float
  %1125 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not271 = icmp eq i32 %1125, 0
  %1126 = call float @llvm.nvvm.fma.rn.ftz.f(float %1124, float noundef 0x3E80000000000000, float noundef %i.i.i271.0) #139
  %1127 = call float @llvm.fma.f32(float %1124, float noundef 0x3E80000000000000, float noundef %i.i.i271.0)
  %.0305 = select i1 %.not271, float %1127, float %1126
  %1128 = fadd float %1123, -1.000000e+00
  %1129 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not272 = icmp eq i32 %1129, 0
  %1130 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %1128, float noundef 0x3FC2073EC0000000) #139
  %1131 = call float @llvm.fma.f32(float %1128, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.0306 = select i1 %.not272, float %1131, float %1130
  %1132 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not273 = icmp eq i32 %1132, 0
  %1133 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0306, float %1128, float noundef 0xBFBF19B980000000) #139
  %1134 = call float @llvm.fma.f32(float %.0306, float %1128, float noundef 0xBFBF19B980000000)
  %.0308 = select i1 %.not273, float %1134, float %1133
  %1135 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not274 = icmp eq i32 %1135, 0
  %1136 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0308, float %1128, float noundef 0x3FC1E52AA0000000) #139
  %1137 = call float @llvm.fma.f32(float %.0308, float %1128, float noundef 0x3FC1E52AA0000000)
  %.0309 = select i1 %.not274, float %1137, float %1136
  %1138 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not275 = icmp eq i32 %1138, 0
  %1139 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0309, float %1128, float noundef 0xBFC55B1720000000) #139
  %1140 = call float @llvm.fma.f32(float %.0309, float %1128, float noundef 0xBFC55B1720000000)
  %.0310 = select i1 %.not275, float %1140, float %1139
  %1141 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not276 = icmp eq i32 %1141, 0
  %1142 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0310, float %1128, float noundef 0x3FC99DA160000000) #139
  %1143 = call float @llvm.fma.f32(float %.0310, float %1128, float noundef 0x3FC99DA160000000)
  %.0312 = select i1 %.not276, float %1143, float %1142
  %1144 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not277 = icmp eq i32 %1144, 0
  %1145 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0312, float %1128, float noundef 0xBFCFFFE440000000) #139
  %1146 = call float @llvm.fma.f32(float %.0312, float %1128, float noundef 0xBFCFFFE440000000)
  %.0313 = select i1 %.not277, float %1146, float %1145
  %1147 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not278 = icmp eq i32 %1147, 0
  %1148 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0313, float %1128, float noundef 0x3FD5554F00000000) #139
  %1149 = call float @llvm.fma.f32(float %.0313, float %1128, float noundef 0x3FD5554F00000000)
  %.0314 = select i1 %.not278, float %1149, float %1148
  %1150 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not279 = icmp eq i32 %1150, 0
  %1151 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0314, float %1128, float noundef -5.000000e-01) #139
  %1152 = call float @llvm.fma.f32(float %.0314, float %1128, float noundef -5.000000e-01)
  %.0304 = select i1 %.not279, float %1152, float %1151
  %1153 = fmul float %1128, %.0304
  %1154 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not280 = icmp eq i32 %1154, 0
  %1155 = call float @llvm.nvvm.fma.rn.ftz.f(float %1153, float %1128, float %1128) #139
  %1156 = call float @llvm.fma.f32(float %1153, float %1128, float %1128)
  %.0302 = select i1 %.not280, float %1156, float %1155
  %1157 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not281 = icmp eq i32 %1157, 0
  %1158 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0305, float noundef 0x3FE62E4300000000, float %.0302) #139
  %1159 = call float @llvm.fma.f32(float %.0305, float noundef 0x3FE62E4300000000, float %.0302)
  %.0301 = select i1 %.not281, float %1159, float %1158
  %1160 = icmp ugt i32 %1119, 2139095039
  br i1 %1160, label %__nv_fmaf_rn.exit.i.i298, label %__internal_accurate_logf.exit.i299

__nv_fmaf_rn.exit.i.i298:                         ; preds = %__internal_fmad.exit.i282
  %1161 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not335 = icmp eq i32 %1161, 0
  %1162 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0299, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %1163 = call float @llvm.fma.f32(float %.0299, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.0300 = select i1 %.not335, float %1163, float %1162
  br label %__internal_accurate_logf.exit.i299

__internal_accurate_logf.exit.i299:               ; preds = %__nv_fmaf_rn.exit.i.i298, %__internal_fmad.exit.i282
  %r.i.i270.0 = phi float [ %.0300, %__nv_fmaf_rn.exit.i.i298 ], [ %.0301, %__internal_fmad.exit.i282 ]
  %1164 = fcmp oeq float %.0299, 0.000000e+00
  %r.i.i270.0.op = fmul float %r.i.i270.0, 0x3FE45F3060000000
  %1165 = select i1 %1164, float 0xFFF0000000000000, float %r.i.i270.0.op
  %1166 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not282 = icmp eq i32 %1166, 0
  %1167 = call float @llvm.nvvm.fabs.ftz.f(float %.0291) #139
  %1168 = call float @llvm.fabs.f32(float %.0291)
  %.0317 = select i1 %.not282, float %1168, float %1167
  %1169 = fcmp ugt float %.0317, 8.000000e+00
  br i1 %1169, label %__nv_isinff.exit.i.i314, label %__internal_fmad.exit.i.i301

__internal_fmad.exit.i.i301:                      ; preds = %__internal_accurate_logf.exit.i299
  %1170 = fadd float %.0317, 0xC0033D1520000000
  %1171 = fadd float %1170, 0xBE7D2E3660000000
  %1172 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not322 = icmp eq i32 %1172, 0
  %1173 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBCD6771CE0000000, float %1171, float noundef 0xBD35953660000000) #139
  %1174 = call float @llvm.fma.f32(float %1171, float noundef 0xBCD6771CE0000000, float noundef 0xBD35953660000000)
  %.0318 = select i1 %.not322, float %1174, float %1173
  %1175 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not323 = icmp eq i32 %1175, 0
  %1176 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0318, float %1171, float noundef 0x3D87E1C300000000) #139
  %1177 = call float @llvm.fma.f32(float %.0318, float %1171, float noundef 0x3D87E1C300000000)
  %.0319 = select i1 %.not323, float %1177, float %1176
  %1178 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not324 = icmp eq i32 %1178, 0
  %1179 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0319, float %1171, float noundef 0xBD9A82F020000000) #139
  %1180 = call float @llvm.fma.f32(float %.0319, float %1171, float noundef 0xBD9A82F020000000)
  %.0321 = select i1 %.not324, float %1180, float %1179
  %1181 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not325 = icmp eq i32 %1181, 0
  %1182 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0321, float %1171, float noundef 0xBDFD21E700000000) #139
  %1183 = call float @llvm.fma.f32(float %.0321, float %1171, float noundef 0xBDFD21E700000000)
  %.0322 = select i1 %.not325, float %1183, float %1182
  %1184 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not326 = icmp eq i32 %1184, 0
  %1185 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0322, float %1171, float noundef 0x3E04060B60000000) #139
  %1186 = call float @llvm.fma.f32(float %.0322, float %1171, float noundef 0x3E04060B60000000)
  %.0323 = select i1 %.not326, float %1186, float %1185
  %1187 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not327 = icmp eq i32 %1187, 0
  %1188 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0323, float %1171, float noundef 0x3E6F2E2860000000) #139
  %1189 = call float @llvm.fma.f32(float %.0323, float %1171, float noundef 0x3E6F2E2860000000)
  %.0325 = select i1 %.not327, float %1189, float %1188
  %1190 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not328 = icmp eq i32 %1190, 0
  %1191 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0325, float %1171, float noundef 0x3E1EEDF0A0000000) #139
  %1192 = call float @llvm.fma.f32(float %.0325, float %1171, float noundef 0x3E1EEDF0A0000000)
  %.0326 = select i1 %.not328, float %1192, float %1191
  %1193 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not329 = icmp eq i32 %1193, 0
  %1194 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0326, float %1171, float noundef 0xBED6DBF8C0000000) #139
  %1195 = call float @llvm.fma.f32(float %.0326, float %1171, float noundef 0xBED6DBF8C0000000)
  %.0327 = select i1 %.not329, float %1195, float %1194
  %1196 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not330 = icmp eq i32 %1196, 0
  %1197 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0327, float %1171, float noundef 0xBEDECCB920000000) #139
  %1198 = call float @llvm.fma.f32(float %.0327, float %1171, float noundef 0xBEDECCB920000000)
  %.0329 = select i1 %.not330, float %1198, float %1197
  %1199 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not331 = icmp eq i32 %1199, 0
  %1200 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0329, float %1171, float noundef 0x3F33C5BD60000000) #139
  %1201 = call float @llvm.fma.f32(float %.0329, float %1171, float noundef 0x3F33C5BD60000000)
  %.0330 = select i1 %.not331, float %1201, float %1200
  %1202 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not332 = icmp eq i32 %1202, 0
  %1203 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0330, float %1171, float noundef 0x3F495C6680000000) #139
  %1204 = call float @llvm.fma.f32(float %.0330, float %1171, float noundef 0x3F495C6680000000)
  %.0331 = select i1 %.not332, float %1204, float %1203
  %1205 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not333 = icmp eq i32 %1205, 0
  %1206 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0331, float %1171, float noundef 0xBF7DD54360000000) #139
  %1207 = call float @llvm.fma.f32(float %.0331, float %1171, float noundef 0xBF7DD54360000000)
  %.0333 = select i1 %.not333, float %1207, float %1206
  %1208 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not334 = icmp eq i32 %1208, 0
  %1209 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0333, float %1171, float noundef 0xBF9B4EE8E0000000) #139
  %1210 = call float @llvm.fma.f32(float %.0333, float %1171, float noundef 0xBF9B4EE8E0000000)
  %.0334 = select i1 %.not334, float %1210, float %1209
  %1211 = fmul float %1171, %.0334
  %1212 = fadd float %.0317, 0xC016148F60000000
  %1213 = fadd float %1212, 0x3E734F46E0000000
  %1214 = fmul float %1213, %1211
  %1215 = fadd float %.0317, 0xC0214EB560000000
  %1216 = fadd float %1215, 0xBE9999BDA0000000
  %1217 = fmul float %1216, %1214
  br label %__nv_j0f.exit.i345

__nv_isinff.exit.i.i314:                          ; preds = %__internal_accurate_logf.exit.i299
  %1218 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not283 = icmp eq i32 %1218, 0
  %1219 = call float @llvm.nvvm.fabs.ftz.f(float %.0317) #139
  %1220 = call float @llvm.fabs.f32(float %.0317)
  %.0335 = select i1 %.not283, float %1220, float %1219
  %1221 = fcmp oeq float %.0335, 0x7FF0000000000000
  br i1 %1221, label %__nv_j0f.exit.i345, label %__internal_fmad.exit13.i.i315

__internal_fmad.exit13.i.i315:                    ; preds = %__nv_isinff.exit.i.i314
  %1222 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0317) #141, !srcloc !73
  %1223 = fmul float %1222, %1222
  %1224 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not284 = icmp eq i32 %1224, 0
  %1225 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400ADFD260000000, float %1223, float noundef 0xBFE076F840000000) #139
  %1226 = call float @llvm.fma.f32(float %1223, float noundef 0x400ADFD260000000, float noundef 0xBFE076F840000000)
  %.0336 = select i1 %.not284, float %1226, float %1225
  %1227 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not285 = icmp eq i32 %1227, 0
  %1228 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0336, float %1223, float noundef 0x3FBA767E60000000) #139
  %1229 = call float @llvm.fma.f32(float %.0336, float %1223, float noundef 0x3FBA767E60000000)
  %.0338 = select i1 %.not285, float %1229, float %1228
  %1230 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not286 = icmp eq i32 %1230, 0
  %1231 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0338, float %1223, float noundef 0xBFAFFFF6C0000000) #139
  %1232 = call float @llvm.fma.f32(float %.0338, float %1223, float noundef 0xBFAFFFF6C0000000)
  %.0339 = select i1 %.not286, float %1232, float %1231
  %1233 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not287 = icmp eq i32 %1233, 0
  %1234 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0339, float %1223, float noundef 1.000000e+00) #139
  %1235 = call float @llvm.fma.f32(float %.0339, float %1223, float noundef 1.000000e+00)
  %.0340 = select i1 %.not287, float %1235, float %1234
  %1236 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not288 = icmp eq i32 %1236, 0
  %1237 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF23C0120000000, float %1223, float noundef 0xBFCA4825A0000000) #139
  %1238 = call float @llvm.fma.f32(float %1223, float noundef 0x3FF23C0120000000, float noundef 0xBFCA4825A0000000)
  %.0341 = select i1 %.not288, float %1238, float %1237
  %1239 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not289 = icmp eq i32 %1239, 0
  %1240 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0341, float %1223, float noundef 0x3FB0A9DA20000000) #139
  %1241 = call float @llvm.fma.f32(float %.0341, float %1223, float noundef 0x3FB0A9DA20000000)
  %.0342 = select i1 %.not289, float %1241, float %1240
  %1242 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not290 = icmp eq i32 %1242, 0
  %1243 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0342, float %1223, float noundef 0xBFBFFFFFE0000000) #139
  %1244 = call float @llvm.fma.f32(float %.0342, float %1223, float noundef 0xBFBFFFFFE0000000)
  %.0344 = select i1 %.not290, float %1244, float %1243
  %1245 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not291 = icmp eq i32 %1245, 0
  %1246 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0344, float %1222, float %.0317) #139
  %1247 = call float @llvm.fma.f32(float %.0344, float %1222, float %.0317)
  %.0345 = select i1 %.not291, float %1247, float %1246
  %1248 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not292 = icmp eq i32 %1248, 0
  br i1 %.not292, label %1251, label %1249

1249:                                             ; preds = %__internal_fmad.exit13.i.i315
  %1250 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0317) #139
  br label %__nv_rsqrtf.exit.i.i323

1251:                                             ; preds = %__internal_fmad.exit13.i.i315
  %1252 = call float @llvm.nvvm.rsqrt.approx.f(float %.0317) #139
  br label %__nv_rsqrtf.exit.i.i323

__nv_rsqrtf.exit.i.i323:                          ; preds = %1251, %1249
  %.0346 = phi float [ %1250, %1249 ], [ %1252, %1251 ]
  %1253 = fmul float %.0346, 0x3FE9884540000000
  %1254 = fmul float %.0340, %1253
  %1255 = fmul float %.0345, 0x3FE45F3060000000
  %1256 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not293 = icmp eq i32 %1256, 0
  %1257 = call i32 @llvm.nvvm.f2i.rn.ftz(float %1255) #139
  %1258 = call i32 @llvm.nvvm.f2i.rn(float %1255) #139
  %.0347 = select i1 %.not293, i32 %1258, i32 %1257
  %1259 = sitofp i32 %.0347 to float
  %1260 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not294 = icmp eq i32 %1260, 0
  %1261 = call float @llvm.nvvm.fma.rn.ftz.f(float %1259, float noundef 0xBFF921FB40000000, float %.0345) #139
  %1262 = call float @llvm.fma.f32(float %1259, float noundef 0xBFF921FB40000000, float %.0345)
  %.0348 = select i1 %.not294, float %1262, float %1261
  %1263 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not295 = icmp eq i32 %1263, 0
  %1264 = call float @llvm.nvvm.fma.rn.ftz.f(float %1259, float noundef 0xBE74442D00000000, float %.0348) #139
  %1265 = call float @llvm.fma.f32(float %1259, float noundef 0xBE74442D00000000, float %.0348)
  %.0349 = select i1 %.not295, float %1265, float %1264
  %1266 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not296 = icmp eq i32 %1266, 0
  %1267 = call float @llvm.nvvm.fma.rn.ftz.f(float %1259, float noundef 0xBCF84698A0000000, float %.0349) #139
  %1268 = call float @llvm.fma.f32(float %1259, float noundef 0xBCF84698A0000000, float %.0349)
  %.0351 = select i1 %.not296, float %1268, float %1267
  %1269 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not297 = icmp eq i32 %1269, 0
  %1270 = call float @llvm.nvvm.fabs.ftz.f(float %.0345) #139
  %1271 = call float @llvm.fabs.f32(float %.0345)
  %.0352 = select i1 %.not297, float %1271, float %1270
  %1272 = fcmp ult float %.0352, 1.056150e+05
  br i1 %1272, label %__internal_trig_reduction_kernel.exit.i.i.i332, label %__nv_isinff.exit.i.i.i.i329

__nv_isinff.exit.i.i.i.i329:                      ; preds = %__nv_rsqrtf.exit.i.i323
  %1273 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not315 = icmp eq i32 %1273, 0
  %.0353 = select i1 %.not315, float %1271, float %1270
  %1274 = fcmp oeq float %.0353, 0x7FF0000000000000
  br i1 %1274, label %__nv_fmul_rn.exit.i.i.i.i330, label %1278

__nv_fmul_rn.exit.i.i.i.i330:                     ; preds = %__nv_isinff.exit.i.i.i.i329
  %1275 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not321 = icmp eq i32 %1275, 0
  %1276 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0345, float noundef 0.000000e+00) #139
  %1277 = fmul float %.0345, 0.000000e+00
  %.0354 = select i1 %.not321, float %1277, float %1276
  br label %__internal_trig_reduction_kernel.exit.i.i.i332

1278:                                             ; preds = %__nv_isinff.exit.i.i.i.i329
  %1279 = bitcast float %.0345 to i32
  %1280 = lshr i32 %1279, 23
  %1281 = and i32 %1280, 255
  %1282 = add nsw i32 %1281, -128
  %1283 = shl i32 %1279, 8
  %1284 = or i32 %1283, -2147483648
  %1285 = lshr i32 %1282, 5
  br label %1286

1286:                                             ; preds = %1278, %1286
  %iq.i.i.i.i.i240.0559 = phi i32 [ 0, %1278 ], [ %1294, %1286 ]
  %hi.i.i.i.i.i235.0558 = phi i32 [ 0, %1278 ], [ %1292, %1286 ]
  %1287 = zext i32 %iq.i.i.i.i.i240.0559 to i64
  %1288 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %1287
  %1289 = load i32, i32 addrspace(1)* %1288, align 4
  %1290 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %1289, i32 %1284, i32 %hi.i.i.i.i.i235.0558) #141, !srcloc !67
  %1291 = extractvalue { i32, i32 } %1290, 0
  %1292 = extractvalue { i32, i32 } %1290, 1
  %1293 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1287
  store i32 %1291, i32* %1293, align 4, !noalias !68
  %1294 = add nuw nsw i32 %iq.i.i.i.i.i240.0559, 1
  %exitcond595.not = icmp eq i32 %1294, 6
  br i1 %exitcond595.not, label %1295, label %1286, !llvm.loop !71

1295:                                             ; preds = %1286
  %1296 = and i32 %1279, -2147483648
  %1297 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %1292, i32* %1297, align 4, !noalias !68
  %1298 = and i32 %1280, 31
  %1299 = sub nsw i32 6, %1285
  %1300 = sext i32 %1299 to i64
  %1301 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1300
  %1302 = load i32, i32* %1301, align 4
  %1303 = sub nsw i32 5, %1285
  %1304 = sext i32 %1303 to i64
  %1305 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1304
  %1306 = load i32, i32* %1305, align 4
  %.not316 = icmp eq i32 %1298, 0
  br i1 %.not316, label %__internal_trig_reduction_slowpath.exit.i.i.i.i331, label %1307

1307:                                             ; preds = %1295
  %1308 = sub nsw i32 4, %1285
  %1309 = sub nuw nsw i32 32, %1298
  %1310 = shl i32 %1302, %1298
  %1311 = lshr i32 %1306, %1309
  %1312 = add i32 %1311, %1310
  %1313 = shl i32 %1306, %1298
  %1314 = sext i32 %1308 to i64
  %1315 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1314
  %1316 = load i32, i32* %1315, align 4
  %1317 = lshr i32 %1316, %1309
  %1318 = add i32 %1317, %1313
  br label %__internal_trig_reduction_slowpath.exit.i.i.i.i331

__internal_trig_reduction_slowpath.exit.i.i.i.i331: ; preds = %1307, %1295
  %hi.i.i.i.i.i235.1 = phi i32 [ %1312, %1307 ], [ %1302, %1295 ]
  %lo.i.i.i.i.i236.0 = phi i32 [ %1318, %1307 ], [ %1306, %1295 ]
  %1319 = lshr i32 %hi.i.i.i.i.i235.1, 30
  %1320 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.i235.1, i32 %lo.i.i.i.i.i236.0, i32 noundef 2)
  %1321 = shl i32 %lo.i.i.i.i.i236.0, 2
  %1322 = lshr i32 %1320, 31
  %1323 = add nuw nsw i32 %1322, %1319
  %.not317 = icmp eq i32 %1296, 0
  %1324 = sub nsw i32 0, %1323
  %spec.select4 = select i1 %.not317, i32 %1323, i32 %1324
  %.not318 = icmp sgt i32 %1320, -1
  %1325 = xor i32 %1296, -2147483648
  %s.i.i.i.i.i233.0 = select i1 %.not318, i32 %1296, i32 %1325
  %not..not318 = xor i1 %.not318, true
  %1326 = sext i1 %not..not318 to i32
  %hi.i.i.i.i.i235.2 = xor i32 %1320, %1326
  %lo.i.i.i.i.i236.1 = xor i32 %1321, %1326
  %1327 = zext i32 %hi.i.i.i.i.i235.2 to i64
  %1328 = shl nuw i64 %1327, 32
  %1329 = zext i32 %lo.i.i.i.i.i236.1 to i64
  %1330 = or i64 %1328, %1329
  %1331 = sitofp i64 %1330 to double
  %1332 = fmul double %1331, 0x3BF921FB54442D19
  %1333 = fptrunc double %1332 to float
  %.not320 = icmp eq i32 %s.i.i.i.i.i233.0, 0
  %1334 = fneg float %1333
  %r.i.i.i.i.i244.0 = select i1 %.not320, float %1333, float %1334
  br label %__internal_trig_reduction_kernel.exit.i.i.i332

__internal_trig_reduction_kernel.exit.i.i.i332:   ; preds = %__nv_fmul_rn.exit.i.i.i.i330, %__internal_trig_reduction_slowpath.exit.i.i.i.i331, %__nv_rsqrtf.exit.i.i323
  %i.i.i.i250.1 = phi i32 [ %.0347, %__nv_rsqrtf.exit.i.i323 ], [ 0, %__nv_fmul_rn.exit.i.i.i.i330 ], [ %spec.select4, %__internal_trig_reduction_slowpath.exit.i.i.i.i331 ]
  %t.i.i.i.i247.1 = phi float [ %.0351, %__nv_rsqrtf.exit.i.i323 ], [ %.0354, %__nv_fmul_rn.exit.i.i.i.i330 ], [ %r.i.i.i.i.i244.0, %__internal_trig_reduction_slowpath.exit.i.i.i.i331 ]
  %1335 = and i32 %i.i.i.i250.1, 3
  %1336 = uitofp i32 %1335 to float
  %1337 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not298 = icmp eq i32 %1337, 0
  %1338 = call float @llvm.nvvm.fma.rn.ftz.f(float %1336, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000) #139
  %1339 = call float @llvm.fma.f32(float %1336, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000)
  %.0356 = select i1 %.not298, float %1339, float %1338
  %1340 = fadd float %t.i.i.i.i247.1, %.0356
  %1341 = fmul float %1340, 0x3FE45F3060000000
  %1342 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not299 = icmp eq i32 %1342, 0
  %1343 = call i32 @llvm.nvvm.f2i.rn.ftz(float %1341) #139
  %1344 = call i32 @llvm.nvvm.f2i.rn(float %1341) #139
  %.0357 = select i1 %.not299, i32 %1344, i32 %1343
  %1345 = sitofp i32 %.0357 to float
  %1346 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not300 = icmp eq i32 %1346, 0
  %1347 = call float @llvm.nvvm.fma.rn.ftz.f(float %1345, float noundef 0xBFF921FB40000000, float %1340) #139
  %1348 = call float @llvm.fma.f32(float %1345, float noundef 0xBFF921FB40000000, float %1340)
  %.0358 = select i1 %.not300, float %1348, float %1347
  %1349 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not301 = icmp eq i32 %1349, 0
  %1350 = call float @llvm.nvvm.fma.rn.ftz.f(float %1345, float noundef 0xBE74442D00000000, float %.0358) #139
  %1351 = call float @llvm.fma.f32(float %1345, float noundef 0xBE74442D00000000, float %.0358)
  %.0359 = select i1 %.not301, float %1351, float %1350
  %1352 = add i32 %.0357, 1
  %1353 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not302 = icmp eq i32 %1353, 0
  %1354 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0359, float %.0359) #139
  %1355 = fmul float %.0359, %.0359
  %.0361 = select i1 %.not302, float %1355, float %1354
  %1356 = and i32 %1352, 1
  %.not303 = icmp eq i32 %1356, 0
  %1357 = select i1 %.not303, float %.0359, float 1.000000e+00
  %1358 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not304 = icmp eq i32 %1358, 0
  %1359 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0361, float %1357, float noundef 0.000000e+00) #139
  %1360 = call float @llvm.fma.f32(float %.0361, float %1357, float noundef 0.000000e+00)
  %.0362 = select i1 %.not304, float %1360, float %1359
  br i1 %.not303, label %__internal_fmad.exit2.i.i.i.i340, label %__internal_fmad.exit1.i.i.i.i339

__internal_fmad.exit1.i.i.i.i339:                 ; preds = %__internal_trig_reduction_kernel.exit.i.i.i332
  %1361 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not314 = icmp eq i32 %1361, 0
  %1362 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0361, float noundef 0xBF56C0FDA0000000) #139
  %1363 = call float @llvm.fma.f32(float %.0361, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0363 = select i1 %.not314, float %1363, float %1362
  br label %__internal_fmad.exit2.i.i.i.i340

__internal_fmad.exit2.i.i.i.i340:                 ; preds = %__internal_trig_reduction_kernel.exit.i.i.i332, %__internal_fmad.exit1.i.i.i.i339
  %1364 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i.i332 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i.i339 ]
  %1365 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i.i332 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i.i339 ]
  %1366 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i.i332 ], [ %.0363, %__internal_fmad.exit1.i.i.i.i339 ]
  %1367 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not307 = icmp eq i32 %1367, 0
  %1368 = call float @llvm.nvvm.fma.rn.ftz.f(float %1366, float %.0361, float noundef %1365) #139
  %1369 = call float @llvm.fma.f32(float %1366, float %.0361, float noundef %1365)
  %.0364 = select i1 %.not307, float %1369, float %1368
  %1370 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not309 = icmp eq i32 %1370, 0
  %1371 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0364, float %.0361, float noundef %1364) #139
  %1372 = call float @llvm.fma.f32(float %.0364, float %.0361, float noundef %1364)
  %.0365 = select i1 %.not309, float %1372, float %1371
  %1373 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not310 = icmp eq i32 %1373, 0
  %1374 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0365, float %.0362, float %1357) #139
  %1375 = call float @llvm.fma.f32(float %.0365, float %.0362, float %1357)
  %.0367 = select i1 %.not310, float %1375, float %1374
  %1376 = and i32 %1352, 2
  %.not311 = icmp eq i32 %1376, 0
  br i1 %.not311, label %__internal_cos_offset_f.exit.i.i344, label %__internal_fmad.exit5.i.i.i.i343

__internal_fmad.exit5.i.i.i.i343:                 ; preds = %__internal_fmad.exit2.i.i.i.i340
  %1377 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not313 = icmp eq i32 %1377, 0
  %1378 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0367, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %1379 = call float @llvm.fma.f32(float %.0367, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0368 = select i1 %.not313, float %1379, float %1378
  br label %__internal_cos_offset_f.exit.i.i344

__internal_cos_offset_f.exit.i.i344:              ; preds = %__internal_fmad.exit5.i.i.i.i343, %__internal_fmad.exit2.i.i.i.i340
  %z.i.i.i.i226.0 = phi float [ %.0368, %__internal_fmad.exit5.i.i.i.i343 ], [ %.0367, %__internal_fmad.exit2.i.i.i.i340 ]
  %1380 = fmul float %1254, %z.i.i.i.i226.0
  br label %__nv_j0f.exit.i345

__nv_j0f.exit.i345:                               ; preds = %__internal_cos_offset_f.exit.i.i344, %__nv_isinff.exit.i.i314, %__internal_fmad.exit.i.i301
  %r.i5.i256.1 = phi float [ %1217, %__internal_fmad.exit.i.i301 ], [ %1380, %__internal_cos_offset_f.exit.i.i344 ], [ 0.000000e+00, %__nv_isinff.exit.i.i314 ]
  %1381 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not312 = icmp eq i32 %1381, 0
  %1382 = call float @llvm.nvvm.fma.rn.ftz.f(float %1165, float %r.i5.i256.1, float %.0297) #139
  %1383 = call float @llvm.fma.f32(float %1165, float %r.i5.i256.1, float %.0297)
  %.0369 = select i1 %.not312, float %1383, float %1382
  br label %__nv_y0f.exit406

1384:                                             ; preds = %__nv_fabsf.exit.i281
  %1385 = fcmp ugt float %.0291, 0x3FFECF5F00000000
  br i1 %1385, label %1425, label %__internal_fmad.exit9.i347

__internal_fmad.exit9.i347:                       ; preds = %1384
  %1386 = fadd float %.0291, 0xBFEC982EC0000000
  %1387 = fadd float %1386, 0x3E4CAFA060000000
  %1388 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not254 = icmp eq i32 %1388, 0
  %1389 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F9BB91660000000, float %1387, float noundef 0xBFC04AE980000000) #139
  %1390 = call float @llvm.fma.f32(float %1387, float noundef 0x3F9BB91660000000, float noundef 0xBFC04AE980000000)
  %.0370 = select i1 %.not254, float %1390, float %1389
  %1391 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not255 = icmp eq i32 %1391, 0
  %1392 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0370, float %1387, float noundef 0x3FCFE59920000000) #139
  %1393 = call float @llvm.fma.f32(float %.0370, float %1387, float noundef 0x3FCFE59920000000)
  %.0371 = select i1 %.not255, float %1393, float %1392
  %1394 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not256 = icmp eq i32 %1394, 0
  %1395 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0371, float %1387, float noundef 0xBFD17E5360000000) #139
  %1396 = call float @llvm.fma.f32(float %.0371, float %1387, float noundef 0xBFD17E5360000000)
  %.0372 = select i1 %.not256, float %1396, float %1395
  %1397 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not257 = icmp eq i32 %1397, 0
  %1398 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0372, float %1387, float noundef 0x3FCB79D260000000) #139
  %1399 = call float @llvm.fma.f32(float %.0372, float %1387, float noundef 0x3FCB79D260000000)
  %.0374 = select i1 %.not257, float %1399, float %1398
  %1400 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not258 = icmp eq i32 %1400, 0
  %1401 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0374, float %1387, float noundef 0xBFC7189FE0000000) #139
  %1402 = call float @llvm.fma.f32(float %.0374, float %1387, float noundef 0xBFC7189FE0000000)
  %.0375 = select i1 %.not258, float %1402, float %1401
  %1403 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not259 = icmp eq i32 %1403, 0
  %1404 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0375, float %1387, float noundef 0x3FC84EE9A0000000) #139
  %1405 = call float @llvm.fma.f32(float %.0375, float %1387, float noundef 0x3FC84EE9A0000000)
  %.0376 = select i1 %.not259, float %1405, float %1404
  %1406 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not260 = icmp eq i32 %1406, 0
  %1407 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0376, float %1387, float noundef 0xBFCA4B9640000000) #139
  %1408 = call float @llvm.fma.f32(float %.0376, float %1387, float noundef 0xBFCA4B9640000000)
  %.0377 = select i1 %.not260, float %1408, float %1407
  %1409 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not261 = icmp eq i32 %1409, 0
  %1410 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0377, float %1387, float noundef 0x3FCC1E87A0000000) #139
  %1411 = call float @llvm.fma.f32(float %.0377, float %1387, float noundef 0x3FCC1E87A0000000)
  %.0378 = select i1 %.not261, float %1411, float %1410
  %1412 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not262 = icmp eq i32 %1412, 0
  %1413 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0378, float %1387, float noundef 0xBFCCF228A0000000) #139
  %1414 = call float @llvm.fma.f32(float %.0378, float %1387, float noundef 0xBFCCF228A0000000)
  %.0379 = select i1 %.not262, float %1414, float %1413
  %1415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not263 = icmp eq i32 %1415, 0
  %1416 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0379, float %1387, float noundef 0x3FCC3A4940000000) #139
  %1417 = call float @llvm.fma.f32(float %.0379, float %1387, float noundef 0x3FCC3A4940000000)
  %.0381 = select i1 %.not263, float %1417, float %1416
  %1418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not264 = icmp eq i32 %1418, 0
  %1419 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0381, float %1387, float noundef 0xBFDF7E35A0000000) #139
  %1420 = call float @llvm.fma.f32(float %.0381, float %1387, float noundef 0xBFDF7E35A0000000)
  %.0382 = select i1 %.not264, float %1420, float %1419
  %1421 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not265 = icmp eq i32 %1421, 0
  %1422 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0382, float %1387, float noundef 0x3FEC243760000000) #139
  %1423 = call float @llvm.fma.f32(float %.0382, float %1387, float noundef 0x3FEC243760000000)
  %.0383 = select i1 %.not265, float %1423, float %1422
  %1424 = fmul float %1387, %.0383
  br label %__nv_y0f.exit406

1425:                                             ; preds = %1384
  %1426 = fcmp ugt float %.0291, 0x40161663C0000000
  br i1 %1426, label %1457, label %__internal_fmad.exit33.i359

__internal_fmad.exit33.i359:                      ; preds = %1425
  %1427 = fadd float %.0291, 0xC00FA95340000000
  %1428 = fadd float %1427, 0xBE7B30AD40000000
  %1429 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not245 = icmp eq i32 %1429, 0
  %1430 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBE893BA7E0000000, float %1428, float noundef 0xBE8BC4C0E0000000) #139
  %1431 = call float @llvm.fma.f32(float %1428, float noundef 0xBE893BA7E0000000, float noundef 0xBE8BC4C0E0000000)
  %.0384 = select i1 %.not245, float %1431, float %1430
  %1432 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not246 = icmp eq i32 %1432, 0
  %1433 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0384, float %1428, float noundef 0xBED0AE0C80000000) #139
  %1434 = call float @llvm.fma.f32(float %.0384, float %1428, float noundef 0xBED0AE0C80000000)
  %.0385 = select i1 %.not246, float %1434, float %1433
  %1435 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not247 = icmp eq i32 %1435, 0
  %1436 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0385, float %1428, float noundef 0x3F0AA8C200000000) #139
  %1437 = call float @llvm.fma.f32(float %.0385, float %1428, float noundef 0x3F0AA8C200000000)
  %.0386 = select i1 %.not247, float %1437, float %1436
  %1438 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not248 = icmp eq i32 %1438, 0
  %1439 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0386, float %1428, float noundef 0x3F2959DAE0000000) #139
  %1440 = call float @llvm.fma.f32(float %.0386, float %1428, float noundef 0x3F2959DAE0000000)
  %.0388 = select i1 %.not248, float %1440, float %1439
  %1441 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not249 = icmp eq i32 %1441, 0
  %1442 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0388, float %1428, float noundef 0xBF61E34180000000) #139
  %1443 = call float @llvm.fma.f32(float %.0388, float %1428, float noundef 0xBF61E34180000000)
  %.0389 = select i1 %.not249, float %1443, float %1442
  %1444 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not250 = icmp eq i32 %1444, 0
  %1445 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0389, float %1428, float noundef 0xBF7C0FE5C0000000) #139
  %1446 = call float @llvm.fma.f32(float %.0389, float %1428, float noundef 0xBF7C0FE5C0000000)
  %.0390 = select i1 %.not250, float %1446, float %1445
  %1447 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not251 = icmp eq i32 %1447, 0
  %1448 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0390, float %1428, float noundef 0x3FADF6D6A0000000) #139
  %1449 = call float @llvm.fma.f32(float %.0390, float %1428, float noundef 0x3FADF6D6A0000000)
  %.0391 = select i1 %.not251, float %1449, float %1448
  %1450 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not252 = icmp eq i32 %1450, 0
  %1451 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0391, float %1428, float noundef 0x3FAA09BE20000000) #139
  %1452 = call float @llvm.fma.f32(float %.0391, float %1428, float noundef 0x3FAA09BE20000000)
  %.0392 = select i1 %.not252, float %1452, float %1451
  %1453 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not253 = icmp eq i32 %1453, 0
  %1454 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0392, float %1428, float noundef 0xBFD9C34260000000) #139
  %1455 = call float @llvm.fma.f32(float %.0392, float %1428, float noundef 0xBFD9C34260000000)
  %.0393 = select i1 %.not253, float %1455, float %1454
  %1456 = fmul float %1428, %.0393
  br label %__nv_y0f.exit406

1457:                                             ; preds = %1425
  %1458 = fcmp ugt float %.0291, 0x40214EF300000000
  br i1 %1458, label %__nv_isinff.exit.i376, label %__internal_fmad.exit51.i368

__internal_fmad.exit51.i368:                      ; preds = %1457
  %1459 = fadd float %.0291, 0xC01C581DC0000000
  %1460 = fadd float %1459, 0xBE739C8400000000
  %1461 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not237 = icmp eq i32 %1461, 0
  %1462 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EA219D7C0000000, float %1460, float noundef 0x3EC53B4D80000000) #139
  %1463 = call float @llvm.fma.f32(float %1460, float noundef 0x3EA219D7C0000000, float noundef 0x3EC53B4D80000000)
  %.0395 = select i1 %.not237, float %1463, float %1462
  %1464 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not238 = icmp eq i32 %1464, 0
  %1465 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0395, float %1460, float noundef 0xBF080A9800000000) #139
  %1466 = call float @llvm.fma.f32(float %.0395, float %1460, float noundef 0xBF080A9800000000)
  %.0396 = select i1 %.not238, float %1466, float %1465
  %1467 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not239 = icmp eq i32 %1467, 0
  %1468 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0396, float %1460, float noundef 0xBF22631560000000) #139
  %1469 = call float @llvm.fma.f32(float %.0396, float %1460, float noundef 0xBF22631560000000)
  %.0397 = select i1 %.not239, float %1469, float %1468
  %1470 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not240 = icmp eq i32 %1470, 0
  %1471 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0397, float %1460, float noundef 0x3F61D32420000000) #139
  %1472 = call float @llvm.fma.f32(float %.0397, float %1460, float noundef 0x3F61D32420000000)
  %.0398 = select i1 %.not240, float %1472, float %1471
  %1473 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not241 = icmp eq i32 %1473, 0
  %1474 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0398, float %1460, float noundef 0x3F6B2E9AA0000000) #139
  %1475 = call float @llvm.fma.f32(float %.0398, float %1460, float noundef 0x3F6B2E9AA0000000)
  %.0399 = select i1 %.not241, float %1475, float %1474
  %1476 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not242 = icmp eq i32 %1476, 0
  %1477 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0399, float %1460, float noundef 0xBFA8969AE0000000) #139
  %1478 = call float @llvm.fma.f32(float %.0399, float %1460, float noundef 0xBFA8969AE0000000)
  %.0400 = select i1 %.not242, float %1478, float %1477
  %1479 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not243 = icmp eq i32 %1479, 0
  %1480 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0400, float %1460, float noundef 0xBF95AEF320000000) #139
  %1481 = call float @llvm.fma.f32(float %.0400, float %1460, float noundef 0xBF95AEF320000000)
  %.0402 = select i1 %.not243, float %1481, float %1480
  %1482 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not244 = icmp eq i32 %1482, 0
  %1483 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0402, float %1460, float noundef 0x3FD334CCA0000000) #139
  %1484 = call float @llvm.fma.f32(float %.0402, float %1460, float noundef 0x3FD334CCA0000000)
  %.0403 = select i1 %.not244, float %1484, float %1483
  %1485 = fmul float %1460, %.0403
  br label %__nv_y0f.exit406

__nv_isinff.exit.i376:                            ; preds = %1457
  %1486 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not200 = icmp eq i32 %1486, 0
  %1487 = call float @llvm.nvvm.fabs.ftz.f(float %.0291) #139
  %1488 = call float @llvm.fabs.f32(float %.0291)
  %.0404 = select i1 %.not200, float %1488, float %1487
  %1489 = fcmp oeq float %.0404, 0x7FF0000000000000
  br i1 %1489, label %__nv_y0f.exit406, label %__internal_fmad.exit70.i377

__internal_fmad.exit70.i377:                      ; preds = %__nv_isinff.exit.i376
  %1490 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0291) #141, !srcloc !73
  %1491 = fmul float %1490, %1490
  %1492 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not201 = icmp eq i32 %1492, 0
  %1493 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFD98D3E60000000, float %1491, float noundef 0x3FBA1ABE00000000) #139
  %1494 = call float @llvm.fma.f32(float %1491, float noundef 0xBFD98D3E60000000, float noundef 0x3FBA1ABE00000000)
  %.0405 = select i1 %.not201, float %1494, float %1493
  %1495 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not202 = icmp eq i32 %1495, 0
  %1496 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0405, float %1491, float noundef 0xBFAFFF0AA0000000) #139
  %1497 = call float @llvm.fma.f32(float %.0405, float %1491, float noundef 0xBFAFFF0AA0000000)
  %.0406 = select i1 %.not202, float %1497, float %1496
  %1498 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not203 = icmp eq i32 %1498, 0
  %1499 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0406, float %1491, float noundef 1.000000e+00) #139
  %1500 = call float @llvm.fma.f32(float %.0406, float %1491, float noundef 1.000000e+00)
  %.0407 = select i1 %.not203, float %1500, float %1499
  %1501 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not204 = icmp eq i32 %1501, 0
  %1502 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF199AC20000000, float %1491, float noundef 0xBFCA1A6380000000) #139
  %1503 = call float @llvm.fma.f32(float %1491, float noundef 0x3FF199AC20000000, float noundef 0xBFCA1A6380000000)
  %.0408 = select i1 %.not204, float %1503, float %1502
  %1504 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not205 = icmp eq i32 %1504, 0
  %1505 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0408, float %1491, float noundef 0x3FB0A8F060000000) #139
  %1506 = call float @llvm.fma.f32(float %.0408, float %1491, float noundef 0x3FB0A8F060000000)
  %.0409 = select i1 %.not205, float %1506, float %1505
  %1507 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not206 = icmp eq i32 %1507, 0
  %1508 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0409, float %1491, float noundef 0xBFBFFFFF60000000) #139
  %1509 = call float @llvm.fma.f32(float %.0409, float %1491, float noundef 0xBFBFFFFF60000000)
  %.0411 = select i1 %.not206, float %1509, float %1508
  %1510 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not207 = icmp eq i32 %1510, 0
  %1511 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0411, float %1490, float %.0291) #139
  %1512 = call float @llvm.fma.f32(float %.0411, float %1490, float %.0291)
  %.0412 = select i1 %.not207, float %1512, float %1511
  %1513 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not208 = icmp eq i32 %1513, 0
  br i1 %.not208, label %1516, label %1514

1514:                                             ; preds = %__internal_fmad.exit70.i377
  %1515 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0291) #139
  br label %__nv_rsqrtf.exit.i384

1516:                                             ; preds = %__internal_fmad.exit70.i377
  %1517 = call float @llvm.nvvm.rsqrt.approx.f(float %.0291) #139
  br label %__nv_rsqrtf.exit.i384

__nv_rsqrtf.exit.i384:                            ; preds = %1516, %1514
  %.0413 = phi float [ %1515, %1514 ], [ %1517, %1516 ]
  %1518 = fmul float %.0413, 0x3FE9884540000000
  %1519 = fmul float %.0407, %1518
  %1520 = fmul float %.0412, 0x3FE45F3060000000
  %1521 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not209 = icmp eq i32 %1521, 0
  %1522 = call i32 @llvm.nvvm.f2i.rn.ftz(float %1520) #139
  %1523 = call i32 @llvm.nvvm.f2i.rn(float %1520) #139
  %.0417 = select i1 %.not209, i32 %1523, i32 %1522
  %1524 = sitofp i32 %.0417 to float
  %1525 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not210 = icmp eq i32 %1525, 0
  %1526 = call float @llvm.nvvm.fma.rn.ftz.f(float %1524, float noundef 0xBFF921FB40000000, float %.0412) #139
  %1527 = call float @llvm.fma.f32(float %1524, float noundef 0xBFF921FB40000000, float %.0412)
  %.0418 = select i1 %.not210, float %1527, float %1526
  %1528 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not211 = icmp eq i32 %1528, 0
  %1529 = call float @llvm.nvvm.fma.rn.ftz.f(float %1524, float noundef 0xBE74442D00000000, float %.0418) #139
  %1530 = call float @llvm.fma.f32(float %1524, float noundef 0xBE74442D00000000, float %.0418)
  %.0419 = select i1 %.not211, float %1530, float %1529
  %1531 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not212 = icmp eq i32 %1531, 0
  %1532 = call float @llvm.nvvm.fma.rn.ftz.f(float %1524, float noundef 0xBCF84698A0000000, float %.0419) #139
  %1533 = call float @llvm.fma.f32(float %1524, float noundef 0xBCF84698A0000000, float %.0419)
  %.0420 = select i1 %.not212, float %1533, float %1532
  %1534 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not213 = icmp eq i32 %1534, 0
  %1535 = call float @llvm.nvvm.fabs.ftz.f(float %.0412) #139
  %1536 = call float @llvm.fabs.f32(float %.0412)
  %.0421 = select i1 %.not213, float %1536, float %1535
  %1537 = fcmp ult float %.0421, 1.056150e+05
  br i1 %1537, label %__internal_trig_reduction_kernel.exit.i.i393, label %__nv_isinff.exit.i.i.i390

__nv_isinff.exit.i.i.i390:                        ; preds = %__nv_rsqrtf.exit.i384
  %1538 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not230 = icmp eq i32 %1538, 0
  %.0422 = select i1 %.not230, float %1536, float %1535
  %1539 = fcmp oeq float %.0422, 0x7FF0000000000000
  br i1 %1539, label %__nv_fmul_rn.exit.i.i.i391, label %1543

__nv_fmul_rn.exit.i.i.i391:                       ; preds = %__nv_isinff.exit.i.i.i390
  %1540 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not236 = icmp eq i32 %1540, 0
  %1541 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0412, float noundef 0.000000e+00) #139
  %1542 = fmul float %.0412, 0.000000e+00
  %.0423 = select i1 %.not236, float %1542, float %1541
  br label %__internal_trig_reduction_kernel.exit.i.i393

1543:                                             ; preds = %__nv_isinff.exit.i.i.i390
  %1544 = bitcast float %.0412 to i32
  %1545 = lshr i32 %1544, 23
  %1546 = and i32 %1545, 255
  %1547 = add nsw i32 %1546, -128
  %1548 = shl i32 %1544, 8
  %1549 = or i32 %1548, -2147483648
  %1550 = lshr i32 %1547, 5
  br label %1551

1551:                                             ; preds = %1543, %1551
  %iq.i.i.i.i210.0557 = phi i32 [ 0, %1543 ], [ %1559, %1551 ]
  %hi.i.i.i.i205.0556 = phi i32 [ 0, %1543 ], [ %1557, %1551 ]
  %1552 = zext i32 %iq.i.i.i.i210.0557 to i64
  %1553 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %1552
  %1554 = load i32, i32 addrspace(1)* %1553, align 4
  %1555 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %1554, i32 %1549, i32 %hi.i.i.i.i205.0556) #141, !srcloc !67
  %1556 = extractvalue { i32, i32 } %1555, 0
  %1557 = extractvalue { i32, i32 } %1555, 1
  %1558 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1552
  store i32 %1556, i32* %1558, align 4, !noalias !68
  %1559 = add nuw nsw i32 %iq.i.i.i.i210.0557, 1
  %exitcond.not = icmp eq i32 %1559, 6
  br i1 %exitcond.not, label %1560, label %1551, !llvm.loop !71

1560:                                             ; preds = %1551
  %1561 = and i32 %1544, -2147483648
  %1562 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %1557, i32* %1562, align 4, !noalias !68
  %1563 = and i32 %1545, 31
  %1564 = sub nsw i32 6, %1550
  %1565 = sext i32 %1564 to i64
  %1566 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1565
  %1567 = load i32, i32* %1566, align 4
  %1568 = sub nsw i32 5, %1550
  %1569 = sext i32 %1568 to i64
  %1570 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1569
  %1571 = load i32, i32* %1570, align 4
  %.not231 = icmp eq i32 %1563, 0
  br i1 %.not231, label %__internal_trig_reduction_slowpath.exit.i.i.i392, label %1572

1572:                                             ; preds = %1560
  %1573 = sub nsw i32 4, %1550
  %1574 = sub nuw nsw i32 32, %1563
  %1575 = shl i32 %1567, %1563
  %1576 = lshr i32 %1571, %1574
  %1577 = add i32 %1576, %1575
  %1578 = shl i32 %1571, %1563
  %1579 = sext i32 %1573 to i64
  %1580 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1579
  %1581 = load i32, i32* %1580, align 4
  %1582 = lshr i32 %1581, %1574
  %1583 = add i32 %1582, %1578
  br label %__internal_trig_reduction_slowpath.exit.i.i.i392

__internal_trig_reduction_slowpath.exit.i.i.i392: ; preds = %1572, %1560
  %hi.i.i.i.i205.1 = phi i32 [ %1577, %1572 ], [ %1567, %1560 ]
  %lo.i.i.i.i206.0 = phi i32 [ %1583, %1572 ], [ %1571, %1560 ]
  %1584 = lshr i32 %hi.i.i.i.i205.1, 30
  %1585 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i205.1, i32 %lo.i.i.i.i206.0, i32 noundef 2)
  %1586 = shl i32 %lo.i.i.i.i206.0, 2
  %1587 = lshr i32 %1585, 31
  %1588 = add nuw nsw i32 %1587, %1584
  %.not232 = icmp eq i32 %1561, 0
  %1589 = sub nsw i32 0, %1588
  %spec.select5 = select i1 %.not232, i32 %1588, i32 %1589
  %.not233 = icmp sgt i32 %1585, -1
  %1590 = xor i32 %1561, -2147483648
  %s.i.i.i.i204.0 = select i1 %.not233, i32 %1561, i32 %1590
  %not..not233 = xor i1 %.not233, true
  %1591 = sext i1 %not..not233 to i32
  %hi.i.i.i.i205.2 = xor i32 %1585, %1591
  %lo.i.i.i.i206.1 = xor i32 %1586, %1591
  %1592 = zext i32 %hi.i.i.i.i205.2 to i64
  %1593 = shl nuw i64 %1592, 32
  %1594 = zext i32 %lo.i.i.i.i206.1 to i64
  %1595 = or i64 %1593, %1594
  %1596 = sitofp i64 %1595 to double
  %1597 = fmul double %1596, 0x3BF921FB54442D19
  %1598 = fptrunc double %1597 to float
  %.not235 = icmp eq i32 %s.i.i.i.i204.0, 0
  %1599 = fneg float %1598
  %r.i.i.i.i214.0 = select i1 %.not235, float %1598, float %1599
  br label %__internal_trig_reduction_kernel.exit.i.i393

__internal_trig_reduction_kernel.exit.i.i393:     ; preds = %__nv_fmul_rn.exit.i.i.i391, %__internal_trig_reduction_slowpath.exit.i.i.i392, %__nv_rsqrtf.exit.i384
  %i.i85.i220.1 = phi i32 [ %.0417, %__nv_rsqrtf.exit.i384 ], [ 0, %__nv_fmul_rn.exit.i.i.i391 ], [ %spec.select5, %__internal_trig_reduction_slowpath.exit.i.i.i392 ]
  %t.i.i.i217.1 = phi float [ %.0420, %__nv_rsqrtf.exit.i384 ], [ %.0423, %__nv_fmul_rn.exit.i.i.i391 ], [ %r.i.i.i.i214.0, %__internal_trig_reduction_slowpath.exit.i.i.i392 ]
  %1600 = and i32 %i.i85.i220.1, 3
  %1601 = uitofp i32 %1600 to float
  %1602 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not214 = icmp eq i32 %1602, 0
  %1603 = call float @llvm.nvvm.fma.rn.ftz.f(float %1601, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %1604 = call float @llvm.fma.f32(float %1601, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.0424 = select i1 %.not214, float %1604, float %1603
  %1605 = fadd float %t.i.i.i217.1, %.0424
  %1606 = fmul float %1605, 0x3FE45F3060000000
  %1607 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not215 = icmp eq i32 %1607, 0
  %1608 = call i32 @llvm.nvvm.f2i.rn.ftz(float %1606) #139
  %1609 = call i32 @llvm.nvvm.f2i.rn(float %1606) #139
  %.0425 = select i1 %.not215, i32 %1609, i32 %1608
  %1610 = sitofp i32 %.0425 to float
  %1611 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not216 = icmp eq i32 %1611, 0
  %1612 = call float @llvm.nvvm.fma.rn.ftz.f(float %1610, float noundef 0xBFF921FB40000000, float %1605) #139
  %1613 = call float @llvm.fma.f32(float %1610, float noundef 0xBFF921FB40000000, float %1605)
  %.0426 = select i1 %.not216, float %1613, float %1612
  %1614 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not217 = icmp eq i32 %1614, 0
  %1615 = call float @llvm.nvvm.fma.rn.ftz.f(float %1610, float noundef 0xBE74442D00000000, float %.0426) #139
  %1616 = call float @llvm.fma.f32(float %1610, float noundef 0xBE74442D00000000, float %.0426)
  %.0427 = select i1 %.not217, float %1616, float %1615
  %1617 = add i32 %.0425, 1
  %1618 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not218 = icmp eq i32 %1618, 0
  %1619 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0427, float %.0427) #139
  %1620 = fmul float %.0427, %.0427
  %.0428 = select i1 %.not218, float %1620, float %1619
  %1621 = and i32 %1617, 1
  %.not219 = icmp eq i32 %1621, 0
  %1622 = select i1 %.not219, float %.0427, float 1.000000e+00
  %1623 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not220 = icmp eq i32 %1623, 0
  %1624 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0428, float %1622, float noundef 0.000000e+00) #139
  %1625 = call float @llvm.fma.f32(float %.0428, float %1622, float noundef 0.000000e+00)
  %.0429 = select i1 %.not220, float %1625, float %1624
  br i1 %.not219, label %__internal_fmad.exit2.i.i.i401, label %__internal_fmad.exit1.i.i.i400

__internal_fmad.exit1.i.i.i400:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i393
  %1626 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not229 = icmp eq i32 %1626, 0
  %1627 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0428, float noundef 0xBF56C0FDA0000000) #139
  %1628 = call float @llvm.fma.f32(float %.0428, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0431 = select i1 %.not229, float %1628, float %1627
  br label %__internal_fmad.exit2.i.i.i401

__internal_fmad.exit2.i.i.i401:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i393, %__internal_fmad.exit1.i.i.i400
  %1629 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i393 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i400 ]
  %1630 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i393 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i400 ]
  %1631 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i393 ], [ %.0431, %__internal_fmad.exit1.i.i.i400 ]
  %1632 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not223 = icmp eq i32 %1632, 0
  %1633 = call float @llvm.nvvm.fma.rn.ftz.f(float %1631, float %.0428, float noundef %1630) #139
  %1634 = call float @llvm.fma.f32(float %1631, float %.0428, float noundef %1630)
  %.0432 = select i1 %.not223, float %1634, float %1633
  %1635 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not225 = icmp eq i32 %1635, 0
  %1636 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0432, float %.0428, float noundef %1629) #139
  %1637 = call float @llvm.fma.f32(float %.0432, float %.0428, float noundef %1629)
  %.0433 = select i1 %.not225, float %1637, float %1636
  %1638 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not226 = icmp eq i32 %1638, 0
  %1639 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0433, float %.0429, float %1622) #139
  %1640 = call float @llvm.fma.f32(float %.0433, float %.0429, float %1622)
  %.0434 = select i1 %.not226, float %1640, float %1639
  %1641 = and i32 %1617, 2
  %.not227 = icmp eq i32 %1641, 0
  br i1 %.not227, label %__internal_cos_offset_f.exit.i405, label %__internal_fmad.exit5.i.i.i404

__internal_fmad.exit5.i.i.i404:                   ; preds = %__internal_fmad.exit2.i.i.i401
  %1642 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not228 = icmp eq i32 %1642, 0
  %1643 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0434, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %1644 = call float @llvm.fma.f32(float %.0434, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0436 = select i1 %.not228, float %1644, float %1643
  br label %__internal_cos_offset_f.exit.i405

__internal_cos_offset_f.exit.i405:                ; preds = %__internal_fmad.exit5.i.i.i404, %__internal_fmad.exit2.i.i.i401
  %z.i.i83.i197.0 = phi float [ %.0436, %__internal_fmad.exit5.i.i.i404 ], [ %.0434, %__internal_fmad.exit2.i.i.i401 ]
  %1645 = fmul float %1519, %z.i.i83.i197.0
  br label %__nv_y0f.exit406

__nv_y0f.exit406:                                 ; preds = %__internal_fmad.exit9.i347, %__internal_fmad.exit51.i368, %__nv_isinff.exit.i376, %__internal_cos_offset_f.exit.i405, %__internal_fmad.exit33.i359, %__nv_j0f.exit.i345
  %r.i274.4 = phi float [ %.0369, %__nv_j0f.exit.i345 ], [ %1424, %__internal_fmad.exit9.i347 ], [ %1456, %__internal_fmad.exit33.i359 ], [ %1485, %__internal_fmad.exit51.i368 ], [ %1645, %__internal_cos_offset_f.exit.i405 ], [ 0.000000e+00, %__nv_isinff.exit.i376 ]
  %r.i274.5 = select i1 %551, float 0x7FFFFFFFE0000000, float %r.i274.4
  br label %2229

__nv_fabsf.exit.i496:                             ; preds = %__nv_y0f.exit
  %1646 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not65 = icmp eq i32 %1646, 0
  %.0439 = select i1 %.not65, float %2, float %1
  %1647 = fcmp olt float %.0439, 0x3810000000000000
  br i1 %1647, label %1648, label %1650

1648:                                             ; preds = %__nv_fabsf.exit.i496
  %1649 = fdiv float 0xBFE45F3060000000, %.0439
  br label %__nv_y1f.exit618

1650:                                             ; preds = %__nv_fabsf.exit.i496
  %1651 = fcmp ugt float %.0439, 0x3FFB2D5880000000
  br i1 %1651, label %1942, label %__internal_fmad.exit.i497

__internal_fmad.exit.i497:                        ; preds = %1650
  %1652 = fmul float %.0439, %.0439
  %1653 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not129 = icmp eq i32 %1653, 0
  %1654 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E428C5980000000, float %1652, float noundef 0xBEB0A4FB40000000) #139
  %1655 = call float @llvm.fma.f32(float %1652, float noundef 0x3E428C5980000000, float noundef 0xBEB0A4FB40000000)
  %.0441 = select i1 %.not129, float %1655, float %1654
  %1656 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not130 = icmp eq i32 %1656, 0
  %1657 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0441, float %1652, float noundef 0x3F12C7D2A0000000) #139
  %1658 = call float @llvm.fma.f32(float %.0441, float %1652, float noundef 0x3F12C7D2A0000000)
  %.0442 = select i1 %.not130, float %1658, float %1657
  %1659 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not131 = icmp eq i32 %1659, 0
  %1660 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0442, float %1652, float noundef 0xBF6835B960000000) #139
  %1661 = call float @llvm.fma.f32(float %.0442, float %1652, float noundef 0xBF6835B960000000)
  %.0443 = select i1 %.not131, float %1661, float %1660
  %1662 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not132 = icmp eq i32 %1662, 0
  %1663 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0443, float %1652, float noundef 0x3FABD39760000000) #139
  %1664 = call float @llvm.fma.f32(float %.0443, float %1652, float noundef 0x3FABD39760000000)
  %.0445 = select i1 %.not132, float %1664, float %1663
  %1665 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not133 = icmp eq i32 %1665, 0
  %1666 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0445, float %1652, float noundef 0xBFC9186620000000) #139
  %1667 = call float @llvm.fma.f32(float %.0445, float %1652, float noundef 0xBFC9186620000000)
  %.0446 = select i1 %.not133, float %1667, float %1666
  %1668 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not134 = icmp eq i32 %1668, 0
  %1669 = call float @llvm.nvvm.fabs.ftz.f(float %.0439) #139
  %1670 = call float @llvm.fabs.f32(float %.0439)
  %.0448 = select i1 %.not134, float %1670, float %1669
  %1671 = fcmp ugt float %.0448, 8.000000e+00
  br i1 %1671, label %__nv_isinff.exit.i.i515, label %__internal_fmad.exit.i.i503

__internal_fmad.exit.i.i503:                      ; preds = %__internal_fmad.exit.i497
  %1672 = fadd float %.0448, 0xC00EA75580000000
  %1673 = fadd float %1672, 0x3E74A121E0000000
  %1674 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not187 = icmp eq i32 %1674, 0
  %1675 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3D35E68C60000000, float %1673, float noundef 0x3D7037E840000000) #139
  %1676 = call float @llvm.fma.f32(float %1673, float noundef 0x3D35E68C60000000, float noundef 0x3D7037E840000000)
  %.0449 = select i1 %.not187, float %1676, float %1675
  %1677 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not188 = icmp eq i32 %1677, 0
  %1678 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0449, float %1673, float noundef 0xBDBC43D820000000) #139
  %1679 = call float @llvm.fma.f32(float %.0449, float %1673, float noundef 0xBDBC43D820000000)
  %.0450 = select i1 %.not188, float %1679, float %1678
  %1680 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not189 = icmp eq i32 %1680, 0
  %1681 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0450, float %1673, float noundef 0xBDEBBBDFE0000000) #139
  %1682 = call float @llvm.fma.f32(float %.0450, float %1673, float noundef 0xBDEBBBDFE0000000)
  %.0451 = select i1 %.not189, float %1682, float %1681
  %1683 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not190 = icmp eq i32 %1683, 0
  %1684 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0451, float %1673, float noundef 0x3E336193A0000000) #139
  %1685 = call float @llvm.fma.f32(float %.0451, float %1673, float noundef 0x3E336193A0000000)
  %.0453 = select i1 %.not190, float %1685, float %1684
  %1686 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not191 = icmp eq i32 %1686, 0
  %1687 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0453, float %1673, float noundef 0x3E5D022E60000000) #139
  %1688 = call float @llvm.fma.f32(float %.0453, float %1673, float noundef 0x3E5D022E60000000)
  %.0455 = select i1 %.not191, float %1688, float %1687
  %1689 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not192 = icmp eq i32 %1689, 0
  %1690 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0455, float %1673, float noundef 0xBEA1F1B900000000) #139
  %1691 = call float @llvm.fma.f32(float %.0455, float %1673, float noundef 0xBEA1F1B900000000)
  %.0456 = select i1 %.not192, float %1691, float %1690
  %1692 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not193 = icmp eq i32 %1692, 0
  %1693 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0456, float %1673, float noundef 0xBEC3CCA7A0000000) #139
  %1694 = call float @llvm.fma.f32(float %.0456, float %1673, float noundef 0xBEC3CCA7A0000000)
  %.0457 = select i1 %.not193, float %1694, float %1693
  %1695 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not194 = icmp eq i32 %1695, 0
  %1696 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0457, float %1673, float noundef 0x3F059B38A0000000) #139
  %1697 = call float @llvm.fma.f32(float %.0457, float %1673, float noundef 0x3F059B38A0000000)
  %.0458 = select i1 %.not194, float %1697, float %1696
  %1698 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not195 = icmp eq i32 %1698, 0
  %1699 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0458, float %1673, float noundef 0x3F1F3D6200000000) #139
  %1700 = call float @llvm.fma.f32(float %.0458, float %1673, float noundef 0x3F1F3D6200000000)
  %.0459 = select i1 %.not195, float %1700, float %1699
  %1701 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not196 = icmp eq i32 %1701, 0
  %1702 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0459, float %1673, float noundef 0xBF5D9D7380000000) #139
  %1703 = call float @llvm.fma.f32(float %.0459, float %1673, float noundef 0xBF5D9D7380000000)
  %.0460 = select i1 %.not196, float %1703, float %1702
  %1704 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not197 = icmp eq i32 %1704, 0
  %1705 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0460, float %1673, float noundef 0xBF64EDFFA0000000) #139
  %1706 = call float @llvm.fma.f32(float %.0460, float %1673, float noundef 0xBF64EDFFA0000000)
  %.0462 = select i1 %.not197, float %1706, float %1705
  %1707 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not198 = icmp eq i32 %1707, 0
  %1708 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0462, float %1673, float noundef 0x3FA0E73260000000) #139
  %1709 = call float @llvm.fma.f32(float %.0462, float %1673, float noundef 0x3FA0E73260000000)
  %.0463 = select i1 %.not198, float %1709, float %1708
  %1710 = fadd float %.0448, 0xC01C0FF600000000
  %1711 = fadd float %1710, 0x3E88971B60000000
  %1712 = fmul float %1711, %.0463
  %1713 = fmul float %1673, %1712
  %1714 = fmul float %.0448, %1713
  br label %1878

__nv_isinff.exit.i.i515:                          ; preds = %__internal_fmad.exit.i497
  %1715 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not135 = icmp eq i32 %1715, 0
  %1716 = call float @llvm.nvvm.fabs.ftz.f(float %.0448) #139
  %1717 = call float @llvm.fabs.f32(float %.0448)
  %.0464 = select i1 %.not135, float %1717, float %1716
  %1718 = fcmp oeq float %.0464, 0x7FF0000000000000
  br i1 %1718, label %1878, label %__internal_fmad.exit12.i.i516

__internal_fmad.exit12.i.i516:                    ; preds = %__nv_isinff.exit.i.i515
  %1719 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0448) #141, !srcloc !73
  %1720 = fmul float %1719, %1719
  %1721 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not136 = icmp eq i32 %1721, 0
  %1722 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC0105966E0000000, float %1720, float noundef 0x3FE7FEFD20000000) #139
  %1723 = call float @llvm.fma.f32(float %1720, float noundef 0xC0105966E0000000, float noundef 0x3FE7FEFD20000000)
  %.0465 = select i1 %.not136, float %1723, float %1722
  %1724 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not137 = icmp eq i32 %1724, 0
  %1725 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0465, float %1720, float noundef 0xBFC8B175C0000000) #139
  %1726 = call float @llvm.fma.f32(float %.0465, float %1720, float noundef 0xBFC8B175C0000000)
  %.0466 = select i1 %.not137, float %1726, float %1725
  %1727 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not138 = icmp eq i32 %1727, 0
  %1728 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0466, float %1720, float noundef 0x3FC7FFF160000000) #139
  %1729 = call float @llvm.fma.f32(float %.0466, float %1720, float noundef 0x3FC7FFF160000000)
  %.0467 = select i1 %.not138, float %1729, float %1728
  %1730 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not139 = icmp eq i32 %1730, 0
  %1731 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0467, float %1720, float noundef 1.000000e+00) #139
  %1732 = call float @llvm.fma.f32(float %.0467, float %1720, float noundef 1.000000e+00)
  %.0469 = select i1 %.not139, float %1732, float %1731
  %1733 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not140 = icmp eq i32 %1733, 0
  %1734 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFF9477440000000, float %1720, float noundef 0x3FD72295A0000000) #139
  %1735 = call float @llvm.fma.f32(float %1720, float noundef 0xBFF9477440000000, float noundef 0x3FD72295A0000000)
  %.0470 = select i1 %.not140, float %1735, float %1734
  %1736 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not141 = icmp eq i32 %1736, 0
  %1737 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0470, float %1720, float noundef 0xBFC4FE5D80000000) #139
  %1738 = call float @llvm.fma.f32(float %.0470, float %1720, float noundef 0xBFC4FE5D80000000)
  %.0468 = select i1 %.not141, float %1738, float %1737
  %1739 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not142 = icmp eq i32 %1739, 0
  %1740 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0468, float %1720, float noundef 0x3FD7FFFFA0000000) #139
  %1741 = call float @llvm.fma.f32(float %.0468, float %1720, float noundef 0x3FD7FFFFA0000000)
  %.0461 = select i1 %.not142, float %1741, float %1740
  %1742 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not143 = icmp eq i32 %1742, 0
  %1743 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0461, float %1719, float %.0448) #139
  %1744 = call float @llvm.fma.f32(float %.0461, float %1719, float %.0448)
  %.0454 = select i1 %.not143, float %1744, float %1743
  %1745 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not144 = icmp eq i32 %1745, 0
  br i1 %.not144, label %1748, label %1746

1746:                                             ; preds = %__internal_fmad.exit12.i.i516
  %1747 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0448) #139
  br label %__nv_rsqrtf.exit.i.i524

1748:                                             ; preds = %__internal_fmad.exit12.i.i516
  %1749 = call float @llvm.nvvm.rsqrt.approx.f(float %.0448) #139
  br label %__nv_rsqrtf.exit.i.i524

__nv_rsqrtf.exit.i.i524:                          ; preds = %1748, %1746
  %.0452 = phi float [ %1747, %1746 ], [ %1749, %1748 ]
  %1750 = fmul float %.0452, 0x3FE9884540000000
  %1751 = fmul float %.0469, %1750
  %1752 = fmul float %.0454, 0x3FE45F3060000000
  %1753 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not145 = icmp eq i32 %1753, 0
  %1754 = call i32 @llvm.nvvm.f2i.rn.ftz(float %1752) #139
  %1755 = call i32 @llvm.nvvm.f2i.rn(float %1752) #139
  %.0447 = select i1 %.not145, i32 %1755, i32 %1754
  %1756 = sitofp i32 %.0447 to float
  %1757 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not146 = icmp eq i32 %1757, 0
  %1758 = call float @llvm.nvvm.fma.rn.ftz.f(float %1756, float noundef 0xBFF921FB40000000, float %.0454) #139
  %1759 = call float @llvm.fma.f32(float %1756, float noundef 0xBFF921FB40000000, float %.0454)
  %.0444 = select i1 %.not146, float %1759, float %1758
  %1760 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not147 = icmp eq i32 %1760, 0
  %1761 = call float @llvm.nvvm.fma.rn.ftz.f(float %1756, float noundef 0xBE74442D00000000, float %.0444) #139
  %1762 = call float @llvm.fma.f32(float %1756, float noundef 0xBE74442D00000000, float %.0444)
  %.0440 = select i1 %.not147, float %1762, float %1761
  %1763 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not148 = icmp eq i32 %1763, 0
  %1764 = call float @llvm.nvvm.fma.rn.ftz.f(float %1756, float noundef 0xBCF84698A0000000, float %.0440) #139
  %1765 = call float @llvm.fma.f32(float %1756, float noundef 0xBCF84698A0000000, float %.0440)
  %.0438 = select i1 %.not148, float %1765, float %1764
  %1766 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not149 = icmp eq i32 %1766, 0
  %1767 = call float @llvm.nvvm.fabs.ftz.f(float %.0454) #139
  %1768 = call float @llvm.fabs.f32(float %.0454)
  %.0437 = select i1 %.not149, float %1768, float %1767
  %1769 = fcmp ult float %.0437, 1.056150e+05
  br i1 %1769, label %__internal_trig_reduction_kernel.exit.i.i.i533, label %__nv_isinff.exit.i.i.i.i530

__nv_isinff.exit.i.i.i.i530:                      ; preds = %__nv_rsqrtf.exit.i.i524
  %1770 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not180 = icmp eq i32 %1770, 0
  %.0435 = select i1 %.not180, float %1768, float %1767
  %1771 = fcmp oeq float %.0435, 0x7FF0000000000000
  br i1 %1771, label %__nv_fmul_rn.exit.i.i.i.i531, label %1775

__nv_fmul_rn.exit.i.i.i.i531:                     ; preds = %__nv_isinff.exit.i.i.i.i530
  %1772 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not186 = icmp eq i32 %1772, 0
  %1773 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0454, float noundef 0.000000e+00) #139
  %1774 = fmul float %.0454, 0.000000e+00
  %.0430 = select i1 %.not186, float %1774, float %1773
  br label %__internal_trig_reduction_kernel.exit.i.i.i533

1775:                                             ; preds = %__nv_isinff.exit.i.i.i.i530
  %1776 = bitcast float %.0454 to i32
  %1777 = lshr i32 %1776, 23
  %1778 = and i32 %1777, 255
  %1779 = add nsw i32 %1778, -128
  %1780 = shl i32 %1776, 8
  %1781 = or i32 %1780, -2147483648
  %1782 = lshr i32 %1779, 5
  br label %1783

1783:                                             ; preds = %1775, %1783
  %hi.i.i.i.i.i461.0563 = phi i32 [ 0, %1775 ], [ %1789, %1783 ]
  %iq.i.i.i.i.i466.0562 = phi i32 [ 0, %1775 ], [ %1791, %1783 ]
  %1784 = zext i32 %iq.i.i.i.i.i466.0562 to i64
  %1785 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %1784
  %1786 = load i32, i32 addrspace(1)* %1785, align 4
  %1787 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %1786, i32 %1781, i32 %hi.i.i.i.i.i461.0563) #141, !srcloc !67
  %1788 = extractvalue { i32, i32 } %1787, 0
  %1789 = extractvalue { i32, i32 } %1787, 1
  %1790 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1784
  store i32 %1788, i32* %1790, align 4, !noalias !68
  %1791 = add nuw nsw i32 %iq.i.i.i.i.i466.0562, 1
  %exitcond597.not = icmp eq i32 %1791, 6
  br i1 %exitcond597.not, label %1792, label %1783, !llvm.loop !71

1792:                                             ; preds = %1783
  %1793 = and i32 %1776, -2147483648
  %1794 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %1789, i32* %1794, align 4, !noalias !68
  %1795 = and i32 %1777, 31
  %1796 = sub nsw i32 6, %1782
  %1797 = sext i32 %1796 to i64
  %1798 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1797
  %1799 = load i32, i32* %1798, align 4
  %1800 = sub nsw i32 5, %1782
  %1801 = sext i32 %1800 to i64
  %1802 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1801
  %1803 = load i32, i32* %1802, align 4
  %.not181 = icmp eq i32 %1795, 0
  br i1 %.not181, label %__internal_trig_reduction_slowpath.exit.i.i.i.i532, label %1804

1804:                                             ; preds = %1792
  %1805 = sub nsw i32 4, %1782
  %1806 = sub nuw nsw i32 32, %1795
  %1807 = shl i32 %1799, %1795
  %1808 = lshr i32 %1803, %1806
  %1809 = add i32 %1808, %1807
  %1810 = shl i32 %1803, %1795
  %1811 = sext i32 %1805 to i64
  %1812 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %1811
  %1813 = load i32, i32* %1812, align 4
  %1814 = lshr i32 %1813, %1806
  %1815 = add i32 %1814, %1810
  br label %__internal_trig_reduction_slowpath.exit.i.i.i.i532

__internal_trig_reduction_slowpath.exit.i.i.i.i532: ; preds = %1804, %1792
  %lo.i.i.i.i.i462.0 = phi i32 [ %1815, %1804 ], [ %1803, %1792 ]
  %hi.i.i.i.i.i461.1 = phi i32 [ %1809, %1804 ], [ %1799, %1792 ]
  %1816 = lshr i32 %hi.i.i.i.i.i461.1, 30
  %1817 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.i461.1, i32 %lo.i.i.i.i.i462.0, i32 noundef 2)
  %1818 = shl i32 %lo.i.i.i.i.i462.0, 2
  %1819 = lshr i32 %1817, 31
  %1820 = add nuw nsw i32 %1819, %1816
  %.not182 = icmp eq i32 %1793, 0
  %1821 = sub nsw i32 0, %1820
  %spec.select6 = select i1 %.not182, i32 %1820, i32 %1821
  %.not183 = icmp sgt i32 %1817, -1
  %1822 = xor i32 %1793, -2147483648
  %not..not183 = xor i1 %.not183, true
  %1823 = sext i1 %not..not183 to i32
  %lo.i.i.i.i.i462.1 = xor i32 %1818, %1823
  %hi.i.i.i.i.i461.2 = xor i32 %1817, %1823
  %s.i.i.i.i.i459.0 = select i1 %.not183, i32 %1793, i32 %1822
  %1824 = zext i32 %hi.i.i.i.i.i461.2 to i64
  %1825 = shl nuw i64 %1824, 32
  %1826 = zext i32 %lo.i.i.i.i.i462.1 to i64
  %1827 = or i64 %1825, %1826
  %1828 = sitofp i64 %1827 to double
  %1829 = fmul double %1828, 0x3BF921FB54442D19
  %1830 = fptrunc double %1829 to float
  %.not185 = icmp eq i32 %s.i.i.i.i.i459.0, 0
  %1831 = fneg float %1830
  %r.i.i.i.i.i470.0 = select i1 %.not185, float %1830, float %1831
  br label %__internal_trig_reduction_kernel.exit.i.i.i533

__internal_trig_reduction_kernel.exit.i.i.i533:   ; preds = %__nv_fmul_rn.exit.i.i.i.i531, %__internal_trig_reduction_slowpath.exit.i.i.i.i532, %__nv_rsqrtf.exit.i.i524
  %i.i.i.i476.1 = phi i32 [ %.0447, %__nv_rsqrtf.exit.i.i524 ], [ 0, %__nv_fmul_rn.exit.i.i.i.i531 ], [ %spec.select6, %__internal_trig_reduction_slowpath.exit.i.i.i.i532 ]
  %t.i.i.i.i473.1 = phi float [ %.0438, %__nv_rsqrtf.exit.i.i524 ], [ %.0430, %__nv_fmul_rn.exit.i.i.i.i531 ], [ %r.i.i.i.i.i470.0, %__internal_trig_reduction_slowpath.exit.i.i.i.i532 ]
  %1832 = and i32 %i.i.i.i476.1, 3
  %1833 = uitofp i32 %1832 to float
  %1834 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not150 = icmp eq i32 %1834, 0
  %1835 = call float @llvm.nvvm.fma.rn.ftz.f(float %1833, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %1836 = call float @llvm.fma.f32(float %1833, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.0416 = select i1 %.not150, float %1836, float %1835
  %1837 = fadd float %t.i.i.i.i473.1, %.0416
  %1838 = fmul float %1837, 0x3FE45F3060000000
  %1839 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not151 = icmp eq i32 %1839, 0
  %1840 = call i32 @llvm.nvvm.f2i.rn.ftz(float %1838) #139
  %1841 = call i32 @llvm.nvvm.f2i.rn(float %1838) #139
  %.0415 = select i1 %.not151, i32 %1841, i32 %1840
  %1842 = sitofp i32 %.0415 to float
  %1843 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not152 = icmp eq i32 %1843, 0
  %1844 = call float @llvm.nvvm.fma.rn.ftz.f(float %1842, float noundef 0xBFF921FB40000000, float %1837) #139
  %1845 = call float @llvm.fma.f32(float %1842, float noundef 0xBFF921FB40000000, float %1837)
  %.0414 = select i1 %.not152, float %1845, float %1844
  %1846 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not153 = icmp eq i32 %1846, 0
  %1847 = call float @llvm.nvvm.fma.rn.ftz.f(float %1842, float noundef 0xBE74442D00000000, float %.0414) #139
  %1848 = call float @llvm.fma.f32(float %1842, float noundef 0xBE74442D00000000, float %.0414)
  %.0410 = select i1 %.not153, float %1848, float %1847
  %1849 = add i32 %.0415, 1
  %1850 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not154 = icmp eq i32 %1850, 0
  %1851 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0410, float %.0410) #139
  %1852 = fmul float %.0410, %.0410
  %.0401 = select i1 %.not154, float %1852, float %1851
  %1853 = and i32 %1849, 1
  %.not155 = icmp eq i32 %1853, 0
  %1854 = select i1 %.not155, float %.0410, float 1.000000e+00
  %1855 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not156 = icmp eq i32 %1855, 0
  %1856 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0401, float %1854, float noundef 0.000000e+00) #139
  %1857 = call float @llvm.fma.f32(float %.0401, float %1854, float noundef 0.000000e+00)
  %.0394 = select i1 %.not156, float %1857, float %1856
  br i1 %.not155, label %__internal_fmad.exit2.i.i.i.i541, label %__internal_fmad.exit1.i.i.i.i540

__internal_fmad.exit1.i.i.i.i540:                 ; preds = %__internal_trig_reduction_kernel.exit.i.i.i533
  %1858 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not179 = icmp eq i32 %1858, 0
  %1859 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0401, float noundef 0xBF56C0FDA0000000) #139
  %1860 = call float @llvm.fma.f32(float %.0401, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0387 = select i1 %.not179, float %1860, float %1859
  br label %__internal_fmad.exit2.i.i.i.i541

__internal_fmad.exit2.i.i.i.i541:                 ; preds = %__internal_trig_reduction_kernel.exit.i.i.i533, %__internal_fmad.exit1.i.i.i.i540
  %1861 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i.i533 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i.i540 ]
  %1862 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i.i533 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i.i540 ]
  %1863 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i.i533 ], [ %.0387, %__internal_fmad.exit1.i.i.i.i540 ]
  %1864 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not159 = icmp eq i32 %1864, 0
  %1865 = call float @llvm.nvvm.fma.rn.ftz.f(float %1863, float %.0401, float noundef %1862) #139
  %1866 = call float @llvm.fma.f32(float %1863, float %.0401, float noundef %1862)
  %.0380 = select i1 %.not159, float %1866, float %1865
  %1867 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not161 = icmp eq i32 %1867, 0
  %1868 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0380, float %.0401, float noundef %1861) #139
  %1869 = call float @llvm.fma.f32(float %.0380, float %.0401, float noundef %1861)
  %.0373 = select i1 %.not161, float %1869, float %1868
  %1870 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not162 = icmp eq i32 %1870, 0
  %1871 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0373, float %.0394, float %1854) #139
  %1872 = call float @llvm.fma.f32(float %.0373, float %.0394, float %1854)
  %.0366 = select i1 %.not162, float %1872, float %1871
  %1873 = and i32 %1849, 2
  %.not163 = icmp eq i32 %1873, 0
  br i1 %.not163, label %__internal_cos_offset_f.exit.i.i545, label %__internal_fmad.exit5.i.i.i.i544

__internal_fmad.exit5.i.i.i.i544:                 ; preds = %__internal_fmad.exit2.i.i.i.i541
  %1874 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not178 = icmp eq i32 %1874, 0
  %1875 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0366, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %1876 = call float @llvm.fma.f32(float %.0366, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0360 = select i1 %.not178, float %1876, float %1875
  br label %__internal_cos_offset_f.exit.i.i545

__internal_cos_offset_f.exit.i.i545:              ; preds = %__internal_fmad.exit5.i.i.i.i544, %__internal_fmad.exit2.i.i.i.i541
  %z.i.i.i.i452.0 = phi float [ %.0360, %__internal_fmad.exit5.i.i.i.i544 ], [ %.0366, %__internal_fmad.exit2.i.i.i.i541 ]
  %1877 = fmul float %1751, %z.i.i.i.i452.0
  br label %1878

1878:                                             ; preds = %__internal_cos_offset_f.exit.i.i545, %__nv_isinff.exit.i.i515, %__internal_fmad.exit.i.i503
  %r.i.i482.1 = phi float [ %1714, %__internal_fmad.exit.i.i503 ], [ %1877, %__internal_cos_offset_f.exit.i.i545 ], [ 0.000000e+00, %__nv_isinff.exit.i.i515 ]
  %1879 = fcmp olt float %.0439, 0.000000e+00
  %1880 = fneg float %r.i.i482.1
  %r.i.i482.2 = select i1 %1879, float %1880, float %r.i.i482.1
  %1881 = fcmp olt float %.0448, 0x39B4484C00000000
  br i1 %1881, label %1882, label %.__nv_j1f.exit.i546_crit_edge

.__nv_j1f.exit.i546_crit_edge:                    ; preds = %1878
  %.pre = bitcast float %.0439 to i32
  br label %__nv_j1f.exit.i546

1882:                                             ; preds = %1878
  %1883 = bitcast float %.0439 to i32
  %1884 = and i32 %1883, -2147483648
  %1885 = bitcast float %r.i.i482.2 to i32
  %1886 = and i32 %1885, 2147483647
  %1887 = or i32 %1886, %1884
  %1888 = bitcast i32 %1887 to float
  br label %__nv_j1f.exit.i546

__nv_j1f.exit.i546:                               ; preds = %.__nv_j1f.exit.i546_crit_edge, %1882
  %.pre-phi = phi i32 [ %.pre, %.__nv_j1f.exit.i546_crit_edge ], [ %1883, %1882 ]
  %r.i.i482.3 = phi float [ %r.i.i482.2, %.__nv_j1f.exit.i546_crit_edge ], [ %1888, %1882 ]
  %1889 = add i32 %.pre-phi, -1059760811
  %1890 = and i32 %1889, -8388608
  %1891 = sub i32 %.pre-phi, %1890
  %1892 = bitcast i32 %1891 to float
  %1893 = sitofp i32 %1890 to float
  %1894 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not164 = icmp eq i32 %1894, 0
  %1895 = call float @llvm.nvvm.fma.rn.ftz.f(float %1893, float noundef 0x3E80000000000000, float noundef 0.000000e+00) #139
  %1896 = call float @llvm.fma.f32(float %1893, float noundef 0x3E80000000000000, float noundef 0.000000e+00)
  %.0328 = select i1 %.not164, float %1896, float %1895
  %1897 = fadd float %1892, -1.000000e+00
  %1898 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not165 = icmp eq i32 %1898, 0
  %1899 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %1897, float noundef 0x3FC2073EC0000000) #139
  %1900 = call float @llvm.fma.f32(float %1897, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.0324 = select i1 %.not165, float %1900, float %1899
  %1901 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not166 = icmp eq i32 %1901, 0
  %1902 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0324, float %1897, float noundef 0xBFBF19B980000000) #139
  %1903 = call float @llvm.fma.f32(float %.0324, float %1897, float noundef 0xBFBF19B980000000)
  %.0320 = select i1 %.not166, float %1903, float %1902
  %1904 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not167 = icmp eq i32 %1904, 0
  %1905 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0320, float %1897, float noundef 0x3FC1E52AA0000000) #139
  %1906 = call float @llvm.fma.f32(float %.0320, float %1897, float noundef 0x3FC1E52AA0000000)
  %.0316 = select i1 %.not167, float %1906, float %1905
  %1907 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not168 = icmp eq i32 %1907, 0
  %1908 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0316, float %1897, float noundef 0xBFC55B1720000000) #139
  %1909 = call float @llvm.fma.f32(float %.0316, float %1897, float noundef 0xBFC55B1720000000)
  %.0315 = select i1 %.not168, float %1909, float %1908
  %1910 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not169 = icmp eq i32 %1910, 0
  %1911 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0315, float %1897, float noundef 0x3FC99DA160000000) #139
  %1912 = call float @llvm.fma.f32(float %.0315, float %1897, float noundef 0x3FC99DA160000000)
  %.0311 = select i1 %.not169, float %1912, float %1911
  %1913 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not170 = icmp eq i32 %1913, 0
  %1914 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0311, float %1897, float noundef 0xBFCFFFE440000000) #139
  %1915 = call float @llvm.fma.f32(float %.0311, float %1897, float noundef 0xBFCFFFE440000000)
  %.0307 = select i1 %.not170, float %1915, float %1914
  %1916 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not171 = icmp eq i32 %1916, 0
  %1917 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0307, float %1897, float noundef 0x3FD5554F00000000) #139
  %1918 = call float @llvm.fma.f32(float %.0307, float %1897, float noundef 0x3FD5554F00000000)
  %.0303 = select i1 %.not171, float %1918, float %1917
  %1919 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not172 = icmp eq i32 %1919, 0
  %1920 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0303, float %1897, float noundef -5.000000e-01) #139
  %1921 = call float @llvm.fma.f32(float %.0303, float %1897, float noundef -5.000000e-01)
  %.0332 = select i1 %.not172, float %1921, float %1920
  %1922 = fmul float %1897, %.0332
  %1923 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not173 = icmp eq i32 %1923, 0
  %1924 = call float @llvm.nvvm.fma.rn.ftz.f(float %1922, float %1897, float %1897) #139
  %1925 = call float @llvm.fma.f32(float %1922, float %1897, float %1897)
  %.0337 = select i1 %.not173, float %1925, float %1924
  %1926 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not174 = icmp eq i32 %1926, 0
  %1927 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0328, float noundef 0x3FE62E4300000000, float %.0337) #139
  %1928 = call float @llvm.fma.f32(float %.0328, float noundef 0x3FE62E4300000000, float %.0337)
  %.0343 = select i1 %.not174, float %1928, float %1927
  %1929 = icmp ugt i32 %.pre-phi, 2139095039
  br i1 %1929, label %__nv_fmaf_rn.exit.i.i558, label %__internal_accurate_logf.exit.i559

__nv_fmaf_rn.exit.i.i558:                         ; preds = %__nv_j1f.exit.i546
  %1930 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not177 = icmp eq i32 %1930, 0
  %1931 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0439, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %1932 = call float @llvm.fma.f32(float %.0439, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.0350 = select i1 %.not177, float %1932, float %1931
  br label %__internal_accurate_logf.exit.i559

__internal_accurate_logf.exit.i559:               ; preds = %__nv_fmaf_rn.exit.i.i558, %__nv_j1f.exit.i546
  %r.i5.i445.0 = phi float [ %.0350, %__nv_fmaf_rn.exit.i.i558 ], [ %.0343, %__nv_j1f.exit.i546 ]
  %1933 = fcmp oeq float %.0439, 0.000000e+00
  %r.i5.i445.1 = select i1 %1933, float 0xFFF0000000000000, float %r.i5.i445.0
  %1934 = fdiv float -1.000000e+00, %.0439
  %1935 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not175 = icmp eq i32 %1935, 0
  %1936 = call float @llvm.nvvm.fma.rn.ftz.f(float %r.i.i482.3, float %r.i5.i445.1, float %1934) #139
  %1937 = call float @llvm.fma.f32(float %r.i.i482.3, float %r.i5.i445.1, float %1934)
  %.0298 = select i1 %.not175, float %1937, float %1936
  %1938 = fmul float %.0298, 0x3FE45F3060000000
  %1939 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not176 = icmp eq i32 %1939, 0
  %1940 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0446, float %.0439, float %1938) #139
  %1941 = call float @llvm.fma.f32(float %.0446, float %.0439, float %1938)
  %.0292 = select i1 %.not176, float %1941, float %1940
  br label %__nv_y1f.exit618

1942:                                             ; preds = %1650
  %1943 = fcmp ugt float %.0439, 0x400E81DDC0000000
  br i1 %1943, label %1974, label %__internal_fmad.exit11.i562

__internal_fmad.exit11.i562:                      ; preds = %1942
  %1944 = fadd float %.0439, 0xC00193BEE0000000
  %1945 = fadd float %1944, 0x3E76401B80000000
  %1946 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not120 = icmp eq i32 %1946, 0
  %1947 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBEF13C53A0000000, float %1945, float noundef 0x3F20C95100000000) #139
  %1948 = call float @llvm.fma.f32(float %1945, float noundef 0xBEF13C53A0000000, float noundef 0x3F20C95100000000)
  %.0290 = select i1 %.not120, float %1948, float %1947
  %1949 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not121 = icmp eq i32 %1949, 0
  %1950 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0290, float %1945, float noundef 0xBF3E1561A0000000) #139
  %1951 = call float @llvm.fma.f32(float %.0290, float %1945, float noundef 0xBF3E1561A0000000)
  %.0283 = select i1 %.not121, float %1951, float %1950
  %1952 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not122 = icmp eq i32 %1952, 0
  %1953 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0283, float %1945, float noundef 0x3F51EC2040000000) #139
  %1954 = call float @llvm.fma.f32(float %.0283, float %1945, float noundef 0x3F51EC2040000000)
  %.0277 = select i1 %.not122, float %1954, float %1953
  %1955 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not123 = icmp eq i32 %1955, 0
  %1956 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0277, float %1945, float noundef 0xBF658E08A0000000) #139
  %1957 = call float @llvm.fma.f32(float %.0277, float %1945, float noundef 0xBF658E08A0000000)
  %.0276 = select i1 %.not123, float %1957, float %1956
  %1958 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not124 = icmp eq i32 %1958, 0
  %1959 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0276, float %1945, float noundef 0x3F7E6BBEE0000000) #139
  %1960 = call float @llvm.fma.f32(float %.0276, float %1945, float noundef 0x3F7E6BBEE0000000)
  %.0271 = select i1 %.not124, float %1960, float %1959
  %1961 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not125 = icmp eq i32 %1961, 0
  %1962 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0271, float %1945, float noundef 0xBF73A12F80000000) #139
  %1963 = call float @llvm.fma.f32(float %.0271, float %1945, float noundef 0xBF73A12F80000000)
  %.0267 = select i1 %.not125, float %1963, float %1962
  %1964 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not126 = icmp eq i32 %1964, 0
  %1965 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0267, float %1945, float noundef 0xBFA0D2D140000000) #139
  %1966 = call float @llvm.fma.f32(float %.0267, float %1945, float noundef 0xBFA0D2D140000000)
  %.0262 = select i1 %.not126, float %1966, float %1965
  %1967 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not127 = icmp eq i32 %1967, 0
  %1968 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0262, float %1945, float noundef 0xBFBE56FBE0000000) #139
  %1969 = call float @llvm.fma.f32(float %.0262, float %1945, float noundef 0xBFBE56FBE0000000)
  %.0255 = select i1 %.not127, float %1969, float %1968
  %1970 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not128 = icmp eq i32 %1970, 0
  %1971 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0255, float %1945, float noundef 0x3FE0AA4840000000) #139
  %1972 = call float @llvm.fma.f32(float %.0255, float %1945, float noundef 0x3FE0AA4840000000)
  %.0248 = select i1 %.not128, float %1972, float %1971
  %1973 = fmul float %1945, %.0248
  br label %__nv_y1f.exit618

1974:                                             ; preds = %1942
  %1975 = fcmp ugt float %.0439, 0x401C0D26E0000000
  br i1 %1975, label %2006, label %__internal_fmad.exit29.i571

__internal_fmad.exit29.i571:                      ; preds = %1974
  %1976 = fadd float %.0439, 0xC015B7FE40000000
  %1977 = fadd float %1976, 0xBE8D0F6060000000
  %1978 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not111 = icmp eq i32 %1978, 0
  %1979 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E57CAFA00000000, float %1977, float noundef 0xBEA11482C0000000) #139
  %1980 = call float @llvm.fma.f32(float %1977, float noundef 0x3E57CAFA00000000, float noundef 0xBEA11482C0000000)
  %.0241 = select i1 %.not111, float %1980, float %1979
  %1981 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not112 = icmp eq i32 %1981, 0
  %1982 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0241, float %1977, float noundef 0xBEC7F14280000000) #139
  %1983 = call float @llvm.fma.f32(float %.0241, float %1977, float noundef 0xBEC7F14280000000)
  %.0234 = select i1 %.not112, float %1983, float %1982
  %1984 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not113 = icmp eq i32 %1984, 0
  %1985 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0234, float %1977, float noundef 0x3F084FC040000000) #139
  %1986 = call float @llvm.fma.f32(float %.0234, float %1977, float noundef 0x3F084FC040000000)
  %.0227 = select i1 %.not113, float %1986, float %1985
  %1987 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not114 = icmp eq i32 %1987, 0
  %1988 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0227, float %1977, float noundef 0x3F23376380000000) #139
  %1989 = call float @llvm.fma.f32(float %.0227, float %1977, float noundef 0x3F23376380000000)
  %.0220 = select i1 %.not114, float %1989, float %1988
  %1990 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not115 = icmp eq i32 %1990, 0
  %1991 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0220, float %1977, float noundef 0xBF61BE3FA0000000) #139
  %1992 = call float @llvm.fma.f32(float %.0220, float %1977, float noundef 0xBF61BE3FA0000000)
  %.0213 = select i1 %.not115, float %1992, float %1991
  %1993 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not116 = icmp eq i32 %1993, 0
  %1994 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0213, float %1977, float noundef 0xBF710A3120000000) #139
  %1995 = call float @llvm.fma.f32(float %.0213, float %1977, float noundef 0xBF710A3120000000)
  %.0208 = select i1 %.not116, float %1995, float %1994
  %1996 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not117 = icmp eq i32 %1996, 0
  %1997 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0208, float %1977, float noundef 0x3FAA15D820000000) #139
  %1998 = call float @llvm.fma.f32(float %.0208, float %1977, float noundef 0x3FAA15D820000000)
  %.0207 = select i1 %.not117, float %1998, float %1997
  %1999 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not118 = icmp eq i32 %1999, 0
  %2000 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0207, float %1977, float noundef 0x3FA00B9F80000000) #139
  %2001 = call float @llvm.fma.f32(float %.0207, float %1977, float noundef 0x3FA00B9F80000000)
  %.0201 = select i1 %.not118, float %2001, float %2000
  %2002 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not119 = icmp eq i32 %2002, 0
  %2003 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0201, float %1977, float noundef 0xBFD5C7C560000000) #139
  %2004 = call float @llvm.fma.f32(float %.0201, float %1977, float noundef 0xBFD5C7C560000000)
  %.0198 = select i1 %.not119, float %2004, float %2003
  %2005 = fmul float %1977, %.0198
  br label %__nv_y1f.exit618

2006:                                             ; preds = %1974
  %2007 = fcmp ugt float %.0439, 0x4024585C60000000
  br i1 %2007, label %__nv_isinff.exit.i588, label %__internal_fmad.exit47.i580

__internal_fmad.exit47.i580:                      ; preds = %2006
  %2008 = fadd float %.0439, 0xC0213127A0000000
  %2009 = fadd float %2008, 0xBE9CC2D360000000
  %2010 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not103 = icmp eq i32 %2010, 0
  %2011 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EA19E7060000000, float %2009, float noundef 0x3EC05205C0000000) #139
  %2012 = call float @llvm.fma.f32(float %2009, float noundef 0x3EA19E7060000000, float noundef 0x3EC05205C0000000)
  %.0192 = select i1 %.not103, float %2012, float %2011
  %2013 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not104 = icmp eq i32 %2013, 0
  %2014 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0192, float %2009, float noundef 0xBF06EBEE20000000) #139
  %2015 = call float @llvm.fma.f32(float %.0192, float %2009, float noundef 0xBF06EBEE20000000)
  %.0185 = select i1 %.not104, float %2015, float %2014
  %2016 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not105 = icmp eq i32 %2016, 0
  %2017 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0185, float %2009, float noundef 0xBF1B3F5500000000) #139
  %2018 = call float @llvm.fma.f32(float %.0185, float %2009, float noundef 0xBF1B3F5500000000)
  %.0178 = select i1 %.not105, float %2018, float %2017
  %2019 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not106 = icmp eq i32 %2019, 0
  %2020 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0178, float %2009, float noundef 0x3F607A3340000000) #139
  %2021 = call float @llvm.fma.f32(float %.0178, float %2009, float noundef 0x3F607A3340000000)
  %.0171 = select i1 %.not106, float %2021, float %2020
  %2022 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not107 = icmp eq i32 %2022, 0
  %2023 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0171, float %2009, float noundef 0x3F63CE6DA0000000) #139
  %2024 = call float @llvm.fma.f32(float %.0171, float %2009, float noundef 0x3F63CE6DA0000000)
  %.0167 = select i1 %.not107, float %2024, float %2023
  %2025 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not108 = icmp eq i32 %2025, 0
  %2026 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0167, float %2009, float noundef 0xBFA6395CA0000000) #139
  %2027 = call float @llvm.fma.f32(float %.0167, float %2009, float noundef 0xBFA6395CA0000000)
  %.0160 = select i1 %.not108, float %2027, float %2026
  %2028 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not109 = icmp eq i32 %2028, 0
  %2029 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0160, float %2009, float noundef 0xBF902B36C0000000) #139
  %2030 = call float @llvm.fma.f32(float %.0160, float %2009, float noundef 0xBF902B36C0000000)
  %.0154 = select i1 %.not109, float %2030, float %2029
  %2031 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not110 = icmp eq i32 %2031, 0
  %2032 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0154, float %2009, float noundef 0x3FD15F9940000000) #139
  %2033 = call float @llvm.fma.f32(float %.0154, float %2009, float noundef 0x3FD15F9940000000)
  %.0149 = select i1 %.not110, float %2033, float %2032
  %2034 = fmul float %2009, %.0149
  br label %__nv_y1f.exit618

__nv_isinff.exit.i588:                            ; preds = %2006
  %2035 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not66 = icmp eq i32 %2035, 0
  %2036 = call float @llvm.nvvm.fabs.ftz.f(float %.0439) #139
  %2037 = call float @llvm.fabs.f32(float %.0439)
  %.0146 = select i1 %.not66, float %2037, float %2036
  %2038 = fcmp oeq float %.0146, 0x7FF0000000000000
  br i1 %2038, label %__nv_y1f.exit618, label %__internal_fmad.exit66.i589

__internal_fmad.exit66.i589:                      ; preds = %__nv_isinff.exit.i588
  %2039 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0439) #141, !srcloc !73
  %2040 = fmul float %2039, %2039
  %2041 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not67 = icmp eq i32 %2041, 0
  %2042 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FE4CFEC00000000, float %2040, float noundef 0xBFC8957200000000) #139
  %2043 = call float @llvm.fma.f32(float %2040, float noundef 0x3FE4CFEC00000000, float noundef 0xBFC8957200000000)
  %.0140 = select i1 %.not67, float %2043, float %2042
  %2044 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not68 = icmp eq i32 %2044, 0
  %2045 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0140, float %2040, float noundef 0x3FC7FFD7E0000000) #139
  %2046 = call float @llvm.fma.f32(float %.0140, float %2040, float noundef 0x3FC7FFD7E0000000)
  %.0134 = select i1 %.not68, float %2046, float %2045
  %2047 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not69 = icmp eq i32 %2047, 0
  %2048 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0134, float %2040, float noundef 1.000000e+00) #139
  %2049 = call float @llvm.fma.f32(float %.0134, float %2040, float noundef 1.000000e+00)
  %.0132 = select i1 %.not69, float %2049, float %2048
  %2050 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not70 = icmp eq i32 %2050, 0
  %2051 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFFC9C3560000000, float %2040, float noundef 0x3FD76E7560000000) #139
  %2052 = call float @llvm.fma.f32(float %2040, float noundef 0xBFFC9C3560000000, float noundef 0x3FD76E7560000000)
  %.0128 = select i1 %.not70, float %2052, float %2051
  %2053 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not71 = icmp eq i32 %2053, 0
  %2054 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0128, float %2040, float noundef 0xBFC4FF6DC0000000) #139
  %2055 = call float @llvm.fma.f32(float %.0128, float %2040, float noundef 0xBFC4FF6DC0000000)
  %.0123 = select i1 %.not71, float %2055, float %2054
  %2056 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not72 = icmp eq i32 %2056, 0
  %2057 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0123, float %2040, float noundef 0x3FD7FFFFE0000000) #139
  %2058 = call float @llvm.fma.f32(float %.0123, float %2040, float noundef 0x3FD7FFFFE0000000)
  %.0119 = select i1 %.not72, float %2058, float %2057
  %2059 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not73 = icmp eq i32 %2059, 0
  %2060 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0119, float %2039, float %.0439) #139
  %2061 = call float @llvm.fma.f32(float %.0119, float %2039, float %.0439)
  %.0112 = select i1 %.not73, float %2061, float %2060
  %2062 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not74 = icmp eq i32 %2062, 0
  br i1 %.not74, label %2065, label %2063

2063:                                             ; preds = %__internal_fmad.exit66.i589
  %2064 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0439) #139
  br label %__nv_rsqrtf.exit.i596

2065:                                             ; preds = %__internal_fmad.exit66.i589
  %2066 = call float @llvm.nvvm.rsqrt.approx.f(float %.0439) #139
  br label %__nv_rsqrtf.exit.i596

__nv_rsqrtf.exit.i596:                            ; preds = %2065, %2063
  %.0110 = phi float [ %2064, %2063 ], [ %2066, %2065 ]
  %2067 = fmul float %.0110, 0x3FE9884540000000
  %2068 = fmul float %.0132, %2067
  %2069 = fmul float %.0112, 0x3FE45F3060000000
  %2070 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not75 = icmp eq i32 %2070, 0
  %2071 = call i32 @llvm.nvvm.f2i.rn.ftz(float %2069) #139
  %2072 = call i32 @llvm.nvvm.f2i.rn(float %2069) #139
  %.0103 = select i1 %.not75, i32 %2072, i32 %2071
  %2073 = sitofp i32 %.0103 to float
  %2074 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not76 = icmp eq i32 %2074, 0
  %2075 = call float @llvm.nvvm.fma.rn.ftz.f(float %2073, float noundef 0xBFF921FB40000000, float %.0112) #139
  %2076 = call float @llvm.fma.f32(float %2073, float noundef 0xBFF921FB40000000, float %.0112)
  %.099 = select i1 %.not76, float %2076, float %2075
  %2077 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not77 = icmp eq i32 %2077, 0
  %2078 = call float @llvm.nvvm.fma.rn.ftz.f(float %2073, float noundef 0xBE74442D00000000, float %.099) #139
  %2079 = call float @llvm.fma.f32(float %2073, float noundef 0xBE74442D00000000, float %.099)
  %.095 = select i1 %.not77, float %2079, float %2078
  %2080 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not78 = icmp eq i32 %2080, 0
  %2081 = call float @llvm.nvvm.fma.rn.ftz.f(float %2073, float noundef 0xBCF84698A0000000, float %.095) #139
  %2082 = call float @llvm.fma.f32(float %2073, float noundef 0xBCF84698A0000000, float %.095)
  %.091 = select i1 %.not78, float %2082, float %2081
  %2083 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not79 = icmp eq i32 %2083, 0
  %2084 = call float @llvm.nvvm.fabs.ftz.f(float %.0112) #139
  %2085 = call float @llvm.fabs.f32(float %.0112)
  %.089 = select i1 %.not79, float %2085, float %2084
  %2086 = fcmp ult float %.089, 1.056150e+05
  br i1 %2086, label %__internal_trig_reduction_kernel.exit.i.i605, label %__nv_isinff.exit.i.i.i602

__nv_isinff.exit.i.i.i602:                        ; preds = %__nv_rsqrtf.exit.i596
  %2087 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not96 = icmp eq i32 %2087, 0
  %.086 = select i1 %.not96, float %2085, float %2084
  %2088 = fcmp oeq float %.086, 0x7FF0000000000000
  br i1 %2088, label %__nv_fmul_rn.exit.i.i.i603, label %2092

__nv_fmul_rn.exit.i.i.i603:                       ; preds = %__nv_isinff.exit.i.i.i602
  %2089 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not102 = icmp eq i32 %2089, 0
  %2090 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0112, float noundef 0.000000e+00) #139
  %2091 = fmul float %.0112, 0.000000e+00
  %.081 = select i1 %.not102, float %2091, float %2090
  br label %__internal_trig_reduction_kernel.exit.i.i605

2092:                                             ; preds = %__nv_isinff.exit.i.i.i602
  %2093 = bitcast float %.0112 to i32
  %2094 = lshr i32 %2093, 23
  %2095 = and i32 %2094, 255
  %2096 = add nsw i32 %2095, -128
  %2097 = shl i32 %2093, 8
  %2098 = or i32 %2097, -2147483648
  %2099 = lshr i32 %2096, 5
  br label %2100

2100:                                             ; preds = %2092, %2100
  %hi.i.i.i.i417.0561 = phi i32 [ 0, %2092 ], [ %2106, %2100 ]
  %iq.i.i.i.i422.0560 = phi i32 [ 0, %2092 ], [ %2108, %2100 ]
  %2101 = zext i32 %iq.i.i.i.i422.0560 to i64
  %2102 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %2101
  %2103 = load i32, i32 addrspace(1)* %2102, align 4
  %2104 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %2103, i32 %2098, i32 %hi.i.i.i.i417.0561) #141, !srcloc !67
  %2105 = extractvalue { i32, i32 } %2104, 0
  %2106 = extractvalue { i32, i32 } %2104, 1
  %2107 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %2101
  store i32 %2105, i32* %2107, align 4, !noalias !68
  %2108 = add nuw nsw i32 %iq.i.i.i.i422.0560, 1
  %exitcond596.not = icmp eq i32 %2108, 6
  br i1 %exitcond596.not, label %2109, label %2100, !llvm.loop !71

2109:                                             ; preds = %2100
  %2110 = and i32 %2093, -2147483648
  %2111 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 6
  store i32 %2106, i32* %2111, align 4, !noalias !68
  %2112 = and i32 %2094, 31
  %2113 = sub nsw i32 6, %2099
  %2114 = sext i32 %2113 to i64
  %2115 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %2114
  %2116 = load i32, i32* %2115, align 4
  %2117 = sub nsw i32 5, %2099
  %2118 = sext i32 %2117 to i64
  %2119 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %2118
  %2120 = load i32, i32* %2119, align 4
  %.not97 = icmp eq i32 %2112, 0
  br i1 %.not97, label %__internal_trig_reduction_slowpath.exit.i.i.i604, label %2121

2121:                                             ; preds = %2109
  %2122 = sub nsw i32 4, %2099
  %2123 = sub nuw nsw i32 32, %2112
  %2124 = shl i32 %2116, %2112
  %2125 = lshr i32 %2120, %2123
  %2126 = add i32 %2125, %2124
  %2127 = shl i32 %2120, %2112
  %2128 = sext i32 %2122 to i64
  %2129 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i.i, i64 0, i64 %2128
  %2130 = load i32, i32* %2129, align 4
  %2131 = lshr i32 %2130, %2123
  %2132 = add i32 %2131, %2127
  br label %__internal_trig_reduction_slowpath.exit.i.i.i604

__internal_trig_reduction_slowpath.exit.i.i.i604: ; preds = %2121, %2109
  %lo.i.i.i.i418.0 = phi i32 [ %2132, %2121 ], [ %2120, %2109 ]
  %hi.i.i.i.i417.1 = phi i32 [ %2126, %2121 ], [ %2116, %2109 ]
  %2133 = lshr i32 %hi.i.i.i.i417.1, 30
  %2134 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i417.1, i32 %lo.i.i.i.i418.0, i32 noundef 2)
  %2135 = shl i32 %lo.i.i.i.i418.0, 2
  %2136 = lshr i32 %2134, 31
  %2137 = add nuw nsw i32 %2136, %2133
  %.not98 = icmp eq i32 %2110, 0
  %2138 = sub nsw i32 0, %2137
  %spec.select7 = select i1 %.not98, i32 %2137, i32 %2138
  %.not99 = icmp sgt i32 %2134, -1
  %2139 = xor i32 %2110, -2147483648
  %not..not99 = xor i1 %.not99, true
  %2140 = sext i1 %not..not99 to i32
  %lo.i.i.i.i418.1 = xor i32 %2135, %2140
  %hi.i.i.i.i417.2 = xor i32 %2134, %2140
  %s.i.i.i.i416.0 = select i1 %.not99, i32 %2110, i32 %2139
  %2141 = zext i32 %hi.i.i.i.i417.2 to i64
  %2142 = shl nuw i64 %2141, 32
  %2143 = zext i32 %lo.i.i.i.i418.1 to i64
  %2144 = or i64 %2142, %2143
  %2145 = sitofp i64 %2144 to double
  %2146 = fmul double %2145, 0x3BF921FB54442D19
  %2147 = fptrunc double %2146 to float
  %.not101 = icmp eq i32 %s.i.i.i.i416.0, 0
  %2148 = fneg float %2147
  %r.i.i.i.i426.0 = select i1 %.not101, float %2147, float %2148
  br label %__internal_trig_reduction_kernel.exit.i.i605

__internal_trig_reduction_kernel.exit.i.i605:     ; preds = %__nv_fmul_rn.exit.i.i.i603, %__internal_trig_reduction_slowpath.exit.i.i.i604, %__nv_rsqrtf.exit.i596
  %i.i81.i432.1 = phi i32 [ %.0103, %__nv_rsqrtf.exit.i596 ], [ 0, %__nv_fmul_rn.exit.i.i.i603 ], [ %spec.select7, %__internal_trig_reduction_slowpath.exit.i.i.i604 ]
  %t.i.i.i429.1 = phi float [ %.091, %__nv_rsqrtf.exit.i596 ], [ %.081, %__nv_fmul_rn.exit.i.i.i603 ], [ %r.i.i.i.i426.0, %__internal_trig_reduction_slowpath.exit.i.i.i604 ]
  %2149 = and i32 %i.i81.i432.1, 3
  %2150 = uitofp i32 %2149 to float
  %2151 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not80 = icmp eq i32 %2151, 0
  %2152 = call float @llvm.nvvm.fma.rn.ftz.f(float %2150, float noundef 0x3FF921FB60000000, float noundef 0xC00F6A7A20000000) #139
  %2153 = call float @llvm.fma.f32(float %2150, float noundef 0x3FF921FB60000000, float noundef 0xC00F6A7A20000000)
  %.060 = select i1 %.not80, float %2153, float %2152
  %2154 = fadd float %t.i.i.i429.1, %.060
  %2155 = fmul float %2154, 0x3FE45F3060000000
  %2156 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not81 = icmp eq i32 %2156, 0
  %2157 = call i32 @llvm.nvvm.f2i.rn.ftz(float %2155) #139
  %2158 = call i32 @llvm.nvvm.f2i.rn(float %2155) #139
  %.058 = select i1 %.not81, i32 %2158, i32 %2157
  %2159 = sitofp i32 %.058 to float
  %2160 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not82 = icmp eq i32 %2160, 0
  %2161 = call float @llvm.nvvm.fma.rn.ftz.f(float %2159, float noundef 0xBFF921FB40000000, float %2154) #139
  %2162 = call float @llvm.fma.f32(float %2159, float noundef 0xBFF921FB40000000, float %2154)
  %.054 = select i1 %.not82, float %2162, float %2161
  %2163 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not83 = icmp eq i32 %2163, 0
  %2164 = call float @llvm.nvvm.fma.rn.ftz.f(float %2159, float noundef 0xBE74442D00000000, float %.054) #139
  %2165 = call float @llvm.fma.f32(float %2159, float noundef 0xBE74442D00000000, float %.054)
  %.051 = select i1 %.not83, float %2165, float %2164
  %2166 = add i32 %.058, 1
  %2167 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not84 = icmp eq i32 %2167, 0
  %2168 = call float @llvm.nvvm.mul.rn.ftz.f(float %.051, float %.051) #139
  %2169 = fmul float %.051, %.051
  %.047 = select i1 %.not84, float %2169, float %2168
  %2170 = and i32 %2166, 1
  %.not85 = icmp eq i32 %2170, 0
  %2171 = select i1 %.not85, float %.051, float 1.000000e+00
  %2172 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not86 = icmp eq i32 %2172, 0
  %2173 = call float @llvm.nvvm.fma.rn.ftz.f(float %.047, float %2171, float noundef 0.000000e+00) #139
  %2174 = call float @llvm.fma.f32(float %.047, float %2171, float noundef 0.000000e+00)
  %.043 = select i1 %.not86, float %2174, float %2173
  br i1 %.not85, label %__internal_fmad.exit2.i.i.i613, label %__internal_fmad.exit1.i.i.i612

__internal_fmad.exit1.i.i.i612:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i605
  %2175 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not95 = icmp eq i32 %2175, 0
  %2176 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.047, float noundef 0xBF56C0FDA0000000) #139
  %2177 = call float @llvm.fma.f32(float %.047, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.038 = select i1 %.not95, float %2177, float %2176
  br label %__internal_fmad.exit2.i.i.i613

__internal_fmad.exit2.i.i.i613:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i605, %__internal_fmad.exit1.i.i.i612
  %2178 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i605 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i612 ]
  %2179 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i605 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i612 ]
  %2180 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i605 ], [ %.038, %__internal_fmad.exit1.i.i.i612 ]
  %2181 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not89 = icmp eq i32 %2181, 0
  %2182 = call float @llvm.nvvm.fma.rn.ftz.f(float %2180, float %.047, float noundef %2179) #139
  %2183 = call float @llvm.fma.f32(float %2180, float %.047, float noundef %2179)
  %.034 = select i1 %.not89, float %2183, float %2182
  %2184 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not91 = icmp eq i32 %2184, 0
  %2185 = call float @llvm.nvvm.fma.rn.ftz.f(float %.034, float %.047, float noundef %2178) #139
  %2186 = call float @llvm.fma.f32(float %.034, float %.047, float noundef %2178)
  %.027 = select i1 %.not91, float %2186, float %2185
  %2187 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not92 = icmp eq i32 %2187, 0
  %2188 = call float @llvm.nvvm.fma.rn.ftz.f(float %.027, float %.043, float %2171) #139
  %2189 = call float @llvm.fma.f32(float %.027, float %.043, float %2171)
  %.020 = select i1 %.not92, float %2189, float %2188
  %2190 = and i32 %2166, 2
  %.not93 = icmp eq i32 %2190, 0
  br i1 %.not93, label %__internal_cos_offset_f.exit.i617, label %__internal_fmad.exit5.i.i.i616

__internal_fmad.exit5.i.i.i616:                   ; preds = %__internal_fmad.exit2.i.i.i613
  %2191 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not94 = icmp eq i32 %2191, 0
  %2192 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %2193 = call float @llvm.fma.f32(float %.020, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.013 = select i1 %.not94, float %2193, float %2192
  br label %__internal_cos_offset_f.exit.i617

__internal_cos_offset_f.exit.i617:                ; preds = %__internal_fmad.exit5.i.i.i616, %__internal_fmad.exit2.i.i.i613
  %z.i.i79.i409.0 = phi float [ %.013, %__internal_fmad.exit5.i.i.i616 ], [ %.020, %__internal_fmad.exit2.i.i.i613 ]
  %2194 = fmul float %2068, %z.i.i79.i409.0
  br label %__nv_y1f.exit618

__nv_y1f.exit618:                                 ; preds = %__internal_accurate_logf.exit.i559, %__internal_fmad.exit29.i571, %__internal_cos_offset_f.exit.i617, %__nv_isinff.exit.i588, %__internal_fmad.exit47.i580, %__internal_fmad.exit11.i562, %1648
  %r.i489.5 = phi float [ %1649, %1648 ], [ %.0292, %__internal_accurate_logf.exit.i559 ], [ %1973, %__internal_fmad.exit11.i562 ], [ %2005, %__internal_fmad.exit29.i571 ], [ %2034, %__internal_fmad.exit47.i580 ], [ %2194, %__internal_cos_offset_f.exit.i617 ], [ 0.000000e+00, %__nv_isinff.exit.i588 ]
  %r.i489.6 = select i1 %551, float 0x7FFFFFFFE0000000, float %r.i489.5
  br label %2229

2195:                                             ; preds = %__nv_y0f.exit
  %2196 = icmp slt i32 %n, 0
  %2197 = fcmp ult float %a, 0.000000e+00
  %or.cond = select i1 %2196, i1 true, i1 %2197
  br i1 %or.cond, label %2229, label %__nv_fabsf.exit

__nv_fabsf.exit:                                  ; preds = %2195
  %2198 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not336 = icmp eq i32 %2198, 0
  %.08 = select i1 %.not336, float %2, float %1
  %2199 = fcmp olt float %.08, 0x3810000000000000
  br i1 %2199, label %__nv_fabsf.exit620, label %__internal_fmad.exit.lr.ph

__nv_fabsf.exit620:                               ; preds = %__nv_fabsf.exit
  %2200 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not339 = icmp eq i32 %2200, 0
  %.06 = select i1 %.not339, float %2, float %1
  %2201 = fdiv float -1.000000e+09, %.06
  br label %2229

__internal_fmad.exit.lr.ph:                       ; preds = %__nv_fabsf.exit
  %2202 = fdiv float 2.000000e+00, %a
  %2203 = add nsw i32 %n, -1
  %xtraiter = and i32 %2203, 1
  %2204 = icmp eq i32 %n, 2
  br i1 %2204, label %.__nv_isnanf.exit_crit_edge.unr-lcssa, label %__internal_fmad.exit.lr.ph.new

__internal_fmad.exit.lr.ph.new:                   ; preds = %__internal_fmad.exit.lr.ph
  %unroll_iter = and i32 %2203, -2
  br label %__internal_fmad.exit

__internal_fmad.exit:                             ; preds = %__internal_fmad.exit, %__internal_fmad.exit.lr.ph.new
  %i.0555 = phi i32 [ 1, %__internal_fmad.exit.lr.ph.new ], [ %2218, %__internal_fmad.exit ]
  %yim1.0554 = phi float [ %r.i75.5, %__internal_fmad.exit.lr.ph.new ], [ %.02, %__internal_fmad.exit ]
  %yi.0553 = phi float [ %r.i.6, %__internal_fmad.exit.lr.ph.new ], [ %.02.1, %__internal_fmad.exit ]
  %niter = phi i32 [ 0, %__internal_fmad.exit.lr.ph.new ], [ %niter.next.1, %__internal_fmad.exit ]
  %2205 = sitofp i32 %i.0555 to float
  %2206 = fmul float %2202, %2205
  %2207 = fneg float %yim1.0554
  %2208 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not338 = icmp eq i32 %2208, 0
  %2209 = call float @llvm.nvvm.fma.rn.ftz.f(float %2206, float %yi.0553, float %2207) #139
  %2210 = call float @llvm.fma.f32(float %2206, float %yi.0553, float %2207)
  %.02 = select i1 %.not338, float %2210, float %2209
  %2211 = add nuw nsw i32 %i.0555, 1
  %2212 = sitofp i32 %2211 to float
  %2213 = fmul float %2202, %2212
  %2214 = fneg float %yi.0553
  %2215 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not338.1 = icmp eq i32 %2215, 0
  %2216 = call float @llvm.nvvm.fma.rn.ftz.f(float %2213, float %.02, float %2214) #139
  %2217 = call float @llvm.fma.f32(float %2213, float %.02, float %2214)
  %.02.1 = select i1 %.not338.1, float %2217, float %2216
  %2218 = add nuw nsw i32 %i.0555, 2
  %niter.next.1 = add i32 %niter, 2
  %niter.ncmp.1.not = icmp eq i32 %niter.next.1, %unroll_iter
  br i1 %niter.ncmp.1.not, label %.__nv_isnanf.exit_crit_edge.unr-lcssa.loopexit, label %__internal_fmad.exit

.__nv_isnanf.exit_crit_edge.unr-lcssa.loopexit:   ; preds = %__internal_fmad.exit
  %phi.cast = sitofp i32 %2218 to float
  br label %.__nv_isnanf.exit_crit_edge.unr-lcssa

.__nv_isnanf.exit_crit_edge.unr-lcssa:            ; preds = %.__nv_isnanf.exit_crit_edge.unr-lcssa.loopexit, %__internal_fmad.exit.lr.ph
  %.02.lcssa.ph = phi float [ undef, %__internal_fmad.exit.lr.ph ], [ %.02.1, %.__nv_isnanf.exit_crit_edge.unr-lcssa.loopexit ]
  %i.0555.unr = phi float [ 1.000000e+00, %__internal_fmad.exit.lr.ph ], [ %phi.cast, %.__nv_isnanf.exit_crit_edge.unr-lcssa.loopexit ]
  %yim1.0554.unr = phi float [ %r.i75.5, %__internal_fmad.exit.lr.ph ], [ %.02, %.__nv_isnanf.exit_crit_edge.unr-lcssa.loopexit ]
  %yi.0553.unr = phi float [ %r.i.6, %__internal_fmad.exit.lr.ph ], [ %.02.1, %.__nv_isnanf.exit_crit_edge.unr-lcssa.loopexit ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %__nv_isnanf.exit, label %__internal_fmad.exit.epil.preheader

__internal_fmad.exit.epil.preheader:              ; preds = %.__nv_isnanf.exit_crit_edge.unr-lcssa
  %2219 = fmul float %2202, %i.0555.unr
  %2220 = fneg float %yim1.0554.unr
  %2221 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not338.epil = icmp eq i32 %2221, 0
  %2222 = call float @llvm.nvvm.fma.rn.ftz.f(float %2219, float %yi.0553.unr, float %2220) #139
  %2223 = call float @llvm.fma.f32(float %2219, float %yi.0553.unr, float %2220)
  %.02.epil = select i1 %.not338.epil, float %2223, float %2222
  br label %__nv_isnanf.exit

__nv_isnanf.exit:                                 ; preds = %__internal_fmad.exit.epil.preheader, %.__nv_isnanf.exit_crit_edge.unr-lcssa
  %yip1.0.lcssa = phi float [ %.02.lcssa.ph, %.__nv_isnanf.exit_crit_edge.unr-lcssa ], [ %.02.epil, %__internal_fmad.exit.epil.preheader ]
  %2224 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not337 = icmp eq i32 %2224, 0
  %2225 = call float @llvm.nvvm.fabs.ftz.f(float %yip1.0.lcssa) #139
  %2226 = call float @llvm.fabs.f32(float %yip1.0.lcssa)
  %.0 = select i1 %.not337, float %2226, float %2225
  %2227 = fcmp ugt float %.0, 0x7FF0000000000000
  br i1 %2227, label %2228, label %2229

2228:                                             ; preds = %__nv_isnanf.exit
  br label %2229

2229:                                             ; preds = %2195, %__nv_isnanf.exit, %2228, %__nv_fabsf.exit620, %__nv_y1f.exit618, %__nv_y0f.exit406
  %.01 = phi float [ %r.i274.5, %__nv_y0f.exit406 ], [ %r.i489.6, %__nv_y1f.exit618 ], [ %2201, %__nv_fabsf.exit620 ], [ 0xFFF0000000000000, %2228 ], [ %yip1.0.lcssa, %__nv_isnanf.exit ], [ 0x7FFFFFFFE0000000, %2195 ]
  ret float %.01
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_jnf(i32 %n, float %a) local_unnamed_addr #48 {
  %result.i.i.i.i = alloca [7 x i32], align 4
  switch i32 %n, label %437 [
    i32 0, label %__nv_fabsf.exit.i
    i32 1, label %__nv_fabsf.exit.i40
  ]

__nv_fabsf.exit.i:                                ; preds = %0
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not55 = icmp eq i32 %1, 0
  %2 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %3 = call float @llvm.fabs.f32(float %a)
  %.05 = select i1 %.not55, float %3, float %2
  %4 = fcmp ugt float %.05, 8.000000e+00
  br i1 %4, label %__nv_isinff.exit.i, label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %__nv_fabsf.exit.i
  %5 = fadd float %.05, 0xC0033D1520000000
  %6 = fadd float %5, 0xBE7D2E3660000000
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not94 = icmp eq i32 %7, 0
  %8 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBCD6771CE0000000, float %6, float noundef 0xBD35953660000000) #139
  %9 = call float @llvm.fma.f32(float %6, float noundef 0xBCD6771CE0000000, float noundef 0xBD35953660000000)
  %.06 = select i1 %.not94, float %9, float %8
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not95 = icmp eq i32 %10, 0
  %11 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %6, float noundef 0x3D87E1C300000000) #139
  %12 = call float @llvm.fma.f32(float %.06, float %6, float noundef 0x3D87E1C300000000)
  %.07 = select i1 %.not95, float %12, float %11
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not96 = icmp eq i32 %13, 0
  %14 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %6, float noundef 0xBD9A82F020000000) #139
  %15 = call float @llvm.fma.f32(float %.07, float %6, float noundef 0xBD9A82F020000000)
  %.09 = select i1 %.not96, float %15, float %14
  %16 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not97 = icmp eq i32 %16, 0
  %17 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %6, float noundef 0xBDFD21E700000000) #139
  %18 = call float @llvm.fma.f32(float %.09, float %6, float noundef 0xBDFD21E700000000)
  %.010 = select i1 %.not97, float %18, float %17
  %19 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not98 = icmp eq i32 %19, 0
  %20 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %6, float noundef 0x3E04060B60000000) #139
  %21 = call float @llvm.fma.f32(float %.010, float %6, float noundef 0x3E04060B60000000)
  %.011 = select i1 %.not98, float %21, float %20
  %22 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not99 = icmp eq i32 %22, 0
  %23 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %6, float noundef 0x3E6F2E2860000000) #139
  %24 = call float @llvm.fma.f32(float %.011, float %6, float noundef 0x3E6F2E2860000000)
  %.012 = select i1 %.not99, float %24, float %23
  %25 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not100 = icmp eq i32 %25, 0
  %26 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %6, float noundef 0x3E1EEDF0A0000000) #139
  %27 = call float @llvm.fma.f32(float %.012, float %6, float noundef 0x3E1EEDF0A0000000)
  %.013 = select i1 %.not100, float %27, float %26
  %28 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not101 = icmp eq i32 %28, 0
  %29 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %6, float noundef 0xBED6DBF8C0000000) #139
  %30 = call float @llvm.fma.f32(float %.013, float %6, float noundef 0xBED6DBF8C0000000)
  %.014 = select i1 %.not101, float %30, float %29
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not102 = icmp eq i32 %31, 0
  %32 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %6, float noundef 0xBEDECCB920000000) #139
  %33 = call float @llvm.fma.f32(float %.014, float %6, float noundef 0xBEDECCB920000000)
  %.016 = select i1 %.not102, float %33, float %32
  %34 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not103 = icmp eq i32 %34, 0
  %35 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %6, float noundef 0x3F33C5BD60000000) #139
  %36 = call float @llvm.fma.f32(float %.016, float %6, float noundef 0x3F33C5BD60000000)
  %.017 = select i1 %.not103, float %36, float %35
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not104 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %6, float noundef 0x3F495C6680000000) #139
  %39 = call float @llvm.fma.f32(float %.017, float %6, float noundef 0x3F495C6680000000)
  %.018 = select i1 %.not104, float %39, float %38
  %40 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not105 = icmp eq i32 %40, 0
  %41 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %6, float noundef 0xBF7DD54360000000) #139
  %42 = call float @llvm.fma.f32(float %.018, float %6, float noundef 0xBF7DD54360000000)
  %.019 = select i1 %.not105, float %42, float %41
  %43 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not106 = icmp eq i32 %43, 0
  %44 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %6, float noundef 0xBF9B4EE8E0000000) #139
  %45 = call float @llvm.fma.f32(float %.019, float %6, float noundef 0xBF9B4EE8E0000000)
  %.020 = select i1 %.not106, float %45, float %44
  %46 = fmul float %6, %.020
  %47 = fadd float %.05, 0xC016148F60000000
  %48 = fadd float %47, 0x3E734F46E0000000
  %49 = fmul float %48, %46
  %50 = fadd float %.05, 0xC0214EB560000000
  %51 = fadd float %50, 0xBE9999BDA0000000
  %52 = fmul float %51, %49
  br label %__nv_j0f.exit

__nv_isinff.exit.i:                               ; preds = %__nv_fabsf.exit.i
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not56 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fabs.ftz.f(float %.05) #139
  %55 = call float @llvm.fabs.f32(float %.05)
  %.021 = select i1 %.not56, float %55, float %54
  %56 = fcmp oeq float %.021, 0x7FF0000000000000
  br i1 %56, label %__nv_j0f.exit, label %__internal_fmad.exit13.i

__internal_fmad.exit13.i:                         ; preds = %__nv_isinff.exit.i
  %57 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.05) #141, !srcloc !73
  %58 = fmul float %57, %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not57 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400ADFD260000000, float %58, float noundef 0xBFE076F840000000) #139
  %61 = call float @llvm.fma.f32(float %58, float noundef 0x400ADFD260000000, float noundef 0xBFE076F840000000)
  %.023 = select i1 %.not57, float %61, float %60
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not58 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float %58, float noundef 0x3FBA767E60000000) #139
  %64 = call float @llvm.fma.f32(float %.023, float %58, float noundef 0x3FBA767E60000000)
  %.024 = select i1 %.not58, float %64, float %63
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not59 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %.024, float %58, float noundef 0xBFAFFFF6C0000000) #139
  %67 = call float @llvm.fma.f32(float %.024, float %58, float noundef 0xBFAFFFF6C0000000)
  %.025 = select i1 %.not59, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not60 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %.025, float %58, float noundef 1.000000e+00) #139
  %70 = call float @llvm.fma.f32(float %.025, float %58, float noundef 1.000000e+00)
  %.026 = select i1 %.not60, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not61 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF23C0120000000, float %58, float noundef 0xBFCA4825A0000000) #139
  %73 = call float @llvm.fma.f32(float %58, float noundef 0x3FF23C0120000000, float noundef 0xBFCA4825A0000000)
  %.027 = select i1 %.not61, float %73, float %72
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not62 = icmp eq i32 %74, 0
  %75 = call float @llvm.nvvm.fma.rn.ftz.f(float %.027, float %58, float noundef 0x3FB0A9DA20000000) #139
  %76 = call float @llvm.fma.f32(float %.027, float %58, float noundef 0x3FB0A9DA20000000)
  %.028 = select i1 %.not62, float %76, float %75
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not63 = icmp eq i32 %77, 0
  %78 = call float @llvm.nvvm.fma.rn.ftz.f(float %.028, float %58, float noundef 0xBFBFFFFFE0000000) #139
  %79 = call float @llvm.fma.f32(float %.028, float %58, float noundef 0xBFBFFFFFE0000000)
  %.030 = select i1 %.not63, float %79, float %78
  %80 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not64 = icmp eq i32 %80, 0
  %81 = call float @llvm.nvvm.fma.rn.ftz.f(float %.030, float %57, float %.05) #139
  %82 = call float @llvm.fma.f32(float %.030, float %57, float %.05)
  %.031 = select i1 %.not64, float %82, float %81
  %83 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not65 = icmp eq i32 %83, 0
  br i1 %.not65, label %86, label %84

84:                                               ; preds = %__internal_fmad.exit13.i
  %85 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.05) #139
  br label %__nv_rsqrtf.exit.i

86:                                               ; preds = %__internal_fmad.exit13.i
  %87 = call float @llvm.nvvm.rsqrt.approx.f(float %.05) #139
  br label %__nv_rsqrtf.exit.i

__nv_rsqrtf.exit.i:                               ; preds = %86, %84
  %.032 = phi float [ %85, %84 ], [ %87, %86 ]
  %88 = fmul float %.032, 0x3FE9884540000000
  %89 = fmul float %.026, %88
  %90 = fmul float %.031, 0x3FE45F3060000000
  %91 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not66 = icmp eq i32 %91, 0
  %92 = call i32 @llvm.nvvm.f2i.rn.ftz(float %90) #139
  %93 = call i32 @llvm.nvvm.f2i.rn(float %90) #139
  %.034 = select i1 %.not66, i32 %93, i32 %92
  %94 = sitofp i32 %.034 to float
  %95 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not67 = icmp eq i32 %95, 0
  %96 = call float @llvm.nvvm.fma.rn.ftz.f(float %94, float noundef 0xBFF921FB40000000, float %.031) #139
  %97 = call float @llvm.fma.f32(float %94, float noundef 0xBFF921FB40000000, float %.031)
  %.035 = select i1 %.not67, float %97, float %96
  %98 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not68 = icmp eq i32 %98, 0
  %99 = call float @llvm.nvvm.fma.rn.ftz.f(float %94, float noundef 0xBE74442D00000000, float %.035) #139
  %100 = call float @llvm.fma.f32(float %94, float noundef 0xBE74442D00000000, float %.035)
  %.036 = select i1 %.not68, float %100, float %99
  %101 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not69 = icmp eq i32 %101, 0
  %102 = call float @llvm.nvvm.fma.rn.ftz.f(float %94, float noundef 0xBCF84698A0000000, float %.036) #139
  %103 = call float @llvm.fma.f32(float %94, float noundef 0xBCF84698A0000000, float %.036)
  %.037 = select i1 %.not69, float %103, float %102
  %104 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not70 = icmp eq i32 %104, 0
  %105 = call float @llvm.nvvm.fabs.ftz.f(float %.031) #139
  %106 = call float @llvm.fabs.f32(float %.031)
  %.039 = select i1 %.not70, float %106, float %105
  %107 = fcmp ult float %.039, 1.056150e+05
  br i1 %107, label %__internal_trig_reduction_kernel.exit.i.i, label %__nv_isinff.exit.i.i.i

__nv_isinff.exit.i.i.i:                           ; preds = %__nv_rsqrtf.exit.i
  %108 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not87 = icmp eq i32 %108, 0
  %.040 = select i1 %.not87, float %106, float %105
  %109 = fcmp oeq float %.040, 0x7FF0000000000000
  br i1 %109, label %__nv_fmul_rn.exit.i.i.i, label %113

__nv_fmul_rn.exit.i.i.i:                          ; preds = %__nv_isinff.exit.i.i.i
  %110 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not93 = icmp eq i32 %110, 0
  %111 = call float @llvm.nvvm.mul.rn.ftz.f(float %.031, float noundef 0.000000e+00) #139
  %112 = fmul float %.031, 0.000000e+00
  %.041 = select i1 %.not93, float %112, float %111
  br label %__internal_trig_reduction_kernel.exit.i.i

113:                                              ; preds = %__nv_isinff.exit.i.i.i
  %114 = bitcast float %.031 to i32
  %115 = lshr i32 %114, 23
  %116 = and i32 %115, 255
  %117 = add nsw i32 %116, -128
  %118 = shl i32 %114, 8
  %119 = or i32 %118, -2147483648
  %120 = lshr i32 %117, 5
  br label %121

121:                                              ; preds = %113, %121
  %iq.i.i.i.i.0233 = phi i32 [ 0, %113 ], [ %129, %121 ]
  %hi.i.i.i.i.0232 = phi i32 [ 0, %113 ], [ %127, %121 ]
  %122 = zext i32 %iq.i.i.i.i.0233 to i64
  %123 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %122
  %124 = load i32, i32 addrspace(1)* %123, align 4
  %125 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %124, i32 %119, i32 %hi.i.i.i.i.0232) #141, !srcloc !67
  %126 = extractvalue { i32, i32 } %125, 0
  %127 = extractvalue { i32, i32 } %125, 1
  %128 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %122
  store i32 %126, i32* %128, align 4, !noalias !68
  %129 = add nuw nsw i32 %iq.i.i.i.i.0233, 1
  %exitcond248.not = icmp eq i32 %129, 6
  br i1 %exitcond248.not, label %130, label %121, !llvm.loop !71

130:                                              ; preds = %121
  %131 = and i32 %114, -2147483648
  %132 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %127, i32* %132, align 4, !noalias !68
  %133 = and i32 %115, 31
  %134 = sub nsw i32 6, %120
  %135 = sext i32 %134 to i64
  %136 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %135
  %137 = load i32, i32* %136, align 4
  %138 = sub nsw i32 5, %120
  %139 = sext i32 %138 to i64
  %140 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %139
  %141 = load i32, i32* %140, align 4
  %.not88 = icmp eq i32 %133, 0
  br i1 %.not88, label %__internal_trig_reduction_slowpath.exit.i.i.i, label %142

142:                                              ; preds = %130
  %143 = sub nsw i32 4, %120
  %144 = sub nuw nsw i32 32, %133
  %145 = shl i32 %137, %133
  %146 = lshr i32 %141, %144
  %147 = add i32 %146, %145
  %148 = shl i32 %141, %133
  %149 = sext i32 %143 to i64
  %150 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %149
  %151 = load i32, i32* %150, align 4
  %152 = lshr i32 %151, %144
  %153 = add i32 %152, %148
  br label %__internal_trig_reduction_slowpath.exit.i.i.i

__internal_trig_reduction_slowpath.exit.i.i.i:    ; preds = %142, %130
  %hi.i.i.i.i.1 = phi i32 [ %147, %142 ], [ %137, %130 ]
  %lo.i.i.i.i.0 = phi i32 [ %153, %142 ], [ %141, %130 ]
  %154 = lshr i32 %hi.i.i.i.i.1, 30
  %155 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i.1, i32 %lo.i.i.i.i.0, i32 noundef 2)
  %156 = shl i32 %lo.i.i.i.i.0, 2
  %157 = lshr i32 %155, 31
  %158 = add nuw nsw i32 %157, %154
  %.not89 = icmp eq i32 %131, 0
  %159 = sub nsw i32 0, %158
  %spec.select = select i1 %.not89, i32 %158, i32 %159
  %.not90 = icmp sgt i32 %155, -1
  %160 = xor i32 %131, -2147483648
  %s.i.i.i.i.0 = select i1 %.not90, i32 %131, i32 %160
  %not..not90 = xor i1 %.not90, true
  %161 = sext i1 %not..not90 to i32
  %hi.i.i.i.i.2 = xor i32 %155, %161
  %lo.i.i.i.i.1 = xor i32 %156, %161
  %162 = zext i32 %hi.i.i.i.i.2 to i64
  %163 = shl nuw i64 %162, 32
  %164 = zext i32 %lo.i.i.i.i.1 to i64
  %165 = or i64 %163, %164
  %166 = sitofp i64 %165 to double
  %167 = fmul double %166, 0x3BF921FB54442D19
  %168 = fptrunc double %167 to float
  %.not92 = icmp eq i32 %s.i.i.i.i.0, 0
  %169 = fneg float %168
  %r.i.i.i.i.0 = select i1 %.not92, float %168, float %169
  br label %__internal_trig_reduction_kernel.exit.i.i

__internal_trig_reduction_kernel.exit.i.i:        ; preds = %__nv_fmul_rn.exit.i.i.i, %__internal_trig_reduction_slowpath.exit.i.i.i, %__nv_rsqrtf.exit.i
  %i.i.i.1 = phi i32 [ %.034, %__nv_rsqrtf.exit.i ], [ 0, %__nv_fmul_rn.exit.i.i.i ], [ %spec.select, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %t.i.i.i.1 = phi float [ %.037, %__nv_rsqrtf.exit.i ], [ %.041, %__nv_fmul_rn.exit.i.i.i ], [ %r.i.i.i.i.0, %__internal_trig_reduction_slowpath.exit.i.i.i ]
  %170 = and i32 %i.i.i.1, 3
  %171 = uitofp i32 %170 to float
  %172 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not71 = icmp eq i32 %172, 0
  %173 = call float @llvm.nvvm.fma.rn.ftz.f(float %171, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000) #139
  %174 = call float @llvm.fma.f32(float %171, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000)
  %.044 = select i1 %.not71, float %174, float %173
  %175 = fadd float %t.i.i.i.1, %.044
  %176 = fmul float %175, 0x3FE45F3060000000
  %177 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not72 = icmp eq i32 %177, 0
  %178 = call i32 @llvm.nvvm.f2i.rn.ftz(float %176) #139
  %179 = call i32 @llvm.nvvm.f2i.rn(float %176) #139
  %.045 = select i1 %.not72, i32 %179, i32 %178
  %180 = sitofp i32 %.045 to float
  %181 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not73 = icmp eq i32 %181, 0
  %182 = call float @llvm.nvvm.fma.rn.ftz.f(float %180, float noundef 0xBFF921FB40000000, float %175) #139
  %183 = call float @llvm.fma.f32(float %180, float noundef 0xBFF921FB40000000, float %175)
  %.046 = select i1 %.not73, float %183, float %182
  %184 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not74 = icmp eq i32 %184, 0
  %185 = call float @llvm.nvvm.fma.rn.ftz.f(float %180, float noundef 0xBE74442D00000000, float %.046) #139
  %186 = call float @llvm.fma.f32(float %180, float noundef 0xBE74442D00000000, float %.046)
  %.048 = select i1 %.not74, float %186, float %185
  %187 = add i32 %.045, 1
  %188 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not75 = icmp eq i32 %188, 0
  %189 = call float @llvm.nvvm.mul.rn.ftz.f(float %.048, float %.048) #139
  %190 = fmul float %.048, %.048
  %.050 = select i1 %.not75, float %190, float %189
  %191 = and i32 %187, 1
  %.not76 = icmp eq i32 %191, 0
  %192 = select i1 %.not76, float %.048, float 1.000000e+00
  %193 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not77 = icmp eq i32 %193, 0
  %194 = call float @llvm.nvvm.fma.rn.ftz.f(float %.050, float %192, float noundef 0.000000e+00) #139
  %195 = call float @llvm.fma.f32(float %.050, float %192, float noundef 0.000000e+00)
  %.051 = select i1 %.not77, float %195, float %194
  br i1 %.not76, label %__internal_fmad.exit2.i.i.i, label %__internal_fmad.exit1.i.i.i

__internal_fmad.exit1.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i
  %196 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not86 = icmp eq i32 %196, 0
  %197 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.050, float noundef 0xBF56C0FDA0000000) #139
  %198 = call float @llvm.fma.f32(float %.050, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.052 = select i1 %.not86, float %198, float %197
  br label %__internal_fmad.exit2.i.i.i

__internal_fmad.exit2.i.i.i:                      ; preds = %__internal_trig_reduction_kernel.exit.i.i, %__internal_fmad.exit1.i.i.i
  %199 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i ]
  %200 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i ]
  %201 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i ], [ %.052, %__internal_fmad.exit1.i.i.i ]
  %202 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not80 = icmp eq i32 %202, 0
  %203 = call float @llvm.nvvm.fma.rn.ftz.f(float %201, float %.050, float noundef %200) #139
  %204 = call float @llvm.fma.f32(float %201, float %.050, float noundef %200)
  %.053 = select i1 %.not80, float %204, float %203
  %205 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not82 = icmp eq i32 %205, 0
  %206 = call float @llvm.nvvm.fma.rn.ftz.f(float %.053, float %.050, float noundef %199) #139
  %207 = call float @llvm.fma.f32(float %.053, float %.050, float noundef %199)
  %.054 = select i1 %.not82, float %207, float %206
  %208 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not83 = icmp eq i32 %208, 0
  %209 = call float @llvm.nvvm.fma.rn.ftz.f(float %.054, float %.051, float %192) #139
  %210 = call float @llvm.fma.f32(float %.054, float %.051, float %192)
  %.055 = select i1 %.not83, float %210, float %209
  %211 = and i32 %187, 2
  %.not84 = icmp eq i32 %211, 0
  br i1 %.not84, label %__internal_cos_offset_f.exit.i, label %__internal_fmad.exit5.i.i.i

__internal_fmad.exit5.i.i.i:                      ; preds = %__internal_fmad.exit2.i.i.i
  %212 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not85 = icmp eq i32 %212, 0
  %213 = call float @llvm.nvvm.fma.rn.ftz.f(float %.055, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %214 = call float @llvm.fma.f32(float %.055, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.056 = select i1 %.not85, float %214, float %213
  br label %__internal_cos_offset_f.exit.i

__internal_cos_offset_f.exit.i:                   ; preds = %__internal_fmad.exit5.i.i.i, %__internal_fmad.exit2.i.i.i
  %z.i.i.i.0 = phi float [ %.056, %__internal_fmad.exit5.i.i.i ], [ %.055, %__internal_fmad.exit2.i.i.i ]
  %215 = fmul float %89, %z.i.i.i.0
  br label %__nv_j0f.exit

__nv_fabsf.exit.i40:                              ; preds = %0
  %216 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %216, 0
  %217 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %218 = call float @llvm.fabs.f32(float %a)
  %.057 = select i1 %.not, float %218, float %217
  %219 = fcmp ugt float %.057, 8.000000e+00
  br i1 %219, label %__nv_isinff.exit.i53, label %__internal_fmad.exit.i41

__internal_fmad.exit.i41:                         ; preds = %__nv_fabsf.exit.i40
  %220 = fadd float %.057, 0xC00EA75580000000
  %221 = fadd float %220, 0x3E74A121E0000000
  %222 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %222, 0
  %223 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3D35E68C60000000, float %221, float noundef 0x3D7037E840000000) #139
  %224 = call float @llvm.fma.f32(float %221, float noundef 0x3D35E68C60000000, float noundef 0x3D7037E840000000)
  %.058 = select i1 %.not43, float %224, float %223
  %225 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %225, 0
  %226 = call float @llvm.nvvm.fma.rn.ftz.f(float %.058, float %221, float noundef 0xBDBC43D820000000) #139
  %227 = call float @llvm.fma.f32(float %.058, float %221, float noundef 0xBDBC43D820000000)
  %.059 = select i1 %.not44, float %227, float %226
  %228 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %228, 0
  %229 = call float @llvm.nvvm.fma.rn.ftz.f(float %.059, float %221, float noundef 0xBDEBBBDFE0000000) #139
  %230 = call float @llvm.fma.f32(float %.059, float %221, float noundef 0xBDEBBBDFE0000000)
  %.060 = select i1 %.not45, float %230, float %229
  %231 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %231, 0
  %232 = call float @llvm.nvvm.fma.rn.ftz.f(float %.060, float %221, float noundef 0x3E336193A0000000) #139
  %233 = call float @llvm.fma.f32(float %.060, float %221, float noundef 0x3E336193A0000000)
  %.061 = select i1 %.not46, float %233, float %232
  %234 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %234, 0
  %235 = call float @llvm.nvvm.fma.rn.ftz.f(float %.061, float %221, float noundef 0x3E5D022E60000000) #139
  %236 = call float @llvm.fma.f32(float %.061, float %221, float noundef 0x3E5D022E60000000)
  %.062 = select i1 %.not47, float %236, float %235
  %237 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %237, 0
  %238 = call float @llvm.nvvm.fma.rn.ftz.f(float %.062, float %221, float noundef 0xBEA1F1B900000000) #139
  %239 = call float @llvm.fma.f32(float %.062, float %221, float noundef 0xBEA1F1B900000000)
  %.063 = select i1 %.not48, float %239, float %238
  %240 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %240, 0
  %241 = call float @llvm.nvvm.fma.rn.ftz.f(float %.063, float %221, float noundef 0xBEC3CCA7A0000000) #139
  %242 = call float @llvm.fma.f32(float %.063, float %221, float noundef 0xBEC3CCA7A0000000)
  %.064 = select i1 %.not49, float %242, float %241
  %243 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %243, 0
  %244 = call float @llvm.nvvm.fma.rn.ftz.f(float %.064, float %221, float noundef 0x3F059B38A0000000) #139
  %245 = call float @llvm.fma.f32(float %.064, float %221, float noundef 0x3F059B38A0000000)
  %.065 = select i1 %.not50, float %245, float %244
  %246 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not51 = icmp eq i32 %246, 0
  %247 = call float @llvm.nvvm.fma.rn.ftz.f(float %.065, float %221, float noundef 0x3F1F3D6200000000) #139
  %248 = call float @llvm.fma.f32(float %.065, float %221, float noundef 0x3F1F3D6200000000)
  %.066 = select i1 %.not51, float %248, float %247
  %249 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not52 = icmp eq i32 %249, 0
  %250 = call float @llvm.nvvm.fma.rn.ftz.f(float %.066, float %221, float noundef 0xBF5D9D7380000000) #139
  %251 = call float @llvm.fma.f32(float %.066, float %221, float noundef 0xBF5D9D7380000000)
  %.067 = select i1 %.not52, float %251, float %250
  %252 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not53 = icmp eq i32 %252, 0
  %253 = call float @llvm.nvvm.fma.rn.ftz.f(float %.067, float %221, float noundef 0xBF64EDFFA0000000) #139
  %254 = call float @llvm.fma.f32(float %.067, float %221, float noundef 0xBF64EDFFA0000000)
  %.068 = select i1 %.not53, float %254, float %253
  %255 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not54 = icmp eq i32 %255, 0
  %256 = call float @llvm.nvvm.fma.rn.ftz.f(float %.068, float %221, float noundef 0x3FA0E73260000000) #139
  %257 = call float @llvm.fma.f32(float %.068, float %221, float noundef 0x3FA0E73260000000)
  %.070 = select i1 %.not54, float %257, float %256
  %258 = fadd float %.057, 0xC01C0FF600000000
  %259 = fadd float %258, 0x3E88971B60000000
  %260 = fmul float %259, %.070
  %261 = fmul float %221, %260
  %262 = fmul float %.057, %261
  br label %426

__nv_isinff.exit.i53:                             ; preds = %__nv_fabsf.exit.i40
  %263 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %263, 0
  %264 = call float @llvm.nvvm.fabs.ftz.f(float %.057) #139
  %265 = call float @llvm.fabs.f32(float %.057)
  %.071 = select i1 %.not5, float %265, float %264
  %266 = fcmp oeq float %.071, 0x7FF0000000000000
  br i1 %266, label %426, label %__internal_fmad.exit12.i54

__internal_fmad.exit12.i54:                       ; preds = %__nv_isinff.exit.i53
  %267 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.057) #141, !srcloc !73
  %268 = fmul float %267, %267
  %269 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %269, 0
  %270 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC0105966E0000000, float %268, float noundef 0x3FE7FEFD20000000) #139
  %271 = call float @llvm.fma.f32(float %268, float noundef 0xC0105966E0000000, float noundef 0x3FE7FEFD20000000)
  %.072 = select i1 %.not6, float %271, float %270
  %272 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %272, 0
  %273 = call float @llvm.nvvm.fma.rn.ftz.f(float %.072, float %268, float noundef 0xBFC8B175C0000000) #139
  %274 = call float @llvm.fma.f32(float %.072, float %268, float noundef 0xBFC8B175C0000000)
  %.073 = select i1 %.not7, float %274, float %273
  %275 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %275, 0
  %276 = call float @llvm.nvvm.fma.rn.ftz.f(float %.073, float %268, float noundef 0x3FC7FFF160000000) #139
  %277 = call float @llvm.fma.f32(float %.073, float %268, float noundef 0x3FC7FFF160000000)
  %.075 = select i1 %.not8, float %277, float %276
  %278 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %278, 0
  %279 = call float @llvm.nvvm.fma.rn.ftz.f(float %.075, float %268, float noundef 1.000000e+00) #139
  %280 = call float @llvm.fma.f32(float %.075, float %268, float noundef 1.000000e+00)
  %.076 = select i1 %.not9, float %280, float %279
  %281 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %281, 0
  %282 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFF9477440000000, float %268, float noundef 0x3FD72295A0000000) #139
  %283 = call float @llvm.fma.f32(float %268, float noundef 0xBFF9477440000000, float noundef 0x3FD72295A0000000)
  %.078 = select i1 %.not10, float %283, float %282
  %284 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %284, 0
  %285 = call float @llvm.nvvm.fma.rn.ftz.f(float %.078, float %268, float noundef 0xBFC4FE5D80000000) #139
  %286 = call float @llvm.fma.f32(float %.078, float %268, float noundef 0xBFC4FE5D80000000)
  %.080 = select i1 %.not11, float %286, float %285
  %287 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %287, 0
  %288 = call float @llvm.nvvm.fma.rn.ftz.f(float %.080, float %268, float noundef 0x3FD7FFFFA0000000) #139
  %289 = call float @llvm.fma.f32(float %.080, float %268, float noundef 0x3FD7FFFFA0000000)
  %.081 = select i1 %.not12, float %289, float %288
  %290 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %290, 0
  %291 = call float @llvm.nvvm.fma.rn.ftz.f(float %.081, float %267, float %.057) #139
  %292 = call float @llvm.fma.f32(float %.081, float %267, float %.057)
  %.082 = select i1 %.not13, float %292, float %291
  %293 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %293, 0
  br i1 %.not14, label %296, label %294

294:                                              ; preds = %__internal_fmad.exit12.i54
  %295 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.057) #139
  br label %__nv_rsqrtf.exit.i62

296:                                              ; preds = %__internal_fmad.exit12.i54
  %297 = call float @llvm.nvvm.rsqrt.approx.f(float %.057) #139
  br label %__nv_rsqrtf.exit.i62

__nv_rsqrtf.exit.i62:                             ; preds = %296, %294
  %.084 = phi float [ %295, %294 ], [ %297, %296 ]
  %298 = fmul float %.084, 0x3FE9884540000000
  %299 = fmul float %.076, %298
  %300 = fmul float %.082, 0x3FE45F3060000000
  %301 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %301, 0
  %302 = call i32 @llvm.nvvm.f2i.rn.ftz(float %300) #139
  %303 = call i32 @llvm.nvvm.f2i.rn(float %300) #139
  %.087 = select i1 %.not15, i32 %303, i32 %302
  %304 = sitofp i32 %.087 to float
  %305 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %305, 0
  %306 = call float @llvm.nvvm.fma.rn.ftz.f(float %304, float noundef 0xBFF921FB40000000, float %.082) #139
  %307 = call float @llvm.fma.f32(float %304, float noundef 0xBFF921FB40000000, float %.082)
  %.088 = select i1 %.not16, float %307, float %306
  %308 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %308, 0
  %309 = call float @llvm.nvvm.fma.rn.ftz.f(float %304, float noundef 0xBE74442D00000000, float %.088) #139
  %310 = call float @llvm.fma.f32(float %304, float noundef 0xBE74442D00000000, float %.088)
  %.089 = select i1 %.not17, float %310, float %309
  %311 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %311, 0
  %312 = call float @llvm.nvvm.fma.rn.ftz.f(float %304, float noundef 0xBCF84698A0000000, float %.089) #139
  %313 = call float @llvm.fma.f32(float %304, float noundef 0xBCF84698A0000000, float %.089)
  %.090 = select i1 %.not18, float %313, float %312
  %314 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %314, 0
  %315 = call float @llvm.nvvm.fabs.ftz.f(float %.082) #139
  %316 = call float @llvm.fabs.f32(float %.082)
  %.091 = select i1 %.not19, float %316, float %315
  %317 = fcmp ult float %.091, 1.056150e+05
  br i1 %317, label %__internal_trig_reduction_kernel.exit.i.i71, label %__nv_isinff.exit.i.i.i68

__nv_isinff.exit.i.i.i68:                         ; preds = %__nv_rsqrtf.exit.i62
  %318 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not36 = icmp eq i32 %318, 0
  %.092 = select i1 %.not36, float %316, float %315
  %319 = fcmp oeq float %.092, 0x7FF0000000000000
  br i1 %319, label %__nv_fmul_rn.exit.i.i.i69, label %323

__nv_fmul_rn.exit.i.i.i69:                        ; preds = %__nv_isinff.exit.i.i.i68
  %320 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %320, 0
  %321 = call float @llvm.nvvm.mul.rn.ftz.f(float %.082, float noundef 0.000000e+00) #139
  %322 = fmul float %.082, 0.000000e+00
  %.094 = select i1 %.not42, float %322, float %321
  br label %__internal_trig_reduction_kernel.exit.i.i71

323:                                              ; preds = %__nv_isinff.exit.i.i.i68
  %324 = bitcast float %.082 to i32
  %325 = lshr i32 %324, 23
  %326 = and i32 %325, 255
  %327 = add nsw i32 %326, -128
  %328 = shl i32 %324, 8
  %329 = or i32 %328, -2147483648
  %330 = lshr i32 %327, 5
  br label %331

331:                                              ; preds = %323, %331
  %iq.i.i.i.i18.0235 = phi i32 [ 0, %323 ], [ %339, %331 ]
  %hi.i.i.i.i13.0234 = phi i32 [ 0, %323 ], [ %337, %331 ]
  %332 = zext i32 %iq.i.i.i.i18.0235 to i64
  %333 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %332
  %334 = load i32, i32 addrspace(1)* %333, align 4
  %335 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %334, i32 %329, i32 %hi.i.i.i.i13.0234) #141, !srcloc !67
  %336 = extractvalue { i32, i32 } %335, 0
  %337 = extractvalue { i32, i32 } %335, 1
  %338 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %332
  store i32 %336, i32* %338, align 4, !noalias !68
  %339 = add nuw nsw i32 %iq.i.i.i.i18.0235, 1
  %exitcond249.not = icmp eq i32 %339, 6
  br i1 %exitcond249.not, label %340, label %331, !llvm.loop !71

340:                                              ; preds = %331
  %341 = and i32 %324, -2147483648
  %342 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %337, i32* %342, align 4, !noalias !68
  %343 = and i32 %325, 31
  %344 = sub nsw i32 6, %330
  %345 = sext i32 %344 to i64
  %346 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %345
  %347 = load i32, i32* %346, align 4
  %348 = sub nsw i32 5, %330
  %349 = sext i32 %348 to i64
  %350 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %349
  %351 = load i32, i32* %350, align 4
  %.not37 = icmp eq i32 %343, 0
  br i1 %.not37, label %__internal_trig_reduction_slowpath.exit.i.i.i70, label %352

352:                                              ; preds = %340
  %353 = sub nsw i32 4, %330
  %354 = sub nuw nsw i32 32, %343
  %355 = shl i32 %347, %343
  %356 = lshr i32 %351, %354
  %357 = add i32 %356, %355
  %358 = shl i32 %351, %343
  %359 = sext i32 %353 to i64
  %360 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %359
  %361 = load i32, i32* %360, align 4
  %362 = lshr i32 %361, %354
  %363 = add i32 %362, %358
  br label %__internal_trig_reduction_slowpath.exit.i.i.i70

__internal_trig_reduction_slowpath.exit.i.i.i70:  ; preds = %352, %340
  %hi.i.i.i.i13.1 = phi i32 [ %357, %352 ], [ %347, %340 ]
  %lo.i.i.i.i14.0 = phi i32 [ %363, %352 ], [ %351, %340 ]
  %364 = lshr i32 %hi.i.i.i.i13.1, 30
  %365 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i13.1, i32 %lo.i.i.i.i14.0, i32 noundef 2)
  %366 = shl i32 %lo.i.i.i.i14.0, 2
  %367 = lshr i32 %365, 31
  %368 = add nuw nsw i32 %367, %364
  %.not38 = icmp eq i32 %341, 0
  %369 = sub nsw i32 0, %368
  %spec.select1 = select i1 %.not38, i32 %368, i32 %369
  %.not39 = icmp sgt i32 %365, -1
  %370 = xor i32 %341, -2147483648
  %s.i.i.i.i11.0 = select i1 %.not39, i32 %341, i32 %370
  %not..not39 = xor i1 %.not39, true
  %371 = sext i1 %not..not39 to i32
  %hi.i.i.i.i13.2 = xor i32 %365, %371
  %lo.i.i.i.i14.1 = xor i32 %366, %371
  %372 = zext i32 %hi.i.i.i.i13.2 to i64
  %373 = shl nuw i64 %372, 32
  %374 = zext i32 %lo.i.i.i.i14.1 to i64
  %375 = or i64 %373, %374
  %376 = sitofp i64 %375 to double
  %377 = fmul double %376, 0x3BF921FB54442D19
  %378 = fptrunc double %377 to float
  %.not41 = icmp eq i32 %s.i.i.i.i11.0, 0
  %379 = fneg float %378
  %r.i.i.i.i22.0 = select i1 %.not41, float %378, float %379
  br label %__internal_trig_reduction_kernel.exit.i.i71

__internal_trig_reduction_kernel.exit.i.i71:      ; preds = %__nv_fmul_rn.exit.i.i.i69, %__internal_trig_reduction_slowpath.exit.i.i.i70, %__nv_rsqrtf.exit.i62
  %i.i.i28.1 = phi i32 [ %.087, %__nv_rsqrtf.exit.i62 ], [ 0, %__nv_fmul_rn.exit.i.i.i69 ], [ %spec.select1, %__internal_trig_reduction_slowpath.exit.i.i.i70 ]
  %t.i.i.i25.1 = phi float [ %.090, %__nv_rsqrtf.exit.i62 ], [ %.094, %__nv_fmul_rn.exit.i.i.i69 ], [ %r.i.i.i.i22.0, %__internal_trig_reduction_slowpath.exit.i.i.i70 ]
  %380 = and i32 %i.i.i28.1, 3
  %381 = uitofp i32 %380 to float
  %382 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %382, 0
  %383 = call float @llvm.nvvm.fma.rn.ftz.f(float %381, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %384 = call float @llvm.fma.f32(float %381, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.097 = select i1 %.not20, float %384, float %383
  %385 = fadd float %t.i.i.i25.1, %.097
  %386 = fmul float %385, 0x3FE45F3060000000
  %387 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %387, 0
  %388 = call i32 @llvm.nvvm.f2i.rn.ftz(float %386) #139
  %389 = call i32 @llvm.nvvm.f2i.rn(float %386) #139
  %.098 = select i1 %.not21, i32 %389, i32 %388
  %390 = sitofp i32 %.098 to float
  %391 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %391, 0
  %392 = call float @llvm.nvvm.fma.rn.ftz.f(float %390, float noundef 0xBFF921FB40000000, float %385) #139
  %393 = call float @llvm.fma.f32(float %390, float noundef 0xBFF921FB40000000, float %385)
  %.099 = select i1 %.not22, float %393, float %392
  %394 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %394, 0
  %395 = call float @llvm.nvvm.fma.rn.ftz.f(float %390, float noundef 0xBE74442D00000000, float %.099) #139
  %396 = call float @llvm.fma.f32(float %390, float noundef 0xBE74442D00000000, float %.099)
  %.0100 = select i1 %.not23, float %396, float %395
  %397 = add i32 %.098, 1
  %398 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %398, 0
  %399 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0100, float %.0100) #139
  %400 = fmul float %.0100, %.0100
  %.0102 = select i1 %.not24, float %400, float %399
  %401 = and i32 %397, 1
  %.not25 = icmp eq i32 %401, 0
  %402 = select i1 %.not25, float %.0100, float 1.000000e+00
  %403 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not26 = icmp eq i32 %403, 0
  %404 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0102, float %402, float noundef 0.000000e+00) #139
  %405 = call float @llvm.fma.f32(float %.0102, float %402, float noundef 0.000000e+00)
  %.0103 = select i1 %.not26, float %405, float %404
  br i1 %.not25, label %__internal_fmad.exit2.i.i.i79, label %__internal_fmad.exit1.i.i.i78

__internal_fmad.exit1.i.i.i78:                    ; preds = %__internal_trig_reduction_kernel.exit.i.i71
  %406 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not35 = icmp eq i32 %406, 0
  %407 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0102, float noundef 0xBF56C0FDA0000000) #139
  %408 = call float @llvm.fma.f32(float %.0102, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0104 = select i1 %.not35, float %408, float %407
  br label %__internal_fmad.exit2.i.i.i79

__internal_fmad.exit2.i.i.i79:                    ; preds = %__internal_trig_reduction_kernel.exit.i.i71, %__internal_fmad.exit1.i.i.i78
  %409 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i71 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i78 ]
  %410 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i71 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i78 ]
  %411 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i71 ], [ %.0104, %__internal_fmad.exit1.i.i.i78 ]
  %412 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not29 = icmp eq i32 %412, 0
  %413 = call float @llvm.nvvm.fma.rn.ftz.f(float %411, float %.0102, float noundef %410) #139
  %414 = call float @llvm.fma.f32(float %411, float %.0102, float noundef %410)
  %.0105 = select i1 %.not29, float %414, float %413
  %415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %415, 0
  %416 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0105, float %.0102, float noundef %409) #139
  %417 = call float @llvm.fma.f32(float %.0105, float %.0102, float noundef %409)
  %.0106 = select i1 %.not31, float %417, float %416
  %418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %418, 0
  %419 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0106, float %.0103, float %402) #139
  %420 = call float @llvm.fma.f32(float %.0106, float %.0103, float %402)
  %.0108 = select i1 %.not32, float %420, float %419
  %421 = and i32 %397, 2
  %.not33 = icmp eq i32 %421, 0
  br i1 %.not33, label %__internal_cos_offset_f.exit.i83, label %__internal_fmad.exit5.i.i.i82

__internal_fmad.exit5.i.i.i82:                    ; preds = %__internal_fmad.exit2.i.i.i79
  %422 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not34 = icmp eq i32 %422, 0
  %423 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0108, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %424 = call float @llvm.fma.f32(float %.0108, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0109 = select i1 %.not34, float %424, float %423
  br label %__internal_cos_offset_f.exit.i83

__internal_cos_offset_f.exit.i83:                 ; preds = %__internal_fmad.exit5.i.i.i82, %__internal_fmad.exit2.i.i.i79
  %z.i.i.i4.0 = phi float [ %.0109, %__internal_fmad.exit5.i.i.i82 ], [ %.0108, %__internal_fmad.exit2.i.i.i79 ]
  %425 = fmul float %299, %z.i.i.i4.0
  br label %426

426:                                              ; preds = %__internal_cos_offset_f.exit.i83, %__nv_isinff.exit.i53, %__internal_fmad.exit.i41
  %r.i34.1 = phi float [ %262, %__internal_fmad.exit.i41 ], [ %425, %__internal_cos_offset_f.exit.i83 ], [ 0.000000e+00, %__nv_isinff.exit.i53 ]
  %427 = fcmp olt float %a, 0.000000e+00
  %428 = fneg float %r.i34.1
  %r.i34.2 = select i1 %427, float %428, float %r.i34.1
  %429 = fcmp olt float %.057, 0x39B4484C00000000
  br i1 %429, label %430, label %__nv_j0f.exit

430:                                              ; preds = %426
  %431 = bitcast float %a to i32
  %432 = and i32 %431, -2147483648
  %433 = bitcast float %r.i34.2 to i32
  %434 = and i32 %433, 2147483647
  %435 = or i32 %434, %432
  %436 = bitcast i32 %435 to float
  br label %__nv_j0f.exit

437:                                              ; preds = %0
  %438 = icmp slt i32 %n, 0
  br i1 %438, label %__nv_j0f.exit, label %__nv_fabsf.exit

__nv_fabsf.exit:                                  ; preds = %437
  %439 = icmp ugt i32 %n, 4
  %440 = zext i1 %439 to i32
  %441 = add nuw nsw i32 %440, %n
  %442 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not107 = icmp eq i32 %442, 0
  %443 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %444 = call float @llvm.fabs.f32(float %a)
  %.0110 = select i1 %.not107, float %444, float %443
  %445 = sitofp i32 %441 to float
  %446 = fcmp ogt float %.0110, %445
  br i1 %446, label %__nv_fabsf.exit.i124, label %897

__nv_fabsf.exit.i124:                             ; preds = %__nv_fabsf.exit
  %447 = fdiv float 2.000000e+00, %a
  %448 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not113 = icmp eq i32 %448, 0
  %.0112 = select i1 %.not113, float %444, float %443
  %449 = fcmp ugt float %.0112, 8.000000e+00
  br i1 %449, label %__nv_isinff.exit.i137, label %__internal_fmad.exit.i125

__internal_fmad.exit.i125:                        ; preds = %__nv_fabsf.exit.i124
  %450 = fadd float %.0112, 0xC00EA75580000000
  %451 = fadd float %450, 0x3E74A121E0000000
  %452 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not205 = icmp eq i32 %452, 0
  %453 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3D35E68C60000000, float %451, float noundef 0x3D7037E840000000) #139
  %454 = call float @llvm.fma.f32(float %451, float noundef 0x3D35E68C60000000, float noundef 0x3D7037E840000000)
  %.0113 = select i1 %.not205, float %454, float %453
  %455 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not206 = icmp eq i32 %455, 0
  %456 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0113, float %451, float noundef 0xBDBC43D820000000) #139
  %457 = call float @llvm.fma.f32(float %.0113, float %451, float noundef 0xBDBC43D820000000)
  %.0114 = select i1 %.not206, float %457, float %456
  %458 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not207 = icmp eq i32 %458, 0
  %459 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0114, float %451, float noundef 0xBDEBBBDFE0000000) #139
  %460 = call float @llvm.fma.f32(float %.0114, float %451, float noundef 0xBDEBBBDFE0000000)
  %.0115 = select i1 %.not207, float %460, float %459
  %461 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not208 = icmp eq i32 %461, 0
  %462 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0115, float %451, float noundef 0x3E336193A0000000) #139
  %463 = call float @llvm.fma.f32(float %.0115, float %451, float noundef 0x3E336193A0000000)
  %.0116 = select i1 %.not208, float %463, float %462
  %464 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not209 = icmp eq i32 %464, 0
  %465 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0116, float %451, float noundef 0x3E5D022E60000000) #139
  %466 = call float @llvm.fma.f32(float %.0116, float %451, float noundef 0x3E5D022E60000000)
  %.0117 = select i1 %.not209, float %466, float %465
  %467 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not210 = icmp eq i32 %467, 0
  %468 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0117, float %451, float noundef 0xBEA1F1B900000000) #139
  %469 = call float @llvm.fma.f32(float %.0117, float %451, float noundef 0xBEA1F1B900000000)
  %.0119 = select i1 %.not210, float %469, float %468
  %470 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not211 = icmp eq i32 %470, 0
  %471 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0119, float %451, float noundef 0xBEC3CCA7A0000000) #139
  %472 = call float @llvm.fma.f32(float %.0119, float %451, float noundef 0xBEC3CCA7A0000000)
  %.0120 = select i1 %.not211, float %472, float %471
  %473 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not212 = icmp eq i32 %473, 0
  %474 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0120, float %451, float noundef 0x3F059B38A0000000) #139
  %475 = call float @llvm.fma.f32(float %.0120, float %451, float noundef 0x3F059B38A0000000)
  %.0121 = select i1 %.not212, float %475, float %474
  %476 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not213 = icmp eq i32 %476, 0
  %477 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0121, float %451, float noundef 0x3F1F3D6200000000) #139
  %478 = call float @llvm.fma.f32(float %.0121, float %451, float noundef 0x3F1F3D6200000000)
  %.0122 = select i1 %.not213, float %478, float %477
  %479 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not214 = icmp eq i32 %479, 0
  %480 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0122, float %451, float noundef 0xBF5D9D7380000000) #139
  %481 = call float @llvm.fma.f32(float %.0122, float %451, float noundef 0xBF5D9D7380000000)
  %.0123 = select i1 %.not214, float %481, float %480
  %482 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not215 = icmp eq i32 %482, 0
  %483 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0123, float %451, float noundef 0xBF64EDFFA0000000) #139
  %484 = call float @llvm.fma.f32(float %.0123, float %451, float noundef 0xBF64EDFFA0000000)
  %.0124 = select i1 %.not215, float %484, float %483
  %485 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not216 = icmp eq i32 %485, 0
  %486 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0124, float %451, float noundef 0x3FA0E73260000000) #139
  %487 = call float @llvm.fma.f32(float %.0124, float %451, float noundef 0x3FA0E73260000000)
  %.0126 = select i1 %.not216, float %487, float %486
  %488 = fadd float %.0112, 0xC01C0FF600000000
  %489 = fadd float %488, 0x3E88971B60000000
  %490 = fmul float %489, %.0126
  %491 = fmul float %451, %490
  %492 = fmul float %.0112, %491
  br label %656

__nv_isinff.exit.i137:                            ; preds = %__nv_fabsf.exit.i124
  %493 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not114 = icmp eq i32 %493, 0
  %494 = call float @llvm.nvvm.fabs.ftz.f(float %.0112) #139
  %495 = call float @llvm.fabs.f32(float %.0112)
  %.0127 = select i1 %.not114, float %495, float %494
  %496 = fcmp oeq float %.0127, 0x7FF0000000000000
  br i1 %496, label %656, label %__internal_fmad.exit12.i138

__internal_fmad.exit12.i138:                      ; preds = %__nv_isinff.exit.i137
  %497 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0112) #141, !srcloc !73
  %498 = fmul float %497, %497
  %499 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not115 = icmp eq i32 %499, 0
  %500 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC0105966E0000000, float %498, float noundef 0x3FE7FEFD20000000) #139
  %501 = call float @llvm.fma.f32(float %498, float noundef 0xC0105966E0000000, float noundef 0x3FE7FEFD20000000)
  %.0128 = select i1 %.not115, float %501, float %500
  %502 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not116 = icmp eq i32 %502, 0
  %503 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0128, float %498, float noundef 0xBFC8B175C0000000) #139
  %504 = call float @llvm.fma.f32(float %.0128, float %498, float noundef 0xBFC8B175C0000000)
  %.0129 = select i1 %.not116, float %504, float %503
  %505 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not117 = icmp eq i32 %505, 0
  %506 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0129, float %498, float noundef 0x3FC7FFF160000000) #139
  %507 = call float @llvm.fma.f32(float %.0129, float %498, float noundef 0x3FC7FFF160000000)
  %.0130 = select i1 %.not117, float %507, float %506
  %508 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not118 = icmp eq i32 %508, 0
  %509 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0130, float %498, float noundef 1.000000e+00) #139
  %510 = call float @llvm.fma.f32(float %.0130, float %498, float noundef 1.000000e+00)
  %.0131 = select i1 %.not118, float %510, float %509
  %511 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not119 = icmp eq i32 %511, 0
  %512 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFF9477440000000, float %498, float noundef 0x3FD72295A0000000) #139
  %513 = call float @llvm.fma.f32(float %498, float noundef 0xBFF9477440000000, float noundef 0x3FD72295A0000000)
  %.0133 = select i1 %.not119, float %513, float %512
  %514 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not120 = icmp eq i32 %514, 0
  %515 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0133, float %498, float noundef 0xBFC4FE5D80000000) #139
  %516 = call float @llvm.fma.f32(float %.0133, float %498, float noundef 0xBFC4FE5D80000000)
  %.0134 = select i1 %.not120, float %516, float %515
  %517 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not121 = icmp eq i32 %517, 0
  %518 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0134, float %498, float noundef 0x3FD7FFFFA0000000) #139
  %519 = call float @llvm.fma.f32(float %.0134, float %498, float noundef 0x3FD7FFFFA0000000)
  %.0135 = select i1 %.not121, float %519, float %518
  %520 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not122 = icmp eq i32 %520, 0
  %521 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0135, float %497, float %.0112) #139
  %522 = call float @llvm.fma.f32(float %.0135, float %497, float %.0112)
  %.0136 = select i1 %.not122, float %522, float %521
  %523 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not123 = icmp eq i32 %523, 0
  br i1 %.not123, label %526, label %524

524:                                              ; preds = %__internal_fmad.exit12.i138
  %525 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0112) #139
  br label %__nv_rsqrtf.exit.i146

526:                                              ; preds = %__internal_fmad.exit12.i138
  %527 = call float @llvm.nvvm.rsqrt.approx.f(float %.0112) #139
  br label %__nv_rsqrtf.exit.i146

__nv_rsqrtf.exit.i146:                            ; preds = %526, %524
  %.0137 = phi float [ %525, %524 ], [ %527, %526 ]
  %528 = fmul float %.0137, 0x3FE9884540000000
  %529 = fmul float %.0131, %528
  %530 = fmul float %.0136, 0x3FE45F3060000000
  %531 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not124 = icmp eq i32 %531, 0
  %532 = call i32 @llvm.nvvm.f2i.rn.ftz(float %530) #139
  %533 = call i32 @llvm.nvvm.f2i.rn(float %530) #139
  %.0139 = select i1 %.not124, i32 %533, i32 %532
  %534 = sitofp i32 %.0139 to float
  %535 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not125 = icmp eq i32 %535, 0
  %536 = call float @llvm.nvvm.fma.rn.ftz.f(float %534, float noundef 0xBFF921FB40000000, float %.0136) #139
  %537 = call float @llvm.fma.f32(float %534, float noundef 0xBFF921FB40000000, float %.0136)
  %.0140 = select i1 %.not125, float %537, float %536
  %538 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not126 = icmp eq i32 %538, 0
  %539 = call float @llvm.nvvm.fma.rn.ftz.f(float %534, float noundef 0xBE74442D00000000, float %.0140) #139
  %540 = call float @llvm.fma.f32(float %534, float noundef 0xBE74442D00000000, float %.0140)
  %.0142 = select i1 %.not126, float %540, float %539
  %541 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not127 = icmp eq i32 %541, 0
  %542 = call float @llvm.nvvm.fma.rn.ftz.f(float %534, float noundef 0xBCF84698A0000000, float %.0142) #139
  %543 = call float @llvm.fma.f32(float %534, float noundef 0xBCF84698A0000000, float %.0142)
  %.0143 = select i1 %.not127, float %543, float %542
  %544 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not128 = icmp eq i32 %544, 0
  %545 = call float @llvm.nvvm.fabs.ftz.f(float %.0136) #139
  %546 = call float @llvm.fabs.f32(float %.0136)
  %.0144 = select i1 %.not128, float %546, float %545
  %547 = fcmp ult float %.0144, 1.056150e+05
  br i1 %547, label %__internal_trig_reduction_kernel.exit.i.i155, label %__nv_isinff.exit.i.i.i152

__nv_isinff.exit.i.i.i152:                        ; preds = %__nv_rsqrtf.exit.i146
  %548 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not198 = icmp eq i32 %548, 0
  %.0145 = select i1 %.not198, float %546, float %545
  %549 = fcmp oeq float %.0145, 0x7FF0000000000000
  br i1 %549, label %__nv_fmul_rn.exit.i.i.i153, label %553

__nv_fmul_rn.exit.i.i.i153:                       ; preds = %__nv_isinff.exit.i.i.i152
  %550 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not204 = icmp eq i32 %550, 0
  %551 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0136, float noundef 0.000000e+00) #139
  %552 = fmul float %.0136, 0.000000e+00
  %.0146 = select i1 %.not204, float %552, float %551
  br label %__internal_trig_reduction_kernel.exit.i.i155

553:                                              ; preds = %__nv_isinff.exit.i.i.i152
  %554 = bitcast float %.0136 to i32
  %555 = lshr i32 %554, 23
  %556 = and i32 %555, 255
  %557 = add nsw i32 %556, -128
  %558 = shl i32 %554, 8
  %559 = or i32 %558, -2147483648
  %560 = lshr i32 %557, 5
  br label %561

561:                                              ; preds = %553, %561
  %iq.i.i.i.i102.0223 = phi i32 [ 0, %553 ], [ %569, %561 ]
  %hi.i.i.i.i97.0222 = phi i32 [ 0, %553 ], [ %567, %561 ]
  %562 = zext i32 %iq.i.i.i.i102.0223 to i64
  %563 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %562
  %564 = load i32, i32 addrspace(1)* %563, align 4
  %565 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %564, i32 %559, i32 %hi.i.i.i.i97.0222) #141, !srcloc !67
  %566 = extractvalue { i32, i32 } %565, 0
  %567 = extractvalue { i32, i32 } %565, 1
  %568 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %562
  store i32 %566, i32* %568, align 4, !noalias !68
  %569 = add nuw nsw i32 %iq.i.i.i.i102.0223, 1
  %exitcond247.not = icmp eq i32 %569, 6
  br i1 %exitcond247.not, label %570, label %561, !llvm.loop !71

570:                                              ; preds = %561
  %571 = and i32 %554, -2147483648
  %572 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %567, i32* %572, align 4, !noalias !68
  %573 = and i32 %555, 31
  %574 = sub nsw i32 6, %560
  %575 = sext i32 %574 to i64
  %576 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %575
  %577 = load i32, i32* %576, align 4
  %578 = sub nsw i32 5, %560
  %579 = sext i32 %578 to i64
  %580 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %579
  %581 = load i32, i32* %580, align 4
  %.not199 = icmp eq i32 %573, 0
  br i1 %.not199, label %__internal_trig_reduction_slowpath.exit.i.i.i154, label %582

582:                                              ; preds = %570
  %583 = sub nsw i32 4, %560
  %584 = sub nuw nsw i32 32, %573
  %585 = shl i32 %577, %573
  %586 = lshr i32 %581, %584
  %587 = add i32 %586, %585
  %588 = shl i32 %581, %573
  %589 = sext i32 %583 to i64
  %590 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %589
  %591 = load i32, i32* %590, align 4
  %592 = lshr i32 %591, %584
  %593 = add i32 %592, %588
  br label %__internal_trig_reduction_slowpath.exit.i.i.i154

__internal_trig_reduction_slowpath.exit.i.i.i154: ; preds = %582, %570
  %hi.i.i.i.i97.1 = phi i32 [ %587, %582 ], [ %577, %570 ]
  %lo.i.i.i.i98.0 = phi i32 [ %593, %582 ], [ %581, %570 ]
  %594 = lshr i32 %hi.i.i.i.i97.1, 30
  %595 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i97.1, i32 %lo.i.i.i.i98.0, i32 noundef 2)
  %596 = shl i32 %lo.i.i.i.i98.0, 2
  %597 = lshr i32 %595, 31
  %598 = add nuw nsw i32 %597, %594
  %.not200 = icmp eq i32 %571, 0
  %599 = sub nsw i32 0, %598
  %spec.select2 = select i1 %.not200, i32 %598, i32 %599
  %.not201 = icmp sgt i32 %595, -1
  %600 = xor i32 %571, -2147483648
  %s.i.i.i.i95.0 = select i1 %.not201, i32 %571, i32 %600
  %not..not201 = xor i1 %.not201, true
  %601 = sext i1 %not..not201 to i32
  %hi.i.i.i.i97.2 = xor i32 %595, %601
  %lo.i.i.i.i98.1 = xor i32 %596, %601
  %602 = zext i32 %hi.i.i.i.i97.2 to i64
  %603 = shl nuw i64 %602, 32
  %604 = zext i32 %lo.i.i.i.i98.1 to i64
  %605 = or i64 %603, %604
  %606 = sitofp i64 %605 to double
  %607 = fmul double %606, 0x3BF921FB54442D19
  %608 = fptrunc double %607 to float
  %.not203 = icmp eq i32 %s.i.i.i.i95.0, 0
  %609 = fneg float %608
  %r.i.i.i.i106.0 = select i1 %.not203, float %608, float %609
  br label %__internal_trig_reduction_kernel.exit.i.i155

__internal_trig_reduction_kernel.exit.i.i155:     ; preds = %__nv_fmul_rn.exit.i.i.i153, %__internal_trig_reduction_slowpath.exit.i.i.i154, %__nv_rsqrtf.exit.i146
  %i.i.i112.1 = phi i32 [ %.0139, %__nv_rsqrtf.exit.i146 ], [ 0, %__nv_fmul_rn.exit.i.i.i153 ], [ %spec.select2, %__internal_trig_reduction_slowpath.exit.i.i.i154 ]
  %t.i.i.i109.1 = phi float [ %.0143, %__nv_rsqrtf.exit.i146 ], [ %.0146, %__nv_fmul_rn.exit.i.i.i153 ], [ %r.i.i.i.i106.0, %__internal_trig_reduction_slowpath.exit.i.i.i154 ]
  %610 = and i32 %i.i.i112.1, 3
  %611 = uitofp i32 %610 to float
  %612 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not129 = icmp eq i32 %612, 0
  %613 = call float @llvm.nvvm.fma.rn.ftz.f(float %611, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000) #139
  %614 = call float @llvm.fma.f32(float %611, float noundef 0x3FF921FB60000000, float noundef 0xC002D97C80000000)
  %.0149 = select i1 %.not129, float %614, float %613
  %615 = fadd float %t.i.i.i109.1, %.0149
  %616 = fmul float %615, 0x3FE45F3060000000
  %617 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not130 = icmp eq i32 %617, 0
  %618 = call i32 @llvm.nvvm.f2i.rn.ftz(float %616) #139
  %619 = call i32 @llvm.nvvm.f2i.rn(float %616) #139
  %.0150 = select i1 %.not130, i32 %619, i32 %618
  %620 = sitofp i32 %.0150 to float
  %621 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not131 = icmp eq i32 %621, 0
  %622 = call float @llvm.nvvm.fma.rn.ftz.f(float %620, float noundef 0xBFF921FB40000000, float %615) #139
  %623 = call float @llvm.fma.f32(float %620, float noundef 0xBFF921FB40000000, float %615)
  %.0151 = select i1 %.not131, float %623, float %622
  %624 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not132 = icmp eq i32 %624, 0
  %625 = call float @llvm.nvvm.fma.rn.ftz.f(float %620, float noundef 0xBE74442D00000000, float %.0151) #139
  %626 = call float @llvm.fma.f32(float %620, float noundef 0xBE74442D00000000, float %.0151)
  %.0152 = select i1 %.not132, float %626, float %625
  %627 = add i32 %.0150, 1
  %628 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not133 = icmp eq i32 %628, 0
  %629 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0152, float %.0152) #139
  %630 = fmul float %.0152, %.0152
  %.0154 = select i1 %.not133, float %630, float %629
  %631 = and i32 %627, 1
  %.not134 = icmp eq i32 %631, 0
  %632 = select i1 %.not134, float %.0152, float 1.000000e+00
  %633 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not135 = icmp eq i32 %633, 0
  %634 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0154, float %632, float noundef 0.000000e+00) #139
  %635 = call float @llvm.fma.f32(float %.0154, float %632, float noundef 0.000000e+00)
  %.0155 = select i1 %.not135, float %635, float %634
  br i1 %.not134, label %__internal_fmad.exit2.i.i.i163, label %__internal_fmad.exit1.i.i.i162

__internal_fmad.exit1.i.i.i162:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i155
  %636 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not197 = icmp eq i32 %636, 0
  %637 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0154, float noundef 0xBF56C0FDA0000000) #139
  %638 = call float @llvm.fma.f32(float %.0154, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.0156 = select i1 %.not197, float %638, float %637
  br label %__internal_fmad.exit2.i.i.i163

__internal_fmad.exit2.i.i.i163:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i155, %__internal_fmad.exit1.i.i.i162
  %639 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i155 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i162 ]
  %640 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i155 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i162 ]
  %641 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i155 ], [ %.0156, %__internal_fmad.exit1.i.i.i162 ]
  %642 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not138 = icmp eq i32 %642, 0
  %643 = call float @llvm.nvvm.fma.rn.ftz.f(float %641, float %.0154, float noundef %640) #139
  %644 = call float @llvm.fma.f32(float %641, float %.0154, float noundef %640)
  %.0157 = select i1 %.not138, float %644, float %643
  %645 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not140 = icmp eq i32 %645, 0
  %646 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0157, float %.0154, float noundef %639) #139
  %647 = call float @llvm.fma.f32(float %.0157, float %.0154, float noundef %639)
  %.0158 = select i1 %.not140, float %647, float %646
  %648 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not141 = icmp eq i32 %648, 0
  %649 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0158, float %.0155, float %632) #139
  %650 = call float @llvm.fma.f32(float %.0158, float %.0155, float %632)
  %.0159 = select i1 %.not141, float %650, float %649
  %651 = and i32 %627, 2
  %.not142 = icmp eq i32 %651, 0
  br i1 %.not142, label %__internal_cos_offset_f.exit.i167, label %__internal_fmad.exit5.i.i.i166

__internal_fmad.exit5.i.i.i166:                   ; preds = %__internal_fmad.exit2.i.i.i163
  %652 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not196 = icmp eq i32 %652, 0
  %653 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0159, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %654 = call float @llvm.fma.f32(float %.0159, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.0161 = select i1 %.not196, float %654, float %653
  br label %__internal_cos_offset_f.exit.i167

__internal_cos_offset_f.exit.i167:                ; preds = %__internal_fmad.exit5.i.i.i166, %__internal_fmad.exit2.i.i.i163
  %z.i.i.i88.0 = phi float [ %.0161, %__internal_fmad.exit5.i.i.i166 ], [ %.0159, %__internal_fmad.exit2.i.i.i163 ]
  %655 = fmul float %529, %z.i.i.i88.0
  br label %656

656:                                              ; preds = %__internal_cos_offset_f.exit.i167, %__nv_isinff.exit.i137, %__internal_fmad.exit.i125
  %r.i118.1 = phi float [ %492, %__internal_fmad.exit.i125 ], [ %655, %__internal_cos_offset_f.exit.i167 ], [ 0.000000e+00, %__nv_isinff.exit.i137 ]
  %657 = fcmp olt float %a, 0.000000e+00
  %658 = fneg float %r.i118.1
  %r.i118.2 = select i1 %657, float %658, float %r.i118.1
  %659 = fcmp olt float %.0112, 0x39B4484C00000000
  br i1 %659, label %660, label %__nv_j1f.exit168

660:                                              ; preds = %656
  %661 = bitcast float %a to i32
  %662 = and i32 %661, -2147483648
  %663 = bitcast float %r.i118.2 to i32
  %664 = and i32 %663, 2147483647
  %665 = or i32 %664, %662
  %666 = bitcast i32 %665 to float
  br label %__nv_j1f.exit168

__nv_j1f.exit168:                                 ; preds = %660, %656
  %r.i118.3 = phi float [ %666, %660 ], [ %r.i118.2, %656 ]
  %667 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not143 = icmp eq i32 %667, 0
  %.0162 = select i1 %.not143, float %444, float %443
  %668 = fcmp ugt float %.0162, 8.000000e+00
  br i1 %668, label %__nv_isinff.exit.i220, label %__internal_fmad.exit.i207

__internal_fmad.exit.i207:                        ; preds = %__nv_j1f.exit168
  %669 = fadd float %.0162, 0xC0033D1520000000
  %670 = fadd float %669, 0xBE7D2E3660000000
  %671 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not183 = icmp eq i32 %671, 0
  %672 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBCD6771CE0000000, float %670, float noundef 0xBD35953660000000) #139
  %673 = call float @llvm.fma.f32(float %670, float noundef 0xBCD6771CE0000000, float noundef 0xBD35953660000000)
  %.0163 = select i1 %.not183, float %673, float %672
  %674 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not184 = icmp eq i32 %674, 0
  %675 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0163, float %670, float noundef 0x3D87E1C300000000) #139
  %676 = call float @llvm.fma.f32(float %.0163, float %670, float noundef 0x3D87E1C300000000)
  %.0165 = select i1 %.not184, float %676, float %675
  %677 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not185 = icmp eq i32 %677, 0
  %678 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0165, float %670, float noundef 0xBD9A82F020000000) #139
  %679 = call float @llvm.fma.f32(float %.0165, float %670, float noundef 0xBD9A82F020000000)
  %.0166 = select i1 %.not185, float %679, float %678
  %680 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not186 = icmp eq i32 %680, 0
  %681 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0166, float %670, float noundef 0xBDFD21E700000000) #139
  %682 = call float @llvm.fma.f32(float %.0166, float %670, float noundef 0xBDFD21E700000000)
  %.0167 = select i1 %.not186, float %682, float %681
  %683 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not187 = icmp eq i32 %683, 0
  %684 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0167, float %670, float noundef 0x3E04060B60000000) #139
  %685 = call float @llvm.fma.f32(float %.0167, float %670, float noundef 0x3E04060B60000000)
  %.0168 = select i1 %.not187, float %685, float %684
  %686 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not188 = icmp eq i32 %686, 0
  %687 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0168, float %670, float noundef 0x3E6F2E2860000000) #139
  %688 = call float @llvm.fma.f32(float %.0168, float %670, float noundef 0x3E6F2E2860000000)
  %.0169 = select i1 %.not188, float %688, float %687
  %689 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not189 = icmp eq i32 %689, 0
  %690 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0169, float %670, float noundef 0x3E1EEDF0A0000000) #139
  %691 = call float @llvm.fma.f32(float %.0169, float %670, float noundef 0x3E1EEDF0A0000000)
  %.0170 = select i1 %.not189, float %691, float %690
  %692 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not190 = icmp eq i32 %692, 0
  %693 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0170, float %670, float noundef 0xBED6DBF8C0000000) #139
  %694 = call float @llvm.fma.f32(float %.0170, float %670, float noundef 0xBED6DBF8C0000000)
  %.0164 = select i1 %.not190, float %694, float %693
  %695 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not191 = icmp eq i32 %695, 0
  %696 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0164, float %670, float noundef 0xBEDECCB920000000) #139
  %697 = call float @llvm.fma.f32(float %.0164, float %670, float noundef 0xBEDECCB920000000)
  %.0160 = select i1 %.not191, float %697, float %696
  %698 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not192 = icmp eq i32 %698, 0
  %699 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0160, float %670, float noundef 0x3F33C5BD60000000) #139
  %700 = call float @llvm.fma.f32(float %.0160, float %670, float noundef 0x3F33C5BD60000000)
  %.0153 = select i1 %.not192, float %700, float %699
  %701 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not193 = icmp eq i32 %701, 0
  %702 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0153, float %670, float noundef 0x3F495C6680000000) #139
  %703 = call float @llvm.fma.f32(float %.0153, float %670, float noundef 0x3F495C6680000000)
  %.0148 = select i1 %.not193, float %703, float %702
  %704 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not194 = icmp eq i32 %704, 0
  %705 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0148, float %670, float noundef 0xBF7DD54360000000) #139
  %706 = call float @llvm.fma.f32(float %.0148, float %670, float noundef 0xBF7DD54360000000)
  %.0147 = select i1 %.not194, float %706, float %705
  %707 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not195 = icmp eq i32 %707, 0
  %708 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0147, float %670, float noundef 0xBF9B4EE8E0000000) #139
  %709 = call float @llvm.fma.f32(float %.0147, float %670, float noundef 0xBF9B4EE8E0000000)
  %.0141 = select i1 %.not195, float %709, float %708
  %710 = fmul float %670, %.0141
  %711 = fadd float %.0162, 0xC016148F60000000
  %712 = fadd float %711, 0x3E734F46E0000000
  %713 = fmul float %712, %710
  %714 = fadd float %.0162, 0xC0214EB560000000
  %715 = fadd float %714, 0xBE9999BDA0000000
  %716 = fmul float %715, %713
  br label %__nv_j0f.exit251

__nv_isinff.exit.i220:                            ; preds = %__nv_j1f.exit168
  %717 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not144 = icmp eq i32 %717, 0
  %718 = call float @llvm.nvvm.fabs.ftz.f(float %.0162) #139
  %719 = call float @llvm.fabs.f32(float %.0162)
  %.0138 = select i1 %.not144, float %719, float %718
  %720 = fcmp oeq float %.0138, 0x7FF0000000000000
  br i1 %720, label %__nv_j0f.exit251, label %__internal_fmad.exit13.i221

__internal_fmad.exit13.i221:                      ; preds = %__nv_isinff.exit.i220
  %721 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.0162) #141, !srcloc !73
  %722 = fmul float %721, %721
  %723 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not145 = icmp eq i32 %723, 0
  %724 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400ADFD260000000, float %722, float noundef 0xBFE076F840000000) #139
  %725 = call float @llvm.fma.f32(float %722, float noundef 0x400ADFD260000000, float noundef 0xBFE076F840000000)
  %.0132 = select i1 %.not145, float %725, float %724
  %726 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not146 = icmp eq i32 %726, 0
  %727 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0132, float %722, float noundef 0x3FBA767E60000000) #139
  %728 = call float @llvm.fma.f32(float %.0132, float %722, float noundef 0x3FBA767E60000000)
  %.0125 = select i1 %.not146, float %728, float %727
  %729 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not147 = icmp eq i32 %729, 0
  %730 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0125, float %722, float noundef 0xBFAFFFF6C0000000) #139
  %731 = call float @llvm.fma.f32(float %.0125, float %722, float noundef 0xBFAFFFF6C0000000)
  %.0118 = select i1 %.not147, float %731, float %730
  %732 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not148 = icmp eq i32 %732, 0
  %733 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0118, float %722, float noundef 1.000000e+00) #139
  %734 = call float @llvm.fma.f32(float %.0118, float %722, float noundef 1.000000e+00)
  %.0111 = select i1 %.not148, float %734, float %733
  %735 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not149 = icmp eq i32 %735, 0
  %736 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF23C0120000000, float %722, float noundef 0xBFCA4825A0000000) #139
  %737 = call float @llvm.fma.f32(float %722, float noundef 0x3FF23C0120000000, float noundef 0xBFCA4825A0000000)
  %.0107 = select i1 %.not149, float %737, float %736
  %738 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not150 = icmp eq i32 %738, 0
  %739 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0107, float %722, float noundef 0x3FB0A9DA20000000) #139
  %740 = call float @llvm.fma.f32(float %.0107, float %722, float noundef 0x3FB0A9DA20000000)
  %.0101 = select i1 %.not150, float %740, float %739
  %741 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not151 = icmp eq i32 %741, 0
  %742 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0101, float %722, float noundef 0xBFBFFFFFE0000000) #139
  %743 = call float @llvm.fma.f32(float %.0101, float %722, float noundef 0xBFBFFFFFE0000000)
  %.096 = select i1 %.not151, float %743, float %742
  %744 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not152 = icmp eq i32 %744, 0
  %745 = call float @llvm.nvvm.fma.rn.ftz.f(float %.096, float %721, float %.0162) #139
  %746 = call float @llvm.fma.f32(float %.096, float %721, float %.0162)
  %.095 = select i1 %.not152, float %746, float %745
  %747 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not153 = icmp eq i32 %747, 0
  br i1 %.not153, label %750, label %748

748:                                              ; preds = %__internal_fmad.exit13.i221
  %749 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %.0162) #139
  br label %__nv_rsqrtf.exit.i229

750:                                              ; preds = %__internal_fmad.exit13.i221
  %751 = call float @llvm.nvvm.rsqrt.approx.f(float %.0162) #139
  br label %__nv_rsqrtf.exit.i229

__nv_rsqrtf.exit.i229:                            ; preds = %750, %748
  %.093 = phi float [ %749, %748 ], [ %751, %750 ]
  %752 = fmul float %.093, 0x3FE9884540000000
  %753 = fmul float %.0111, %752
  %754 = fmul float %.095, 0x3FE45F3060000000
  %755 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not154 = icmp eq i32 %755, 0
  %756 = call i32 @llvm.nvvm.f2i.rn.ftz(float %754) #139
  %757 = call i32 @llvm.nvvm.f2i.rn(float %754) #139
  %.086 = select i1 %.not154, i32 %757, i32 %756
  %758 = sitofp i32 %.086 to float
  %759 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not155 = icmp eq i32 %759, 0
  %760 = call float @llvm.nvvm.fma.rn.ftz.f(float %758, float noundef 0xBFF921FB40000000, float %.095) #139
  %761 = call float @llvm.fma.f32(float %758, float noundef 0xBFF921FB40000000, float %.095)
  %.085 = select i1 %.not155, float %761, float %760
  %762 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not156 = icmp eq i32 %762, 0
  %763 = call float @llvm.nvvm.fma.rn.ftz.f(float %758, float noundef 0xBE74442D00000000, float %.085) #139
  %764 = call float @llvm.fma.f32(float %758, float noundef 0xBE74442D00000000, float %.085)
  %.083 = select i1 %.not156, float %764, float %763
  %765 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not157 = icmp eq i32 %765, 0
  %766 = call float @llvm.nvvm.fma.rn.ftz.f(float %758, float noundef 0xBCF84698A0000000, float %.083) #139
  %767 = call float @llvm.fma.f32(float %758, float noundef 0xBCF84698A0000000, float %.083)
  %.079 = select i1 %.not157, float %767, float %766
  %768 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not158 = icmp eq i32 %768, 0
  %769 = call float @llvm.nvvm.fabs.ftz.f(float %.095) #139
  %770 = call float @llvm.fabs.f32(float %.095)
  %.077 = select i1 %.not158, float %770, float %769
  %771 = fcmp ult float %.077, 1.056150e+05
  br i1 %771, label %__internal_trig_reduction_kernel.exit.i.i238, label %__nv_isinff.exit.i.i.i235

__nv_isinff.exit.i.i.i235:                        ; preds = %__nv_rsqrtf.exit.i229
  %772 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not176 = icmp eq i32 %772, 0
  %.074 = select i1 %.not176, float %770, float %769
  %773 = fcmp oeq float %.074, 0x7FF0000000000000
  br i1 %773, label %__nv_fmul_rn.exit.i.i.i236, label %777

__nv_fmul_rn.exit.i.i.i236:                       ; preds = %__nv_isinff.exit.i.i.i235
  %774 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not182 = icmp eq i32 %774, 0
  %775 = call float @llvm.nvvm.mul.rn.ftz.f(float %.095, float noundef 0.000000e+00) #139
  %776 = fmul float %.095, 0.000000e+00
  %.069 = select i1 %.not182, float %776, float %775
  br label %__internal_trig_reduction_kernel.exit.i.i238

777:                                              ; preds = %__nv_isinff.exit.i.i.i235
  %778 = bitcast float %.095 to i32
  %779 = lshr i32 %778, 23
  %780 = and i32 %779, 255
  %781 = add nsw i32 %780, -128
  %782 = shl i32 %778, 8
  %783 = or i32 %782, -2147483648
  %784 = lshr i32 %781, 5
  br label %785

785:                                              ; preds = %777, %785
  %hi.i.i.i.i179.0221 = phi i32 [ 0, %777 ], [ %791, %785 ]
  %iq.i.i.i.i184.0220 = phi i32 [ 0, %777 ], [ %793, %785 ]
  %786 = zext i32 %iq.i.i.i.i184.0220 to i64
  %787 = getelementptr inbounds [6 x i32], [6 x i32] addrspace(1)* @__cudart_i2opi_f, i64 0, i64 %786
  %788 = load i32, i32 addrspace(1)* %787, align 4
  %789 = call { i32, i32 } asm "{\0A\09mad.lo.cc.u32   $0, $2, $3, $4;\0A\09madc.hi.u32     $1, $2, $3,  0;\0A\09}", "=r,=r,r,r,r"(i32 %788, i32 %783, i32 %hi.i.i.i.i179.0221) #141, !srcloc !67
  %790 = extractvalue { i32, i32 } %789, 0
  %791 = extractvalue { i32, i32 } %789, 1
  %792 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %786
  store i32 %790, i32* %792, align 4, !noalias !68
  %793 = add nuw nsw i32 %iq.i.i.i.i184.0220, 1
  %exitcond.not = icmp eq i32 %793, 6
  br i1 %exitcond.not, label %794, label %785, !llvm.loop !71

794:                                              ; preds = %785
  %795 = and i32 %778, -2147483648
  %796 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 6
  store i32 %791, i32* %796, align 4, !noalias !68
  %797 = and i32 %779, 31
  %798 = sub nsw i32 6, %784
  %799 = sext i32 %798 to i64
  %800 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %799
  %801 = load i32, i32* %800, align 4
  %802 = sub nsw i32 5, %784
  %803 = sext i32 %802 to i64
  %804 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %803
  %805 = load i32, i32* %804, align 4
  %.not177 = icmp eq i32 %797, 0
  br i1 %.not177, label %__internal_trig_reduction_slowpath.exit.i.i.i237, label %806

806:                                              ; preds = %794
  %807 = sub nsw i32 4, %784
  %808 = sub nuw nsw i32 32, %797
  %809 = shl i32 %801, %797
  %810 = lshr i32 %805, %808
  %811 = add i32 %810, %809
  %812 = shl i32 %805, %797
  %813 = sext i32 %807 to i64
  %814 = getelementptr inbounds [7 x i32], [7 x i32]* %result.i.i.i.i, i64 0, i64 %813
  %815 = load i32, i32* %814, align 4
  %816 = lshr i32 %815, %808
  %817 = add i32 %816, %812
  br label %__internal_trig_reduction_slowpath.exit.i.i.i237

__internal_trig_reduction_slowpath.exit.i.i.i237: ; preds = %806, %794
  %lo.i.i.i.i180.0 = phi i32 [ %817, %806 ], [ %805, %794 ]
  %hi.i.i.i.i179.1 = phi i32 [ %811, %806 ], [ %801, %794 ]
  %818 = lshr i32 %hi.i.i.i.i179.1, 30
  %819 = call i32 @llvm.fshl.i32(i32 %hi.i.i.i.i179.1, i32 %lo.i.i.i.i180.0, i32 noundef 2)
  %820 = shl i32 %lo.i.i.i.i180.0, 2
  %821 = lshr i32 %819, 31
  %822 = add nuw nsw i32 %821, %818
  %.not178 = icmp eq i32 %795, 0
  %823 = sub nsw i32 0, %822
  %spec.select3 = select i1 %.not178, i32 %822, i32 %823
  %.not179 = icmp sgt i32 %819, -1
  %824 = xor i32 %795, -2147483648
  %not..not179 = xor i1 %.not179, true
  %825 = sext i1 %not..not179 to i32
  %lo.i.i.i.i180.1 = xor i32 %820, %825
  %hi.i.i.i.i179.2 = xor i32 %819, %825
  %s.i.i.i.i177.0 = select i1 %.not179, i32 %795, i32 %824
  %826 = zext i32 %hi.i.i.i.i179.2 to i64
  %827 = shl nuw i64 %826, 32
  %828 = zext i32 %lo.i.i.i.i180.1 to i64
  %829 = or i64 %827, %828
  %830 = sitofp i64 %829 to double
  %831 = fmul double %830, 0x3BF921FB54442D19
  %832 = fptrunc double %831 to float
  %.not181 = icmp eq i32 %s.i.i.i.i177.0, 0
  %833 = fneg float %832
  %r.i.i.i.i188.0 = select i1 %.not181, float %832, float %833
  br label %__internal_trig_reduction_kernel.exit.i.i238

__internal_trig_reduction_kernel.exit.i.i238:     ; preds = %__nv_fmul_rn.exit.i.i.i236, %__internal_trig_reduction_slowpath.exit.i.i.i237, %__nv_rsqrtf.exit.i229
  %i.i.i194.1 = phi i32 [ %.086, %__nv_rsqrtf.exit.i229 ], [ 0, %__nv_fmul_rn.exit.i.i.i236 ], [ %spec.select3, %__internal_trig_reduction_slowpath.exit.i.i.i237 ]
  %t.i.i.i191.1 = phi float [ %.079, %__nv_rsqrtf.exit.i229 ], [ %.069, %__nv_fmul_rn.exit.i.i.i236 ], [ %r.i.i.i.i188.0, %__internal_trig_reduction_slowpath.exit.i.i.i237 ]
  %834 = and i32 %i.i.i194.1, 3
  %835 = uitofp i32 %834 to float
  %836 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not159 = icmp eq i32 %836, 0
  %837 = call float @llvm.nvvm.fma.rn.ftz.f(float %835, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000) #139
  %838 = call float @llvm.fma.f32(float %835, float noundef 0x3FF921FB60000000, float noundef 0xBFE921FB60000000)
  %.049 = select i1 %.not159, float %838, float %837
  %839 = fadd float %t.i.i.i191.1, %.049
  %840 = fmul float %839, 0x3FE45F3060000000
  %841 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not160 = icmp eq i32 %841, 0
  %842 = call i32 @llvm.nvvm.f2i.rn.ftz(float %840) #139
  %843 = call i32 @llvm.nvvm.f2i.rn(float %840) #139
  %.047 = select i1 %.not160, i32 %843, i32 %842
  %844 = sitofp i32 %.047 to float
  %845 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not161 = icmp eq i32 %845, 0
  %846 = call float @llvm.nvvm.fma.rn.ftz.f(float %844, float noundef 0xBFF921FB40000000, float %839) #139
  %847 = call float @llvm.fma.f32(float %844, float noundef 0xBFF921FB40000000, float %839)
  %.043 = select i1 %.not161, float %847, float %846
  %848 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not162 = icmp eq i32 %848, 0
  %849 = call float @llvm.nvvm.fma.rn.ftz.f(float %844, float noundef 0xBE74442D00000000, float %.043) #139
  %850 = call float @llvm.fma.f32(float %844, float noundef 0xBE74442D00000000, float %.043)
  %.042 = select i1 %.not162, float %850, float %849
  %851 = add i32 %.047, 1
  %852 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not163 = icmp eq i32 %852, 0
  %853 = call float @llvm.nvvm.mul.rn.ftz.f(float %.042, float %.042) #139
  %854 = fmul float %.042, %.042
  %.038 = select i1 %.not163, float %854, float %853
  %855 = and i32 %851, 1
  %.not164 = icmp eq i32 %855, 0
  %856 = select i1 %.not164, float %.042, float 1.000000e+00
  %857 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not165 = icmp eq i32 %857, 0
  %858 = call float @llvm.nvvm.fma.rn.ftz.f(float %.038, float %856, float noundef 0.000000e+00) #139
  %859 = call float @llvm.fma.f32(float %.038, float %856, float noundef 0.000000e+00)
  %.033 = select i1 %.not165, float %859, float %858
  br i1 %.not164, label %__internal_fmad.exit2.i.i.i246, label %__internal_fmad.exit1.i.i.i245

__internal_fmad.exit1.i.i.i245:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i238
  %860 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not175 = icmp eq i32 %860, 0
  %861 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.038, float noundef 0xBF56C0FDA0000000) #139
  %862 = call float @llvm.fma.f32(float %.038, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.029 = select i1 %.not175, float %862, float %861
  br label %__internal_fmad.exit2.i.i.i246

__internal_fmad.exit2.i.i.i246:                   ; preds = %__internal_trig_reduction_kernel.exit.i.i238, %__internal_fmad.exit1.i.i.i245
  %863 = phi float [ 0xBFC5555500000000, %__internal_trig_reduction_kernel.exit.i.i238 ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i.i.i245 ]
  %864 = phi float [ 0x3F8110BC80000000, %__internal_trig_reduction_kernel.exit.i.i238 ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i.i.i245 ]
  %865 = phi float [ 0xBF29A82A60000000, %__internal_trig_reduction_kernel.exit.i.i238 ], [ %.029, %__internal_fmad.exit1.i.i.i245 ]
  %866 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not168 = icmp eq i32 %866, 0
  %867 = call float @llvm.nvvm.fma.rn.ftz.f(float %865, float %.038, float noundef %864) #139
  %868 = call float @llvm.fma.f32(float %865, float %.038, float noundef %864)
  %.022 = select i1 %.not168, float %868, float %867
  %869 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not170 = icmp eq i32 %869, 0
  %870 = call float @llvm.nvvm.fma.rn.ftz.f(float %.022, float %.038, float noundef %863) #139
  %871 = call float @llvm.fma.f32(float %.022, float %.038, float noundef %863)
  %.015 = select i1 %.not170, float %871, float %870
  %872 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not171 = icmp eq i32 %872, 0
  %873 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %.033, float %856) #139
  %874 = call float @llvm.fma.f32(float %.015, float %.033, float %856)
  %.08 = select i1 %.not171, float %874, float %873
  %875 = and i32 %851, 2
  %.not172 = icmp eq i32 %875, 0
  br i1 %.not172, label %__internal_cos_offset_f.exit.i250, label %__internal_fmad.exit5.i.i.i249

__internal_fmad.exit5.i.i.i249:                   ; preds = %__internal_fmad.exit2.i.i.i246
  %876 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not174 = icmp eq i32 %876, 0
  %877 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %878 = call float @llvm.fma.f32(float %.08, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.04 = select i1 %.not174, float %878, float %877
  br label %__internal_cos_offset_f.exit.i250

__internal_cos_offset_f.exit.i250:                ; preds = %__internal_fmad.exit5.i.i.i249, %__internal_fmad.exit2.i.i.i246
  %z.i.i.i170.0 = phi float [ %.04, %__internal_fmad.exit5.i.i.i249 ], [ %.08, %__internal_fmad.exit2.i.i.i246 ]
  %879 = fmul float %753, %z.i.i.i170.0
  br label %__nv_j0f.exit251

__nv_j0f.exit251:                                 ; preds = %__internal_cos_offset_f.exit.i250, %__nv_isinff.exit.i220, %__internal_fmad.exit.i207
  %r.i200.1 = phi float [ %716, %__internal_fmad.exit.i207 ], [ %879, %__internal_cos_offset_f.exit.i250 ], [ 0.000000e+00, %__nv_isinff.exit.i220 ]
  %880 = icmp sgt i32 %n, 1
  br i1 %880, label %__internal_fmad.exit.lr.ph, label %__nv_j0f.exit

__internal_fmad.exit.lr.ph:                       ; preds = %__nv_j0f.exit251
  %881 = add nsw i32 %n, -1
  %xtraiter = and i32 %881, 1
  %882 = icmp eq i32 %n, 2
  br i1 %882, label %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa, label %__internal_fmad.exit.lr.ph.new

__internal_fmad.exit.lr.ph.new:                   ; preds = %__internal_fmad.exit.lr.ph
  %unroll_iter = and i32 %881, -2
  br label %__internal_fmad.exit

__internal_fmad.exit:                             ; preds = %__internal_fmad.exit, %__internal_fmad.exit.lr.ph.new
  %jim1.0219 = phi float [ %r.i200.1, %__internal_fmad.exit.lr.ph.new ], [ %.02, %__internal_fmad.exit ]
  %ji.0218 = phi float [ %r.i118.3, %__internal_fmad.exit.lr.ph.new ], [ %.02.1, %__internal_fmad.exit ]
  %i.0217 = phi i32 [ 1, %__internal_fmad.exit.lr.ph.new ], [ %896, %__internal_fmad.exit ]
  %niter = phi i32 [ 0, %__internal_fmad.exit.lr.ph.new ], [ %niter.next.1, %__internal_fmad.exit ]
  %883 = sitofp i32 %i.0217 to float
  %884 = fmul float %ji.0218, %883
  %885 = fneg float %jim1.0219
  %886 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not173 = icmp eq i32 %886, 0
  %887 = call float @llvm.nvvm.fma.rn.ftz.f(float %884, float %447, float %885) #139
  %888 = call float @llvm.fma.f32(float %884, float %447, float %885)
  %.02 = select i1 %.not173, float %888, float %887
  %889 = add nuw nsw i32 %i.0217, 1
  %890 = sitofp i32 %889 to float
  %891 = fmul float %.02, %890
  %892 = fneg float %ji.0218
  %893 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not173.1 = icmp eq i32 %893, 0
  %894 = call float @llvm.nvvm.fma.rn.ftz.f(float %891, float %447, float %892) #139
  %895 = call float @llvm.fma.f32(float %891, float %447, float %892)
  %.02.1 = select i1 %.not173.1, float %895, float %894
  %896 = add nuw nsw i32 %i.0217, 2
  %niter.next.1 = add i32 %niter, 2
  %niter.ncmp.1.not = icmp eq i32 %niter.next.1, %unroll_iter
  br i1 %niter.ncmp.1.not, label %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa.loopexit, label %__internal_fmad.exit

897:                                              ; preds = %__nv_fabsf.exit
  %898 = mul nsw i32 %n, 40
  %899 = sitofp i32 %898 to float
  %900 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not108 = icmp eq i32 %900, 0
  br i1 %.not108, label %907, label %901

901:                                              ; preds = %897
  %902 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not112 = icmp eq i32 %902, 0
  br i1 %.not112, label %905, label %903

903:                                              ; preds = %901
  %904 = call float @llvm.nvvm.sqrt.rn.ftz.f(float %899) #139
  br label %__nv_sqrtf.exit

905:                                              ; preds = %901
  %906 = call float @llvm.nvvm.sqrt.approx.ftz.f(float %899) #139
  br label %__nv_sqrtf.exit

907:                                              ; preds = %897
  %908 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([17 x i8], [17 x i8]* @.str.2, i64 0, i64 0)) #139
  %.not109 = icmp eq i32 %908, 0
  br i1 %.not109, label %911, label %909

909:                                              ; preds = %907
  %910 = call float @llvm.sqrt.f32(float %899)
  br label %__nv_sqrtf.exit

911:                                              ; preds = %907
  %912 = call float @llvm.nvvm.sqrt.approx.f(float %899) #139
  br label %__nv_sqrtf.exit

__nv_sqrtf.exit:                                  ; preds = %911, %909, %905, %903
  %.01 = phi float [ %904, %903 ], [ %906, %905 ], [ %910, %909 ], [ %912, %911 ]
  %913 = fptosi float %.01 to i32
  %914 = add nsw i32 %913, %n
  %915 = and i32 %914, -2
  %916 = icmp sgt i32 %915, 0
  br i1 %916, label %__nv_fabsf.exit252, label %__nv_j0f.exit

__nv_fabsf.exit252:                               ; preds = %__nv_sqrtf.exit, %931
  %ji1.0228 = phi float [ %jim12.0, %931 ], [ 1.000000e+00, %__nv_sqrtf.exit ]
  %i.1227 = phi i32 [ %932, %931 ], [ %915, %__nv_sqrtf.exit ]
  %sum.0226 = phi float [ %sum.2, %931 ], [ 0.000000e+00, %__nv_sqrtf.exit ]
  %lambda.0225 = phi float [ %spec.select4, %931 ], [ 0.000000e+00, %__nv_sqrtf.exit ]
  %jip1.1224 = phi float [ %jip1.2, %931 ], [ 0.000000e+00, %__nv_sqrtf.exit ]
  %917 = sitofp i32 %i.1227 to float
  %918 = fmul float %917, 2.000000e+00
  %919 = fdiv float %918, %a
  %920 = fmul float %ji1.0228, %919
  %921 = fsub float %920, %jip1.1224
  %922 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not110 = icmp eq i32 %922, 0
  %923 = call float @llvm.nvvm.fabs.ftz.f(float %921) #139
  %924 = call float @llvm.fabs.f32(float %921)
  %.0 = select i1 %.not110, float %924, float %923
  %925 = fcmp ogt float %.0, 0x430C6BF520000000
  br i1 %925, label %926, label %931

926:                                              ; preds = %__nv_fabsf.exit252
  %927 = fmul float %921, 0x3CD203AFA0000000
  %928 = fmul float %ji1.0228, 0x3CD203AFA0000000
  %929 = fmul float %lambda.0225, 0x3CD203AFA0000000
  %930 = fmul float %sum.0226, 0x3CD203AFA0000000
  br label %931

931:                                              ; preds = %926, %__nv_fabsf.exit252
  %jip1.2 = phi float [ %928, %926 ], [ %ji1.0228, %__nv_fabsf.exit252 ]
  %lambda.1 = phi float [ %929, %926 ], [ %lambda.0225, %__nv_fabsf.exit252 ]
  %sum.1 = phi float [ %930, %926 ], [ %sum.0226, %__nv_fabsf.exit252 ]
  %jim12.0 = phi float [ %927, %926 ], [ %921, %__nv_fabsf.exit252 ]
  %932 = add nsw i32 %i.1227, -1
  %933 = icmp eq i32 %932, %n
  %spec.select4 = select i1 %933, float %jim12.0, float %lambda.1
  %934 = and i32 %i.1227, 1
  %.not111 = icmp eq i32 %934, 0
  %935 = fmul float %jim12.0, 2.000000e+00
  %936 = select i1 %.not111, float -0.000000e+00, float %935
  %sum.2 = fadd float %sum.1, %936
  %937 = icmp ugt i32 %i.1227, 1
  br i1 %937, label %__nv_fabsf.exit252, label %._crit_edge

._crit_edge:                                      ; preds = %931
  %938 = fsub float %sum.2, %jim12.0
  %939 = fdiv float %spec.select4, %938
  br label %__nv_j0f.exit

.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa.loopexit: ; preds = %__internal_fmad.exit
  %phi.cast = sitofp i32 %896 to float
  br label %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa

.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa:      ; preds = %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa.loopexit, %__internal_fmad.exit.lr.ph
  %.02.lcssa.ph = phi float [ undef, %__internal_fmad.exit.lr.ph ], [ %.02.1, %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa.loopexit ]
  %jim1.0219.unr = phi float [ %r.i200.1, %__internal_fmad.exit.lr.ph ], [ %.02, %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa.loopexit ]
  %ji.0218.unr = phi float [ %r.i118.3, %__internal_fmad.exit.lr.ph ], [ %.02.1, %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa.loopexit ]
  %i.0217.unr = phi float [ 1.000000e+00, %__internal_fmad.exit.lr.ph ], [ %phi.cast, %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa.loopexit ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %__nv_j0f.exit, label %__internal_fmad.exit.epil.preheader

__internal_fmad.exit.epil.preheader:              ; preds = %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa
  %940 = fmul float %ji.0218.unr, %i.0217.unr
  %941 = fneg float %jim1.0219.unr
  %942 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not173.epil = icmp eq i32 %942, 0
  %943 = call float @llvm.nvvm.fma.rn.ftz.f(float %940, float %447, float %941) #139
  %944 = call float @llvm.fma.f32(float %940, float %447, float %941)
  %.02.epil = select i1 %.not173.epil, float %944, float %943
  br label %__nv_j0f.exit

__nv_j0f.exit:                                    ; preds = %__nv_j0f.exit251, %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa, %__internal_fmad.exit.epil.preheader, %__nv_sqrtf.exit, %._crit_edge, %437, %426, %430, %__internal_fmad.exit.i, %__nv_isinff.exit.i, %__internal_cos_offset_f.exit.i
  %.03 = phi float [ %52, %__internal_fmad.exit.i ], [ %215, %__internal_cos_offset_f.exit.i ], [ 0.000000e+00, %__nv_isinff.exit.i ], [ %436, %430 ], [ %r.i34.2, %426 ], [ 0x7FFFFFFFE0000000, %437 ], [ %939, %._crit_edge ], [ -0.000000e+00, %__nv_sqrtf.exit ], [ 0.000000e+00, %__nv_j0f.exit251 ], [ %.02.lcssa.ph, %.__nv_j0f.exit.loopexit_crit_edge.unr-lcssa ], [ %.02.epil, %__internal_fmad.exit.epil.preheader ]
  ret float %.03
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_erff(float %a) local_unnamed_addr #49 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = fcmp oge float %.0, 0x3FF00C1FC0000000
  br i1 %3, label %__nv_fabsf.exit1, label %5

__nv_fabsf.exit1:                                 ; preds = %__nv_fabsf.exit
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %4, 0
  %.01 = select i1 %.not8, float %2, float %1
  br label %__internal_fmad.exit

5:                                                ; preds = %__nv_fabsf.exit
  %6 = fmul float %a, %a
  br label %__internal_fmad.exit

__internal_fmad.exit:                             ; preds = %5, %__nv_fabsf.exit1
  %7 = phi float [ 0x3FC06EBA60000000, %5 ], [ 0x3FE4214280000000, %__nv_fabsf.exit1 ]
  %8 = phi float [ 0xBFD8127580000000, %5 ], [ 0x3FED369F20000000, %__nv_fabsf.exit1 ]
  %9 = phi float [ 0x3FBCE315E0000000, %5 ], [ 0x3FC46AA320000000, %__nv_fabsf.exit1 ]
  %10 = phi float [ 0xBF9B837CE0000000, %5 ], [ 0xBFA49B3340000000, %__nv_fabsf.exit1 ]
  %11 = phi float [ 0x3F755ABD40000000, %5 ], [ 0x3F813233E0000000, %__nv_fabsf.exit1 ]
  %12 = phi float [ 0xBF4AE9A400000000, %5 ], [ 0xBF55C00B60000000, %__nv_fabsf.exit1 ]
  %13 = phi float [ 0x3F163D2D40000000, %5 ], [ 0x3F1D698740000000, %__nv_fabsf.exit1 ]
  %14 = phi float [ %6, %5 ], [ %.01, %__nv_fabsf.exit1 ]
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef %13, float %14, float noundef %12) #139
  %17 = call float @llvm.fma.f32(float noundef %13, float %14, float noundef %12)
  %.02 = select i1 %.not1, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %14, float noundef %11) #139
  %20 = call float @llvm.fma.f32(float %.02, float %14, float noundef %11)
  %.03 = select i1 %.not2, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %14, float noundef %10) #139
  %23 = call float @llvm.fma.f32(float %.03, float %14, float noundef %10)
  %.04 = select i1 %.not3, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %14, float noundef %9) #139
  %26 = call float @llvm.fma.f32(float %.04, float %14, float noundef %9)
  %.05 = select i1 %.not4, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %14, float noundef %8) #139
  %29 = call float @llvm.fma.f32(float %.05, float %14, float noundef %8)
  %.06 = select i1 %.not5, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %14, float noundef %7) #139
  %32 = call float @llvm.fma.f32(float %.06, float %14, float noundef %7)
  %.07 = select i1 %.not6, float %32, float %31
  %33 = fneg float %14
  %34 = select i1 %3, float %33, float %a
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %34, float %34) #139
  %37 = call float @llvm.fma.f32(float %.07, float %34, float %34)
  %.08 = select i1 %.not7, float %37, float %36
  br i1 %3, label %38, label %46

38:                                               ; preds = %__internal_fmad.exit
  %39 = call float @llvm.nvvm.ex2.approx.ftz.f(float %.08) #139
  %40 = fsub float 1.000000e+00, %39
  %41 = bitcast float %40 to i32
  %42 = bitcast float %a to i32
  %43 = and i32 %42, -2147483648
  %44 = or i32 %43, %41
  %45 = bitcast i32 %44 to float
  br label %46

46:                                               ; preds = %38, %__internal_fmad.exit
  %r.0 = phi float [ %45, %38 ], [ %.08, %__internal_fmad.exit ]
  ret float %r.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_erfinvf(float %a) local_unnamed_addr #50 {
__nv_fmaf_rn.exit:
  %0 = fneg float %a
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call float @llvm.nvvm.fma.rn.ftz.f(float %a, float %0, float noundef 1.000000e+00) #139
  %3 = call float @llvm.fma.f32(float %a, float %0, float noundef 1.000000e+00)
  %.0 = select i1 %.not, float %3, float %2
  %4 = call float @llvm.nvvm.lg2.approx.ftz.f(float %.0) #139
  %5 = fneg float %4
  %6 = fcmp olt float %4, 0xC020666660000000
  br i1 %6, label %__internal_fmad.exit, label %__internal_fmad.exit.i

__internal_fmad.exit:                             ; preds = %__nv_fmaf_rn.exit
  %7 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %5) #139
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFE2E09420000000, float %7, float noundef 0xBFE53754A0000000) #139
  %10 = call float @llvm.fma.f32(float %7, float noundef 0xBFE2E09420000000, float noundef 0xBFE53754A0000000)
  %.011 = select i1 %.not10, float %10, float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %7, float noundef 0x3FF98D5B80000000) #139
  %13 = call float @llvm.fma.f32(float %.011, float %7, float noundef 0x3FF98D5B80000000)
  %.012 = select i1 %.not11, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %7, float noundef 0xBFE59B5DA0000000) #139
  %16 = call float @llvm.fma.f32(float %.012, float %7, float noundef 0xBFE59B5DA0000000)
  %.013 = select i1 %.not12, float %16, float %15
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %7, float noundef 0xBFB860A6E0000000) #139
  %19 = call float @llvm.fma.f32(float %.013, float %7, float noundef 0xBFB860A6E0000000)
  %.014 = select i1 %.not13, float %19, float %18
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %7, float noundef 0x3FEABB3720000000) #139
  %22 = call float @llvm.fma.f32(float %.014, float %7, float noundef 0x3FEABB3720000000)
  %.07 = select i1 %.not14, float %22, float %21
  %23 = fdiv float 1.000000e+00, %7
  %24 = fmul float %23, %.07
  %25 = bitcast float %24 to i32
  %26 = bitcast float %a to i32
  %27 = and i32 %26, -2147483648
  %28 = or i32 %27, %25
  %29 = bitcast i32 %28 to float
  br label %58

__internal_fmad.exit.i:                           ; preds = %__nv_fmaf_rn.exit
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBDF14C6E00000000, float %5, float noundef 0x3E443EC8A0000000) #139
  %32 = call float @llvm.fma.f32(float %5, float noundef 0xBDF14C6E00000000, float noundef 0x3E443EC8A0000000)
  %.01 = select i1 %.not1, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %5, float noundef 0xBE802DFB40000000) #139
  %35 = call float @llvm.fma.f32(float %.01, float %5, float noundef 0xBE802DFB40000000)
  %.02 = select i1 %.not2, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %5, float noundef 0x3E8D1F08C0000000) #139
  %38 = call float @llvm.fma.f32(float %.02, float %5, float noundef 0x3E8D1F08C0000000)
  %.03 = select i1 %.not3, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %5, float noundef 0x3EE0E85540000000) #139
  %41 = call float @llvm.fma.f32(float %.03, float %5, float noundef 0x3EE0E85540000000)
  %.04 = select i1 %.not4, float %41, float %40
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %5, float noundef 0xBF009B69A0000000) #139
  %44 = call float @llvm.fma.f32(float %.04, float %5, float noundef 0xBF009B69A0000000)
  %.05 = select i1 %.not5, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %5, float noundef 0xBF495FD420000000) #139
  %47 = call float @llvm.fma.f32(float %.05, float %5, float noundef 0xBF495FD420000000)
  %.06 = select i1 %.not6, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %5, float noundef 0x3F76B804E0000000) #139
  %50 = call float @llvm.fma.f32(float %.06, float %5, float noundef 0x3F76B804E0000000)
  %.08 = select i1 %.not7, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %5, float noundef 0x3FC495C1E0000000) #139
  %53 = call float @llvm.fma.f32(float %.08, float %5, float noundef 0x3FC495C1E0000000)
  %.09 = select i1 %.not8, float %53, float %52
  %54 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %54, 0
  %55 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %5, float noundef 0x3FEC5BF880000000) #139
  %56 = call float @llvm.fma.f32(float %.09, float %5, float noundef 0x3FEC5BF880000000)
  %.010 = select i1 %.not9, float %56, float %55
  %57 = fmul float %.010, %a
  br label %58

58:                                               ; preds = %__internal_fmad.exit.i, %__internal_fmad.exit
  %t.0 = phi float [ %29, %__internal_fmad.exit ], [ %57, %__internal_fmad.exit.i ]
  ret float %t.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_erfcf(float %a) local_unnamed_addr #51 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fadd float %.01, -4.000000e+00
  %4 = fadd float %.01, 4.000000e+00
  %5 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %4) #141, !srcloc !73
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.mul.rn.ftz.f(float %3, float %5) #139
  %8 = fmul float %5, %3
  %.04 = select i1 %.not1, float %8, float %7
  %9 = fadd float %.04, 1.000000e+00
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %10, 0
  %11 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef -4.000000e+00, float %9, float %.01) #139
  %12 = call float @llvm.fma.f32(float %9, float noundef -4.000000e+00, float %.01)
  %.05 = select i1 %.not2, float %12, float %11
  %13 = fneg float %.04
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %13, float %.01, float %.05) #139
  %16 = call float @llvm.fma.f32(float %13, float %.01, float %.05)
  %.06 = select i1 %.not3, float %16, float %15
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fma.rn.ftz.f(float %5, float %.06, float %.04) #139
  %19 = call float @llvm.fma.f32(float %5, float %.06, float %.04)
  %.07 = select i1 %.not4, float %19, float %18
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F4D341220000000, float %.07, float noundef 0x3F7CDC0B60000000) #139
  %22 = call float @llvm.fma.f32(float %.07, float noundef 0x3F4D341220000000, float noundef 0x3F7CDC0B60000000)
  %.08 = select i1 %.not5, float %22, float %21
  %23 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %23, 0
  %24 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %.07, float noundef 0xBF903F6960000000) #139
  %25 = call float @llvm.fma.f32(float %.08, float %.07, float noundef 0xBF903F6960000000)
  %.010 = select i1 %.not6, float %25, float %24
  %26 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %26, 0
  %27 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %.07, float noundef 0x3FA2A6E760000000) #139
  %28 = call float @llvm.fma.f32(float %.010, float %.07, float noundef 0x3FA2A6E760000000)
  %.011 = select i1 %.not7, float %28, float %27
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %29, 0
  %30 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %.07, float noundef 0xBFB10F8B40000000) #139
  %31 = call float @llvm.fma.f32(float %.011, float %.07, float noundef 0xBFB10F8B40000000)
  %.012 = select i1 %.not8, float %31, float %30
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %.07, float noundef 0x3FB8043AA0000000) #139
  %34 = call float @llvm.fma.f32(float %.012, float %.07, float noundef 0x3FB8043AA0000000)
  %.013 = select i1 %.not9, float %34, float %33
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %.07, float noundef 0xBFB9DA8480000000) #139
  %37 = call float @llvm.fma.f32(float %.013, float %.07, float noundef 0xBFB9DA8480000000)
  %.014 = select i1 %.not10, float %37, float %36
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %38, 0
  %39 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %.07, float noundef 0x3FB16E9BC0000000) #139
  %40 = call float @llvm.fma.f32(float %.014, float %.07, float noundef 0x3FB16E9BC0000000)
  %.015 = select i1 %.not11, float %40, float %39
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %41, 0
  %42 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %.07, float noundef 0x3F8F7E2E00000000) #139
  %43 = call float @llvm.fma.f32(float %.015, float %.07, float noundef 0x3F8F7E2E00000000)
  %.016 = select i1 %.not12, float %43, float %42
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %.07, float noundef 0xBFC1DF1A80000000) #139
  %46 = call float @llvm.fma.f32(float %.016, float %.07, float noundef 0xBFC1DF1A80000000)
  %.017 = select i1 %.not13, float %46, float %45
  %47 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %47, 0
  %48 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %.07, float noundef 0x3FF3BA5920000000) #139
  %49 = call float @llvm.fma.f32(float %.017, float %.07, float noundef 0x3FF3BA5920000000)
  %.018 = select i1 %.not14, float %49, float %48
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.000000e+00, float %.01, float noundef 1.000000e+00) #139
  %52 = call float @llvm.fma.f32(float %.01, float noundef 2.000000e+00, float noundef 1.000000e+00)
  %.019 = select i1 %.not15, float %52, float %51
  %53 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.019) #141, !srcloc !73
  %54 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %54, 0
  %55 = call float @llvm.nvvm.mul.rn.ftz.f(float %.018, float %53) #139
  %56 = fmul float %53, %.018
  %.020 = select i1 %.not16, float %56, float %55
  %57 = fmul float %.020, -2.000000e+00
  %58 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %58, 0
  %59 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %57, float %.018) #139
  %60 = call float @llvm.fma.f32(float %.01, float %57, float %.018)
  %.021 = select i1 %.not17, float %60, float %59
  %61 = fsub float %.021, %.020
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %61, float %53, float %.020) #139
  %64 = call float @llvm.fma.f32(float %61, float %53, float %.020)
  %.022 = select i1 %.not18, float %64, float %63
  %65 = fneg float %.01
  %66 = fmul float %.01, %65
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %67, 0
  %68 = call float @llvm.nvvm.mul.rn.ftz.f(float %66, float noundef 0x3FF7154760000000) #139
  %69 = fmul float %66, 0x3FF7154760000000
  %.03 = select i1 %.not19, float %69, float %68
  %70 = call float @llvm.trunc.f32(float %.03)
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fabs.ftz.f(float %70) #139
  %73 = call float @llvm.fabs.f32(float %70)
  %.09 = select i1 %.not20, float %73, float %72
  %74 = fcmp ogt float %.09, 1.260000e+02
  %75 = bitcast float %70 to i32
  %76 = and i32 %75, -2147483648
  %77 = or i32 %76, 1123811328
  %78 = bitcast i32 %77 to float
  %j.i.0 = select i1 %74, float %78, float %70
  %79 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %79, 0
  %80 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.0, float noundef 0xBFE62E4300000000, float %66) #139
  %81 = call float @llvm.fma.f32(float %j.i.0, float noundef 0xBFE62E4300000000, float %66)
  %.023 = select i1 %.not21, float %81, float %80
  %82 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %82, 0
  %83 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.0, float noundef 0x3E205C6100000000, float %.023) #139
  %84 = call float @llvm.fma.f32(float %j.i.0, float noundef 0x3E205C6100000000, float %.023)
  %.024 = select i1 %.not22, float %84, float %83
  %85 = fmul float %.024, 0x3FF7154760000000
  %86 = fadd float %j.i.0, 0x4168000FE0000000
  %87 = bitcast float %86 to i32
  %88 = shl i32 %87, 23
  %89 = bitcast i32 %88 to float
  %90 = call float @llvm.nvvm.ex2.approx.ftz.f(float %85) #139
  %91 = fmul float %90, %89
  %92 = fneg float %66
  %93 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %93, 0
  %94 = call float @llvm.nvvm.fma.rn.ftz.f(float %65, float %.01, float %92) #139
  %95 = call float @llvm.fma.f32(float %65, float %.01, float %92)
  %.02 = select i1 %.not23, float %95, float %94
  %96 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %96, 0
  %97 = call float @llvm.nvvm.fma.rn.ftz.f(float %91, float %.02, float %91) #139
  %98 = call float @llvm.fma.f32(float %91, float %.02, float %91)
  %.0 = select i1 %.not24, float %98, float %97
  %99 = fmul float %.0, %.022
  %100 = fcmp ogt float %.01, 0x40241C2900000000
  %t1.0 = select i1 %100, float 0.000000e+00, float %99
  %101 = fcmp olt float %a, 0.000000e+00
  %102 = fsub float 2.000000e+00, %t1.0
  %103 = select i1 %101, float %102, float %t1.0
  ret float %103
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_erfcxf(float %a) local_unnamed_addr #52 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp olt float %.01, 0x40241C2900000000
  br i1 %3, label %__nv_fmul_rn.exit.i, label %66

__nv_fmul_rn.exit.i:                              ; preds = %__nv_fabsf.exit
  %4 = fadd float %.01, -4.000000e+00
  %5 = fadd float %.01, 4.000000e+00
  %6 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %5) #141, !srcloc !73
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %7, 0
  %8 = call float @llvm.nvvm.mul.rn.ftz.f(float %4, float %6) #139
  %9 = fmul float %4, %6
  %.03 = select i1 %.not14, float %9, float %8
  %10 = fadd float %.03, 1.000000e+00
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef -4.000000e+00, float %10, float %.01) #139
  %13 = call float @llvm.fma.f32(float %10, float noundef -4.000000e+00, float %.01)
  %.04 = select i1 %.not15, float %13, float %12
  %14 = fneg float %.03
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %14, float %.01, float %.04) #139
  %17 = call float @llvm.fma.f32(float %14, float %.01, float %.04)
  %.05 = select i1 %.not16, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %6, float %.05, float %.03) #139
  %20 = call float @llvm.fma.f32(float %6, float %.05, float %.03)
  %.06 = select i1 %.not17, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F4D341220000000, float %.06, float noundef 0x3F7CDC0B60000000) #139
  %23 = call float @llvm.fma.f32(float %.06, float noundef 0x3F4D341220000000, float noundef 0x3F7CDC0B60000000)
  %.07 = select i1 %.not18, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %.06, float noundef 0xBF903F6960000000) #139
  %26 = call float @llvm.fma.f32(float %.07, float %.06, float noundef 0xBF903F6960000000)
  %.08 = select i1 %.not19, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %.06, float noundef 0x3FA2A6E760000000) #139
  %29 = call float @llvm.fma.f32(float %.08, float %.06, float noundef 0x3FA2A6E760000000)
  %.010 = select i1 %.not20, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %.06, float noundef 0xBFB10F8B40000000) #139
  %32 = call float @llvm.fma.f32(float %.010, float %.06, float noundef 0xBFB10F8B40000000)
  %.011 = select i1 %.not21, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %.06, float noundef 0x3FB8043AA0000000) #139
  %35 = call float @llvm.fma.f32(float %.011, float %.06, float noundef 0x3FB8043AA0000000)
  %.012 = select i1 %.not22, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %.06, float noundef 0xBFB9DA8480000000) #139
  %38 = call float @llvm.fma.f32(float %.012, float %.06, float noundef 0xBFB9DA8480000000)
  %.014 = select i1 %.not23, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %.06, float noundef 0x3FB16E9BC0000000) #139
  %41 = call float @llvm.fma.f32(float %.014, float %.06, float noundef 0x3FB16E9BC0000000)
  %.016 = select i1 %.not24, float %41, float %40
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %.06, float noundef 0x3F8F7E2E00000000) #139
  %44 = call float @llvm.fma.f32(float %.016, float %.06, float noundef 0x3F8F7E2E00000000)
  %.017 = select i1 %.not25, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not26 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %.06, float noundef 0xBFC1DF1A80000000) #139
  %47 = call float @llvm.fma.f32(float %.017, float %.06, float noundef 0xBFC1DF1A80000000)
  %.018 = select i1 %.not26, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %.06, float noundef 0x3FF3BA5920000000) #139
  %50 = call float @llvm.fma.f32(float %.018, float %.06, float noundef 0x3FF3BA5920000000)
  %.020 = select i1 %.not27, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.000000e+00, float %.01, float noundef 1.000000e+00) #139
  %53 = call float @llvm.fma.f32(float %.01, float noundef 2.000000e+00, float noundef 1.000000e+00)
  %.021 = select i1 %.not28, float %53, float %52
  %54 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.021) #141, !srcloc !73
  %55 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not29 = icmp eq i32 %55, 0
  %56 = call float @llvm.nvvm.mul.rn.ftz.f(float %.020, float %54) #139
  %57 = fmul float %54, %.020
  %.022 = select i1 %.not29, float %57, float %56
  %58 = fmul float %.022, -2.000000e+00
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %58, float %.020) #139
  %61 = call float @llvm.fma.f32(float %.01, float %58, float %.020)
  %.023 = select i1 %.not30, float %61, float %60
  %62 = fsub float %.023, %.022
  %63 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %63, 0
  %64 = call float @llvm.nvvm.fma.rn.ftz.f(float %62, float %54, float %.022) #139
  %65 = call float @llvm.fma.f32(float %62, float %54, float %.022)
  %.024 = select i1 %.not31, float %65, float %64
  br label %88

66:                                               ; preds = %__nv_fabsf.exit
  %67 = fmul float %.01, 2.500000e-01
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %68, 0
  br i1 %.not1, label %71, label %69

69:                                               ; preds = %66
  %70 = call float @llvm.nvvm.div.approx.ftz.f(float noundef 2.500000e-01, float %67) #139
  br label %__nv_fast_fdividef.exit

71:                                               ; preds = %66
  %72 = call float @llvm.nvvm.div.approx.f(float noundef 2.500000e-01, float %67) #139
  br label %__nv_fast_fdividef.exit

__nv_fast_fdividef.exit:                          ; preds = %71, %69
  %.025 = phi float [ %70, %69 ], [ %72, %71 ]
  %73 = fmul float %.025, %.025
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %74, 0
  %75 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 6.562500e+00, float %73, float noundef -1.875000e+00) #139
  %76 = call float @llvm.fma.f32(float %73, float noundef 6.562500e+00, float noundef -1.875000e+00)
  %.026 = select i1 %.not2, float %76, float %75
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %77, 0
  %78 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %73, float noundef 7.500000e-01) #139
  %79 = call float @llvm.fma.f32(float %.026, float %73, float noundef 7.500000e-01)
  %.027 = select i1 %.not3, float %79, float %78
  %80 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %80, 0
  %81 = call float @llvm.nvvm.fma.rn.ftz.f(float %.027, float %73, float noundef -5.000000e-01) #139
  %82 = call float @llvm.fma.f32(float %.027, float %73, float noundef -5.000000e-01)
  %.028 = select i1 %.not4, float %82, float %81
  %83 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %83, 0
  %84 = call float @llvm.nvvm.fma.rn.ftz.f(float %.028, float %73, float noundef 1.000000e+00) #139
  %85 = call float @llvm.fma.f32(float %.028, float %73, float noundef 1.000000e+00)
  %.029 = select i1 %.not5, float %85, float %84
  %86 = fmul float %.025, 0x3FE20DD760000000
  %87 = fmul float %86, %.029
  br label %88

88:                                               ; preds = %__nv_fast_fdividef.exit, %__nv_fmul_rn.exit.i
  %t1.0 = phi float [ %.024, %__nv_fmul_rn.exit.i ], [ %87, %__nv_fast_fdividef.exit ]
  %89 = fcmp olt float %a, 0.000000e+00
  br i1 %89, label %__nv_fmul_rz.exit, label %125

__nv_fmul_rz.exit:                                ; preds = %88
  %90 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %90, 0
  %91 = call float @llvm.nvvm.mul.rz.ftz.f(float %.01, float %.01) #139
  %92 = call float @llvm.nvvm.mul.rz.f(float %.01, float %.01) #139
  %.030 = select i1 %.not6, float %92, float %91
  %93 = fneg float %.030
  %94 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %94, 0
  %95 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float %.01, float %93) #139
  %96 = call float @llvm.fma.f32(float %.01, float %.01, float %93)
  %.031 = select i1 %.not7, float %96, float %95
  %97 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %97, 0
  %98 = call float @llvm.nvvm.fma.rn.ftz.f(float %.030, float noundef 0x3F777313A0000000, float noundef 5.000000e-01) #139
  %99 = call float @llvm.fma.f32(float %.030, float noundef 0x3F777313A0000000, float noundef 5.000000e-01)
  %.015 = select i1 %.not8, float %99, float %98
  %100 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %100, 0
  %101 = call float @llvm.nvvm.saturate.ftz.f(float %.015) #139
  %102 = call float @llvm.nvvm.saturate.f(float %.015) #139
  %.013 = select i1 %.not9, float %102, float %101
  %103 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %103, 0
  %104 = call float @llvm.nvvm.fma.rm.ftz.f(float %.013, float noundef 2.520000e+02, float noundef 0x4168000020000000) #139
  %105 = call float @llvm.nvvm.fma.rm.f(float %.013, float noundef 2.520000e+02, float noundef 0x4168000020000000) #139
  %.09 = select i1 %.not10, float %105, float %104
  %106 = fadd float %.09, 0xC168000FE0000000
  %107 = fneg float %106
  %108 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %108, 0
  %109 = call float @llvm.nvvm.fma.rn.ftz.f(float %.030, float noundef 0x3FF7154760000000, float %107) #139
  %110 = call float @llvm.fma.f32(float %.030, float noundef 0x3FF7154760000000, float %107)
  %.02 = select i1 %.not11, float %110, float %109
  %111 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %111, 0
  %112 = call float @llvm.nvvm.fma.rn.ftz.f(float %.030, float noundef 0x3E54AE0C00000000, float %.02) #139
  %113 = call float @llvm.fma.f32(float %.030, float noundef 0x3E54AE0C00000000, float %.02)
  %.019 = select i1 %.not12, float %113, float %112
  %114 = bitcast float %.09 to i32
  %115 = shl i32 %114, 23
  %116 = bitcast i32 %115 to float
  %117 = call float @llvm.nvvm.ex2.approx.ftz.f(float %.019) #139
  %118 = fmul float %117, %116
  %119 = fadd float %118, %118
  %120 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %120, 0
  %121 = call float @llvm.nvvm.fma.rn.ftz.f(float %119, float %.031, float %119) #139
  %122 = call float @llvm.fma.f32(float %119, float %.031, float %119)
  %.0 = select i1 %.not13, float %122, float %121
  %123 = fsub float %.0, %t1.0
  %124 = fcmp oeq float %119, 0x7FF0000000000000
  %t1.1 = select i1 %124, float %119, float %123
  br label %125

125:                                              ; preds = %__nv_fmul_rz.exit, %88
  %t1.2 = phi float [ %t1.1, %__nv_fmul_rz.exit ], [ %t1.0, %88 ]
  ret float %t1.2
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_erfcinvf(float %a) local_unnamed_addr #53 {
__nv_fadd_rn.exit:
  %0 = fneg float %a
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call float @llvm.nvvm.add.rn.ftz.f(float noundef 2.000000e+00, float %0) #139
  %3 = fsub float 2.000000e+00, %a
  %.01 = select i1 %.not, float %3, float %2
  %4 = fcmp oge float %a, 0x3F6BDA5120000000
  %5 = fcmp ole float %a, 0x3FFFF212E0000000
  %or.cond = and i1 %4, %5
  br i1 %or.cond, label %__nv_fmul_rn.exit, label %41

__nv_fmul_rn.exit:                                ; preds = %__nv_fadd_rn.exit
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.mul.rn.ftz.f(float %.01, float %a) #139
  %8 = fmul float %.01, %a
  %.02 = select i1 %.not11, float %8, float %7
  %9 = call float @llvm.nvvm.lg2.approx.ftz.f(float %.02) #139
  %10 = fneg float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBDF14C6E00000000, float %10, float noundef 0x3E443EC8A0000000) #139
  %13 = call float @llvm.fma.f32(float %10, float noundef 0xBDF14C6E00000000, float noundef 0x3E443EC8A0000000)
  %.04 = select i1 %.not12, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %10, float noundef 0xBE802DFB40000000) #139
  %16 = call float @llvm.fma.f32(float %.04, float %10, float noundef 0xBE802DFB40000000)
  %.05 = select i1 %.not13, float %16, float %15
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %17, 0
  %18 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %10, float noundef 0x3E8D1F08C0000000) #139
  %19 = call float @llvm.fma.f32(float %.05, float %10, float noundef 0x3E8D1F08C0000000)
  %.07 = select i1 %.not14, float %19, float %18
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %10, float noundef 0x3EE0E85540000000) #139
  %22 = call float @llvm.fma.f32(float %.07, float %10, float noundef 0x3EE0E85540000000)
  %.08 = select i1 %.not15, float %22, float %21
  %23 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %23, 0
  %24 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %10, float noundef 0xBF009B69A0000000) #139
  %25 = call float @llvm.fma.f32(float %.08, float %10, float noundef 0xBF009B69A0000000)
  %.09 = select i1 %.not16, float %25, float %24
  %26 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %26, 0
  %27 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %10, float noundef 0xBF495FD420000000) #139
  %28 = call float @llvm.fma.f32(float %.09, float %10, float noundef 0xBF495FD420000000)
  %.010 = select i1 %.not17, float %28, float %27
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %29, 0
  %30 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %10, float noundef 0x3F76B804E0000000) #139
  %31 = call float @llvm.fma.f32(float %.010, float %10, float noundef 0x3F76B804E0000000)
  %.011 = select i1 %.not18, float %31, float %30
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %10, float noundef 0x3FC495C1E0000000) #139
  %34 = call float @llvm.fma.f32(float %.011, float %10, float noundef 0x3FC495C1E0000000)
  %.012 = select i1 %.not19, float %34, float %33
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %10, float noundef 0x3FEC5BF880000000) #139
  %37 = call float @llvm.fma.f32(float %.012, float %10, float noundef 0x3FEC5BF880000000)
  %.014 = select i1 %.not20, float %37, float %36
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %38, 0
  %39 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %0, float %.014) #139
  %40 = call float @llvm.fma.f32(float %.014, float %0, float %.014)
  %.015 = select i1 %.not21, float %40, float %39
  br label %79

41:                                               ; preds = %__nv_fadd_rn.exit
  %42 = fcmp ogt float %a, 1.000000e+00
  %p.0 = select i1 %42, float %.01, float %a
  %43 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %43, 0
  br i1 %.not1, label %46, label %44

44:                                               ; preds = %41
  %45 = call float @llvm.nvvm.lg2.approx.ftz.f(float %p.0) #139
  br label %__nv_fast_log2f.exit

46:                                               ; preds = %41
  %47 = call float @llvm.nvvm.lg2.approx.f(float %p.0) #139
  br label %__nv_fast_log2f.exit

__nv_fast_log2f.exit:                             ; preds = %46, %44
  %.016 = phi float [ %45, %44 ], [ %47, %46 ]
  %48 = fneg float %.016
  %49 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %48) #139
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC04F8E7E20000000, float %49, float noundef 0x405FDF0520000000) #139
  %52 = call float @llvm.fma.f32(float %49, float noundef 0xC04F8E7E20000000, float noundef 0x405FDF0520000000)
  %.017 = select i1 %.not2, float %52, float %51
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %49, float noundef 0xC05C86C380000000) #139
  %55 = call float @llvm.fma.f32(float %.017, float %49, float noundef 0xC05C86C380000000)
  %.018 = select i1 %.not3, float %55, float %54
  %56 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %56, 0
  %57 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %49, float noundef 0x404E29B360000000) #139
  %58 = call float @llvm.fma.f32(float %.018, float %49, float noundef 0x404E29B360000000)
  %.019 = select i1 %.not4, float %58, float %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %49, float noundef 0xC035CA3660000000) #139
  %61 = call float @llvm.fma.f32(float %.019, float %49, float noundef 0xC035CA3660000000)
  %.021 = select i1 %.not5, float %61, float %60
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %49, float noundef 0x4019DEA080000000) #139
  %64 = call float @llvm.fma.f32(float %.021, float %49, float noundef 0x4019DEA080000000)
  %.020 = select i1 %.not6, float %64, float %63
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %49, float noundef 0xBFFD53C0A0000000) #139
  %67 = call float @llvm.fma.f32(float %.020, float %49, float noundef 0xBFFD53C0A0000000)
  %.013 = select i1 %.not7, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %49, float noundef 0xBF9F0E3E80000000) #139
  %70 = call float @llvm.fma.f32(float %.013, float %49, float noundef 0xBF9F0E3E80000000)
  %.06 = select i1 %.not8, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %49, float noundef 0x3FEAA6EEA0000000) #139
  %73 = call float @llvm.fma.f32(float %.06, float %49, float noundef 0x3FEAA6EEA0000000)
  %.0 = select i1 %.not9, float %73, float %72
  %74 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %49) #141, !srcloc !73
  %75 = fneg float %74
  %t.0 = select i1 %42, float %75, float %74
  %76 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %76, 0
  %77 = call float @llvm.nvvm.mul.rn.ftz.f(float %.0, float %t.0) #139
  %78 = fmul float %t.0, %.0
  %.03 = select i1 %.not10, float %78, float %77
  br label %79

79:                                               ; preds = %__nv_fast_log2f.exit, %__nv_fmul_rn.exit
  %t.1 = phi float [ %.015, %__nv_fmul_rn.exit ], [ %.03, %__nv_fast_log2f.exit ]
  ret float %t.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_normcdfinvf(float %a) local_unnamed_addr #54 {
__nv_fadd_rn.exit.i:
  %0 = fadd float %a, %a
  %1 = fneg float %0
  %2 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %2, 0
  %3 = call float @llvm.nvvm.add.rn.ftz.f(float noundef 2.000000e+00, float %1) #139
  %4 = fsub float 2.000000e+00, %0
  %.01 = select i1 %.not, float %4, float %3
  %5 = fcmp oge float %0, 0x3F6BDA5120000000
  %6 = fcmp ole float %0, 0x3FFFF212E0000000
  %or.cond = and i1 %5, %6
  br i1 %or.cond, label %__nv_fmul_rn.exit.i, label %42

__nv_fmul_rn.exit.i:                              ; preds = %__nv_fadd_rn.exit.i
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %7, 0
  %8 = call float @llvm.nvvm.mul.rn.ftz.f(float %.01, float %0) #139
  %9 = fmul float %0, %.01
  %.03 = select i1 %.not12, float %9, float %8
  %10 = call float @llvm.nvvm.lg2.approx.ftz.f(float %.03) #139
  %11 = fneg float %10
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBDF14C6E00000000, float %11, float noundef 0x3E443EC8A0000000) #139
  %14 = call float @llvm.fma.f32(float %11, float noundef 0xBDF14C6E00000000, float noundef 0x3E443EC8A0000000)
  %.05 = select i1 %.not13, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %11, float noundef 0xBE802DFB40000000) #139
  %17 = call float @llvm.fma.f32(float %.05, float %11, float noundef 0xBE802DFB40000000)
  %.06 = select i1 %.not14, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %11, float noundef 0x3E8D1F08C0000000) #139
  %20 = call float @llvm.fma.f32(float %.06, float %11, float noundef 0x3E8D1F08C0000000)
  %.07 = select i1 %.not15, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %11, float noundef 0x3EE0E85540000000) #139
  %23 = call float @llvm.fma.f32(float %.07, float %11, float noundef 0x3EE0E85540000000)
  %.09 = select i1 %.not16, float %23, float %22
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %11, float noundef 0xBF009B69A0000000) #139
  %26 = call float @llvm.fma.f32(float %.09, float %11, float noundef 0xBF009B69A0000000)
  %.010 = select i1 %.not17, float %26, float %25
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float %11, float noundef 0xBF495FD420000000) #139
  %29 = call float @llvm.fma.f32(float %.010, float %11, float noundef 0xBF495FD420000000)
  %.011 = select i1 %.not18, float %29, float %28
  %30 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %30, 0
  %31 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %11, float noundef 0x3F76B804E0000000) #139
  %32 = call float @llvm.fma.f32(float %.011, float %11, float noundef 0x3F76B804E0000000)
  %.012 = select i1 %.not19, float %32, float %31
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %11, float noundef 0x3FC495C1E0000000) #139
  %35 = call float @llvm.fma.f32(float %.012, float %11, float noundef 0x3FC495C1E0000000)
  %.013 = select i1 %.not20, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %11, float noundef 0x3FEC5BF880000000) #139
  %38 = call float @llvm.fma.f32(float %.013, float %11, float noundef 0x3FEC5BF880000000)
  %.014 = select i1 %.not21, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %1, float %.014) #139
  %41 = call float @llvm.fma.f32(float %.014, float %1, float %.014)
  %.016 = select i1 %.not22, float %41, float %40
  br label %__nv_erfcinvf.exit

42:                                               ; preds = %__nv_fadd_rn.exit.i
  %43 = fcmp ogt float %0, 1.000000e+00
  %p.i.0 = select i1 %43, float %.01, float %0
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %44, 0
  br i1 %.not1, label %47, label %45

45:                                               ; preds = %42
  %46 = call float @llvm.nvvm.lg2.approx.ftz.f(float %p.i.0) #139
  br label %__nv_fast_log2f.exit.i

47:                                               ; preds = %42
  %48 = call float @llvm.nvvm.lg2.approx.f(float %p.i.0) #139
  br label %__nv_fast_log2f.exit.i

__nv_fast_log2f.exit.i:                           ; preds = %47, %45
  %.017 = phi float [ %46, %45 ], [ %48, %47 ]
  %49 = fneg float %.017
  %50 = call float @llvm.nvvm.rsqrt.approx.ftz.f(float %49) #139
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC04F8E7E20000000, float %50, float noundef 0x405FDF0520000000) #139
  %53 = call float @llvm.fma.f32(float %50, float noundef 0xC04F8E7E20000000, float noundef 0x405FDF0520000000)
  %.018 = select i1 %.not2, float %53, float %52
  %54 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %54, 0
  %55 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %50, float noundef 0xC05C86C380000000) #139
  %56 = call float @llvm.fma.f32(float %.018, float %50, float noundef 0xC05C86C380000000)
  %.019 = select i1 %.not3, float %56, float %55
  %57 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %57, 0
  %58 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %50, float noundef 0x404E29B360000000) #139
  %59 = call float @llvm.fma.f32(float %.019, float %50, float noundef 0x404E29B360000000)
  %.020 = select i1 %.not4, float %59, float %58
  %60 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %60, 0
  %61 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %50, float noundef 0xC035CA3660000000) #139
  %62 = call float @llvm.fma.f32(float %.020, float %50, float noundef 0xC035CA3660000000)
  %.021 = select i1 %.not5, float %62, float %61
  %63 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %63, 0
  %64 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %50, float noundef 0x4019DEA080000000) #139
  %65 = call float @llvm.fma.f32(float %.021, float %50, float noundef 0x4019DEA080000000)
  %.022 = select i1 %.not6, float %65, float %64
  %66 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %66, 0
  %67 = call float @llvm.nvvm.fma.rn.ftz.f(float %.022, float %50, float noundef 0xBFFD53C0A0000000) #139
  %68 = call float @llvm.fma.f32(float %.022, float %50, float noundef 0xBFFD53C0A0000000)
  %.015 = select i1 %.not7, float %68, float %67
  %69 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %69, 0
  %70 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %50, float noundef 0xBF9F0E3E80000000) #139
  %71 = call float @llvm.fma.f32(float %.015, float %50, float noundef 0xBF9F0E3E80000000)
  %.08 = select i1 %.not8, float %71, float %70
  %72 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %72, 0
  %73 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %50, float noundef 0x3FEAA6EEA0000000) #139
  %74 = call float @llvm.fma.f32(float %.08, float %50, float noundef 0x3FEAA6EEA0000000)
  %.02 = select i1 %.not9, float %74, float %73
  %75 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %50) #141, !srcloc !73
  %76 = fneg float %75
  %t.i.0 = select i1 %43, float %76, float %75
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %77, 0
  %78 = call float @llvm.nvvm.mul.rn.ftz.f(float %.02, float %t.i.0) #139
  %79 = fmul float %t.i.0, %.02
  %.04 = select i1 %.not10, float %79, float %78
  br label %__nv_erfcinvf.exit

__nv_erfcinvf.exit:                               ; preds = %__nv_fast_log2f.exit.i, %__nv_fmul_rn.exit.i
  %t.i.1 = phi float [ %.016, %__nv_fmul_rn.exit.i ], [ %.04, %__nv_fast_log2f.exit.i ]
  %80 = fmul float %t.i.1, 0xBFF6A09E60000000
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %81, 0
  %82 = call float @llvm.nvvm.add.rn.ftz.f(float %80, float noundef 0.000000e+00) #139
  %83 = fadd float %80, 0.000000e+00
  %.0 = select i1 %.not11, float %83, float %82
  ret float %.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_normcdff(float %a) local_unnamed_addr #55 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.03 = select i1 %.not, float %2, float %1
  %3 = fcmp ogt float %.03, 1.450000e+01
  %4 = bitcast float %a to i32
  %5 = and i32 %4, -2147483648
  %6 = or i32 %5, 1097334784
  %7 = bitcast i32 %6 to float
  %.02 = select i1 %3, float %7, float %a
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.mul.rn.ftz.f(float %.02, float noundef 0xBFE6A09E60000000) #139
  %10 = fmul float %.02, 0xBFE6A09E60000000
  %.06 = select i1 %.not1, float %10, float %9
  %11 = fneg float %.06
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %12, 0
  %13 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float noundef 0xBFE6A09E60000000, float %11) #139
  %14 = call float @llvm.fma.f32(float %.02, float noundef 0xBFE6A09E60000000, float %11)
  %.07 = select i1 %.not2, float %14, float %13
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %15, 0
  %16 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float noundef 0xBE49FCEF40000000, float %.07) #139
  %17 = call float @llvm.fma.f32(float %.02, float noundef 0xBE49FCEF40000000, float %.07)
  %.09 = select i1 %.not3, float %17, float %16
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %18, 0
  %19 = call float @llvm.nvvm.add.rn.ftz.f(float %.06, float %.09) #139
  %20 = fadd float %.06, %.09
  %.010 = select i1 %.not4, float %20, float %19
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fabs.ftz.f(float %.010) #139
  %23 = call float @llvm.fabs.f32(float %.010)
  %.011 = select i1 %.not5, float %23, float %22
  %24 = fadd float %.011, -4.000000e+00
  %25 = fadd float %.011, 4.000000e+00
  %26 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %25) #141, !srcloc !73
  %27 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %27, 0
  %28 = call float @llvm.nvvm.mul.rn.ftz.f(float %24, float %26) #139
  %29 = fmul float %26, %24
  %.013 = select i1 %.not6, float %29, float %28
  %30 = fadd float %.013, 1.000000e+00
  %31 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %31, 0
  %32 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef -4.000000e+00, float %30, float %.011) #139
  %33 = call float @llvm.fma.f32(float %30, float noundef -4.000000e+00, float %.011)
  %.014 = select i1 %.not7, float %33, float %32
  %34 = fneg float %.013
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %34, float %.011, float %.014) #139
  %37 = call float @llvm.fma.f32(float %34, float %.011, float %.014)
  %.015 = select i1 %.not8, float %37, float %36
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %38, 0
  %39 = call float @llvm.nvvm.fma.rn.ftz.f(float %26, float %.015, float %.013) #139
  %40 = call float @llvm.fma.f32(float %26, float %.015, float %.013)
  %.016 = select i1 %.not9, float %40, float %39
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %41, 0
  %42 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F4D341220000000, float %.016, float noundef 0x3F7CDC0B60000000) #139
  %43 = call float @llvm.fma.f32(float %.016, float noundef 0x3F4D341220000000, float noundef 0x3F7CDC0B60000000)
  %.017 = select i1 %.not10, float %43, float %42
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %.016, float noundef 0xBF903F6960000000) #139
  %46 = call float @llvm.fma.f32(float %.017, float %.016, float noundef 0xBF903F6960000000)
  %.018 = select i1 %.not11, float %46, float %45
  %47 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %47, 0
  %48 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %.016, float noundef 0x3FA2A6E760000000) #139
  %49 = call float @llvm.fma.f32(float %.018, float %.016, float noundef 0x3FA2A6E760000000)
  %.019 = select i1 %.not12, float %49, float %48
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %.016, float noundef 0xBFB10F8B40000000) #139
  %52 = call float @llvm.fma.f32(float %.019, float %.016, float noundef 0xBFB10F8B40000000)
  %.020 = select i1 %.not13, float %52, float %51
  %53 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %53, 0
  %54 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %.016, float noundef 0x3FB8043AA0000000) #139
  %55 = call float @llvm.fma.f32(float %.020, float %.016, float noundef 0x3FB8043AA0000000)
  %.021 = select i1 %.not14, float %55, float %54
  %56 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %56, 0
  %57 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %.016, float noundef 0xBFB9DA8480000000) #139
  %58 = call float @llvm.fma.f32(float %.021, float %.016, float noundef 0xBFB9DA8480000000)
  %.022 = select i1 %.not15, float %58, float %57
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float %.022, float %.016, float noundef 0x3FB16E9BC0000000) #139
  %61 = call float @llvm.fma.f32(float %.022, float %.016, float noundef 0x3FB16E9BC0000000)
  %.024 = select i1 %.not16, float %61, float %60
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %.024, float %.016, float noundef 0x3F8F7E2E00000000) #139
  %64 = call float @llvm.fma.f32(float %.024, float %.016, float noundef 0x3F8F7E2E00000000)
  %.025 = select i1 %.not17, float %64, float %63
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %.025, float %.016, float noundef 0xBFC1DF1A80000000) #139
  %67 = call float @llvm.fma.f32(float %.025, float %.016, float noundef 0xBFC1DF1A80000000)
  %.026 = select i1 %.not18, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %.016, float noundef 0x3FF3BA5920000000) #139
  %70 = call float @llvm.fma.f32(float %.026, float %.016, float noundef 0x3FF3BA5920000000)
  %.027 = select i1 %.not19, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 2.000000e+00, float %.011, float noundef 1.000000e+00) #139
  %73 = call float @llvm.fma.f32(float %.011, float noundef 2.000000e+00, float noundef 1.000000e+00)
  %.028 = select i1 %.not20, float %73, float %72
  %74 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.028) #141, !srcloc !73
  %75 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %75, 0
  %76 = call float @llvm.nvvm.mul.rn.ftz.f(float %.027, float %74) #139
  %77 = fmul float %74, %.027
  %.029 = select i1 %.not21, float %77, float %76
  %78 = fmul float %.029, -2.000000e+00
  %79 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %79, 0
  %80 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %78, float %.027) #139
  %81 = call float @llvm.fma.f32(float %.011, float %78, float %.027)
  %.030 = select i1 %.not22, float %81, float %80
  %82 = fsub float %.030, %.029
  %83 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %83, 0
  %84 = call float @llvm.nvvm.fma.rn.ftz.f(float %82, float %74, float %.029) #139
  %85 = call float @llvm.fma.f32(float %82, float %74, float %.029)
  %.031 = select i1 %.not23, float %85, float %84
  %86 = fneg float %.011
  %87 = fmul float %.011, %86
  %88 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %88, 0
  %89 = call float @llvm.nvvm.mul.rn.ftz.f(float %87, float noundef 0x3FF7154760000000) #139
  %90 = fmul float %87, 0x3FF7154760000000
  %.08 = select i1 %.not24, float %90, float %89
  %91 = call float @llvm.trunc.f32(float %.08)
  %92 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %92, 0
  %93 = call float @llvm.nvvm.fabs.ftz.f(float %91) #139
  %94 = call float @llvm.fabs.f32(float %91)
  %.012 = select i1 %.not25, float %94, float %93
  %95 = fcmp ogt float %.012, 1.260000e+02
  %96 = bitcast float %91 to i32
  %97 = and i32 %96, -2147483648
  %98 = or i32 %97, 1123811328
  %99 = bitcast i32 %98 to float
  %j.i.i.0 = select i1 %95, float %99, float %91
  %100 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not26 = icmp eq i32 %100, 0
  %101 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.i.0, float noundef 0xBFE62E4300000000, float %87) #139
  %102 = call float @llvm.fma.f32(float %j.i.i.0, float noundef 0xBFE62E4300000000, float %87)
  %.023 = select i1 %.not26, float %102, float %101
  %103 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %103, 0
  %104 = call float @llvm.nvvm.fma.rn.ftz.f(float %j.i.i.0, float noundef 0x3E205C6100000000, float %.023) #139
  %105 = call float @llvm.fma.f32(float %j.i.i.0, float noundef 0x3E205C6100000000, float %.023)
  %.032 = select i1 %.not27, float %105, float %104
  %106 = fmul float %.032, 0x3FF7154760000000
  %107 = fadd float %j.i.i.0, 0x4168000FE0000000
  %108 = bitcast float %107 to i32
  %109 = shl i32 %108, 23
  %110 = bitcast i32 %109 to float
  %111 = call float @llvm.nvvm.ex2.approx.ftz.f(float %106) #139
  %112 = fmul float %111, %110
  %113 = fneg float %87
  %114 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %114, 0
  %115 = call float @llvm.nvvm.fma.rn.ftz.f(float %86, float %.011, float %113) #139
  %116 = call float @llvm.fma.f32(float %86, float %.011, float %113)
  %.04 = select i1 %.not28, float %116, float %115
  %117 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not29 = icmp eq i32 %117, 0
  %118 = call float @llvm.nvvm.fma.rn.ftz.f(float %112, float %.04, float %112) #139
  %119 = call float @llvm.fma.f32(float %112, float %.04, float %112)
  %.01 = select i1 %.not29, float %119, float %118
  %120 = fmul float %.01, %.031
  %121 = fcmp ogt float %.011, 0x40241C2900000000
  %t1.i.0 = select i1 %121, float 0.000000e+00, float %120
  %122 = fcmp olt float %.010, 0.000000e+00
  %123 = fsub float 2.000000e+00, %t1.i.0
  %124 = select i1 %122, float %123, float %t1.i.0
  %125 = fcmp olt float %.02, -1.000000e+00
  br i1 %125, label %__nv_fadd_rn.exit2, label %135

__nv_fadd_rn.exit2:                               ; preds = %__nv_fabsf.exit
  %126 = fsub float %.06, %.010
  %127 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %127, 0
  %128 = call float @llvm.nvvm.add.rn.ftz.f(float %126, float %.09) #139
  %129 = fadd float %.09, %126
  %.0 = select i1 %.not30, float %129, float %128
  %130 = fmul float %.010, -2.000000e+00
  %131 = fmul float %130, %124
  %132 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %132, 0
  %133 = call float @llvm.nvvm.fma.rn.ftz.f(float %131, float %.0, float %124) #139
  %134 = call float @llvm.fma.f32(float %131, float %.0, float %124)
  %.05 = select i1 %.not31, float %134, float %133
  br label %135

135:                                              ; preds = %__nv_fadd_rn.exit2, %__nv_fabsf.exit
  %z.0 = phi float [ %.05, %__nv_fadd_rn.exit2 ], [ %124, %__nv_fabsf.exit ]
  %136 = fmul float %z.0, 5.000000e-01
  ret float %136
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_lgammaf(float %a) local_unnamed_addr #56 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.02 = select i1 %.not, float %2, float %1
  %3 = fcmp ult float %.02, 3.000000e+00
  br i1 %3, label %103, label %4

4:                                                ; preds = %__nv_fabsf.exit
  %5 = fcmp ult float %.02, 0x401F333340000000
  br i1 %5, label %__internal_fmad.exit5.i, label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %4
  %6 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.02) #141, !srcloc !73
  %7 = fmul float %6, %6
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not88 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F497CEAA0000000, float %7, float noundef 0xBF66C12A60000000) #139
  %10 = call float @llvm.fma.f32(float %7, float noundef 0x3F497CEAA0000000, float noundef 0xBF66C12A60000000)
  %.05 = select i1 %.not88, float %10, float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not89 = icmp eq i32 %11, 0
  %12 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %7, float noundef 0x3FB5555460000000) #139
  %13 = call float @llvm.fma.f32(float %.05, float %7, float noundef 0x3FB5555460000000)
  %.06 = select i1 %.not89, float %13, float %12
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not90 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.fma.rn.ftz.f(float %.06, float %6, float noundef 0x3FED67F1C0000000) #139
  %16 = call float @llvm.fma.f32(float %.06, float %6, float noundef 0x3FED67F1C0000000)
  %.07 = select i1 %.not90, float %16, float %15
  %17 = fcmp olt float %.02, 0x3810000000000000
  %18 = fmul float %.02, 0x4160000000000000
  %.010 = select i1 %17, float %18, float %.02
  %i.i.i.0 = select i1 %17, float -2.300000e+01, float 0.000000e+00
  %19 = bitcast float %.010 to i32
  %20 = add i32 %19, -1059760811
  %21 = and i32 %20, -8388608
  %22 = sub i32 %19, %21
  %23 = bitcast i32 %22 to float
  %24 = sitofp i32 %21 to float
  %25 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not91 = icmp eq i32 %25, 0
  %26 = call float @llvm.nvvm.fma.rn.ftz.f(float %24, float noundef 0x3E80000000000000, float noundef %i.i.i.0) #139
  %27 = call float @llvm.fma.f32(float %24, float noundef 0x3E80000000000000, float noundef %i.i.i.0)
  %.016 = select i1 %.not91, float %27, float %26
  %28 = fadd float %23, -1.000000e+00
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not92 = icmp eq i32 %29, 0
  %30 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %28, float noundef 0x3FC2073EC0000000) #139
  %31 = call float @llvm.fma.f32(float %28, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.018 = select i1 %.not92, float %31, float %30
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not93 = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fma.rn.ftz.f(float %.018, float %28, float noundef 0xBFBF19B980000000) #139
  %34 = call float @llvm.fma.f32(float %.018, float %28, float noundef 0xBFBF19B980000000)
  %.019 = select i1 %.not93, float %34, float %33
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not94 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %28, float noundef 0x3FC1E52AA0000000) #139
  %37 = call float @llvm.fma.f32(float %.019, float %28, float noundef 0x3FC1E52AA0000000)
  %.020 = select i1 %.not94, float %37, float %36
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not95 = icmp eq i32 %38, 0
  %39 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %28, float noundef 0xBFC55B1720000000) #139
  %40 = call float @llvm.fma.f32(float %.020, float %28, float noundef 0xBFC55B1720000000)
  %.022 = select i1 %.not95, float %40, float %39
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not96 = icmp eq i32 %41, 0
  %42 = call float @llvm.nvvm.fma.rn.ftz.f(float %.022, float %28, float noundef 0x3FC99DA160000000) #139
  %43 = call float @llvm.fma.f32(float %.022, float %28, float noundef 0x3FC99DA160000000)
  %.023 = select i1 %.not96, float %43, float %42
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not97 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float %28, float noundef 0xBFCFFFE440000000) #139
  %46 = call float @llvm.fma.f32(float %.023, float %28, float noundef 0xBFCFFFE440000000)
  %.024 = select i1 %.not97, float %46, float %45
  %47 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not98 = icmp eq i32 %47, 0
  %48 = call float @llvm.nvvm.fma.rn.ftz.f(float %.024, float %28, float noundef 0x3FD5554F00000000) #139
  %49 = call float @llvm.fma.f32(float %.024, float %28, float noundef 0x3FD5554F00000000)
  %.026 = select i1 %.not98, float %49, float %48
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not99 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %28, float noundef -5.000000e-01) #139
  %52 = call float @llvm.fma.f32(float %.026, float %28, float noundef -5.000000e-01)
  %.015 = select i1 %.not99, float %52, float %51
  %53 = fmul float %28, %.015
  %54 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not100 = icmp eq i32 %54, 0
  %55 = call float @llvm.nvvm.fma.rn.ftz.f(float %53, float %28, float %28) #139
  %56 = call float @llvm.fma.f32(float %53, float %28, float %28)
  %.014 = select i1 %.not100, float %56, float %55
  %57 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not101 = icmp eq i32 %57, 0
  %58 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float noundef 0x3FE62E4300000000, float %.014) #139
  %59 = call float @llvm.fma.f32(float %.016, float noundef 0x3FE62E4300000000, float %.014)
  %.012 = select i1 %.not101, float %59, float %58
  %60 = icmp ugt i32 %19, 2139095039
  br i1 %60, label %__nv_fmaf_rn.exit.i.i, label %__internal_accurate_logf.exit.i

__nv_fmaf_rn.exit.i.i:                            ; preds = %__internal_fmad.exit.i
  %61 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not104 = icmp eq i32 %61, 0
  %62 = call float @llvm.nvvm.fma.rn.ftz.f(float %.010, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %63 = call float @llvm.fma.f32(float %.010, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.011 = select i1 %.not104, float %63, float %62
  br label %__internal_accurate_logf.exit.i

__internal_accurate_logf.exit.i:                  ; preds = %__nv_fmaf_rn.exit.i.i, %__internal_fmad.exit.i
  %r.i3.i.0 = phi float [ %.011, %__nv_fmaf_rn.exit.i.i ], [ %.012, %__internal_fmad.exit.i ]
  %64 = fcmp oeq float %.010, 0.000000e+00
  %r.i3.i.0.op = fmul float %r.i3.i.0, 5.000000e-01
  %65 = select i1 %64, float 0xFFF0000000000000, float %r.i3.i.0.op
  %66 = fadd float %.02, -5.000000e-01
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not102 = icmp eq i32 %67, 0
  %68 = call float @llvm.nvvm.mul.rn.ftz.f(float %65, float %66) #139
  %69 = fmul float %66, %65
  %.027 = select i1 %.not102, float %69, float %68
  %70 = fsub float %.027, %.02
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not103 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.add.rn.ftz.f(float %.027, float %.07) #139
  %73 = fadd float %.07, %.027
  %.028 = select i1 %.not103, float %73, float %72
  %74 = fadd float %70, %.028
  %75 = fcmp oeq float %.02, 0x7FF0000000000000
  %t.i.0 = select i1 %75, float %.02, float %74
  br label %__internal_lgammaf_pos.exit

__internal_fmad.exit5.i:                          ; preds = %4
  %76 = fadd float %.02, -3.000000e+00
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not80 = icmp eq i32 %77, 0
  %78 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xC087671F60000000, float %76, float noundef 0xC0C81EDF00000000) #139
  %79 = call float @llvm.fma.f32(float %76, float noundef 0xC087671F60000000, float noundef 0xC0C81EDF00000000)
  %.030 = select i1 %.not80, float %79, float %78
  %80 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not81 = icmp eq i32 %80, 0
  %81 = call float @llvm.nvvm.fma.rn.ftz.f(float %.030, float %76, float noundef 0xC0E40CAC00000000) #139
  %82 = call float @llvm.fma.f32(float %.030, float %76, float noundef 0xC0E40CAC00000000)
  %.031 = select i1 %.not81, float %82, float %81
  %83 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not82 = icmp eq i32 %83, 0
  %84 = call float @llvm.nvvm.fma.rn.ftz.f(float %.031, float %76, float noundef 0xC0E796D540000000) #139
  %85 = call float @llvm.fma.f32(float %.031, float %76, float noundef 0xC0E796D540000000)
  %.032 = select i1 %.not82, float %85, float %84
  %86 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not83 = icmp eq i32 %86, 0
  %87 = call float @llvm.nvvm.fma.rn.ftz.f(float %.032, float %76, float noundef 0xC10175CB40000000) #139
  %88 = call float @llvm.fma.f32(float %.032, float %76, float noundef 0xC10175CB40000000)
  %.033 = select i1 %.not83, float %88, float %87
  %89 = fadd float %76, 0xC070340400000000
  %90 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not84 = icmp eq i32 %90, 0
  %91 = call float @llvm.nvvm.fma.rn.ftz.f(float %89, float %76, float noundef 0xC0C50C9700000000) #139
  %92 = call float @llvm.fma.f32(float %89, float %76, float noundef 0xC0C50C9700000000)
  %.035 = select i1 %.not84, float %92, float %91
  %93 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not85 = icmp eq i32 %93, 0
  %94 = call float @llvm.nvvm.fma.rn.ftz.f(float %.035, float %76, float noundef 0xC0F6A0D0C0000000) #139
  %95 = call float @llvm.fma.f32(float %.035, float %76, float noundef 0xC0F6A0D0C0000000)
  %.036 = select i1 %.not85, float %95, float %94
  %96 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not86 = icmp eq i32 %96, 0
  %97 = call float @llvm.nvvm.fma.rn.ftz.f(float %.036, float %76, float noundef 0xC109308CA0000000) #139
  %98 = call float @llvm.fma.f32(float %.036, float %76, float noundef 0xC109308CA0000000)
  %.037 = select i1 %.not86, float %98, float %97
  %99 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.037) #141, !srcloc !73
  %100 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not87 = icmp eq i32 %100, 0
  %101 = call float @llvm.nvvm.fma.rn.ftz.f(float %.033, float %99, float %76) #139
  %102 = call float @llvm.fma.f32(float %.033, float %99, float %76)
  %.038 = select i1 %.not87, float %102, float %101
  br label %__internal_lgammaf_pos.exit

103:                                              ; preds = %__nv_fabsf.exit
  %104 = fcmp ult float %.02, 1.500000e+00
  br i1 %104, label %134, label %__internal_fmad.exit22.i

__internal_fmad.exit22.i:                         ; preds = %103
  %105 = fadd float %.02, -2.000000e+00
  %106 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not71 = icmp eq i32 %106, 0
  %107 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F0A00FF40000000, float %105, float noundef 0xBF2CF40040000000) #139
  %108 = call float @llvm.fma.f32(float %105, float noundef 0x3F0A00FF40000000, float noundef 0xBF2CF40040000000)
  %.039 = select i1 %.not71, float %108, float %107
  %109 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not72 = icmp eq i32 %109, 0
  %110 = call float @llvm.nvvm.fma.rn.ftz.f(float %.039, float %105, float noundef 0x3F41BCDF80000000) #139
  %111 = call float @llvm.fma.f32(float %.039, float %105, float noundef 0x3F41BCDF80000000)
  %.041 = select i1 %.not72, float %111, float %110
  %112 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not73 = icmp eq i32 %112, 0
  %113 = call float @llvm.nvvm.fma.rn.ftz.f(float %.041, float %105, float noundef 0xBF53BC1C40000000) #139
  %114 = call float @llvm.fma.f32(float %.041, float %105, float noundef 0xBF53BC1C40000000)
  %.042 = select i1 %.not73, float %114, float %113
  %115 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not74 = icmp eq i32 %115, 0
  %116 = call float @llvm.nvvm.fma.rn.ftz.f(float %.042, float %105, float noundef 0x3F67A0B6E0000000) #139
  %117 = call float @llvm.fma.f32(float %.042, float %105, float noundef 0x3F67A0B6E0000000)
  %.043 = select i1 %.not74, float %117, float %116
  %118 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not75 = icmp eq i32 %118, 0
  %119 = call float @llvm.nvvm.fma.rn.ftz.f(float %.043, float %105, float noundef 0xBF7E3D6200000000) #139
  %120 = call float @llvm.fma.f32(float %.043, float %105, float noundef 0xBF7E3D6200000000)
  %.044 = select i1 %.not75, float %120, float %119
  %121 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not76 = icmp eq i32 %121, 0
  %122 = call float @llvm.nvvm.fma.rn.ftz.f(float %.044, float %105, float noundef 0x3F95134500000000) #139
  %123 = call float @llvm.fma.f32(float %.044, float %105, float noundef 0x3F95134500000000)
  %.045 = select i1 %.not76, float %123, float %122
  %124 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not77 = icmp eq i32 %124, 0
  %125 = call float @llvm.nvvm.fma.rn.ftz.f(float %.045, float %105, float noundef 0xBFB13E0340000000) #139
  %126 = call float @llvm.fma.f32(float %.045, float %105, float noundef 0xBFB13E0340000000)
  %.046 = select i1 %.not77, float %126, float %125
  %127 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not78 = icmp eq i32 %127, 0
  %128 = call float @llvm.nvvm.fma.rn.ftz.f(float %.046, float %105, float noundef 0x3FD4A34CC0000000) #139
  %129 = call float @llvm.fma.f32(float %.046, float %105, float noundef 0x3FD4A34CC0000000)
  %.047 = select i1 %.not78, float %129, float %128
  %130 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not79 = icmp eq i32 %130, 0
  %131 = call float @llvm.nvvm.fma.rn.ftz.f(float %.047, float %105, float noundef 0x3FDB0EE600000000) #139
  %132 = call float @llvm.fma.f32(float %.047, float %105, float noundef 0x3FDB0EE600000000)
  %.048 = select i1 %.not79, float %132, float %131
  %133 = fmul float %105, %.048
  br label %__internal_lgammaf_pos.exit

134:                                              ; preds = %103
  %135 = fcmp ult float %.02, 0x3FE6666660000000
  br i1 %135, label %__internal_fmad.exit60.i, label %__internal_fmad.exit40.i

__internal_fmad.exit40.i:                         ; preds = %134
  %136 = fsub float 1.000000e+00, %.02
  %137 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not61 = icmp eq i32 %137, 0
  %138 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FA77DEEC0000000, float %136, float noundef 0x3FBA8EAEE0000000) #139
  %139 = call float @llvm.fma.f32(float %136, float noundef 0x3FA77DEEC0000000, float noundef 0x3FBA8EAEE0000000)
  %.049 = select i1 %.not61, float %139, float %138
  %140 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not62 = icmp eq i32 %140, 0
  %141 = call float @llvm.nvvm.fma.rn.ftz.f(float %.049, float %136, float noundef 0x3FBF700F20000000) #139
  %142 = call float @llvm.fma.f32(float %.049, float %136, float noundef 0x3FBF700F20000000)
  %.050 = select i1 %.not62, float %142, float %141
  %143 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not63 = icmp eq i32 %143, 0
  %144 = call float @llvm.nvvm.fma.rn.ftz.f(float %.050, float %136, float noundef 0x3FC052B6A0000000) #139
  %145 = call float @llvm.fma.f32(float %.050, float %136, float noundef 0x3FC052B6A0000000)
  %.051 = select i1 %.not63, float %145, float %144
  %146 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not64 = icmp eq i32 %146, 0
  %147 = call float @llvm.nvvm.fma.rn.ftz.f(float %.051, float %136, float noundef 0x3FC254ECA0000000) #139
  %148 = call float @llvm.fma.f32(float %.051, float %136, float noundef 0x3FC254ECA0000000)
  %.053 = select i1 %.not64, float %148, float %147
  %149 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not65 = icmp eq i32 %149, 0
  %150 = call float @llvm.nvvm.fma.rn.ftz.f(float %.053, float %136, float noundef 0x3FC5AD0CE0000000) #139
  %151 = call float @llvm.fma.f32(float %.053, float %136, float noundef 0x3FC5AD0CE0000000)
  %.054 = select i1 %.not65, float %151, float %150
  %152 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not66 = icmp eq i32 %152, 0
  %153 = call float @llvm.nvvm.fma.rn.ftz.f(float %.054, float %136, float noundef 0x3FCA8C57E0000000) #139
  %154 = call float @llvm.fma.f32(float %.054, float %136, float noundef 0x3FCA8C57E0000000)
  %.055 = select i1 %.not66, float %154, float %153
  %155 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not67 = icmp eq i32 %155, 0
  %156 = call float @llvm.nvvm.fma.rn.ftz.f(float %.055, float %136, float noundef 0x3FD1514E40000000) #139
  %157 = call float @llvm.fma.f32(float %.055, float %136, float noundef 0x3FD1514E40000000)
  %.056 = select i1 %.not67, float %157, float %156
  %158 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not68 = icmp eq i32 %158, 0
  %159 = call float @llvm.nvvm.fma.rn.ftz.f(float %.056, float %136, float noundef 0x3FD9A4D480000000) #139
  %160 = call float @llvm.fma.f32(float %.056, float %136, float noundef 0x3FD9A4D480000000)
  %.057 = select i1 %.not68, float %160, float %159
  %161 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not69 = icmp eq i32 %161, 0
  %162 = call float @llvm.nvvm.fma.rn.ftz.f(float %.057, float %136, float noundef 0x3FEA51A640000000) #139
  %163 = call float @llvm.fma.f32(float %.057, float %136, float noundef 0x3FEA51A640000000)
  %.058 = select i1 %.not69, float %163, float %162
  %164 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not70 = icmp eq i32 %164, 0
  %165 = call float @llvm.nvvm.fma.rn.ftz.f(float %.058, float %136, float noundef 0x3FE2788D00000000) #139
  %166 = call float @llvm.fma.f32(float %.058, float %136, float noundef 0x3FE2788D00000000)
  %.059 = select i1 %.not70, float %166, float %165
  %167 = fmul float %136, %.059
  br label %__internal_lgammaf_pos.exit

__internal_fmad.exit60.i:                         ; preds = %134
  %168 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %168, 0
  %169 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F6D6390C0000000, float %.02, float noundef 0xBF76690F00000000) #139
  %170 = call float @llvm.fma.f32(float %.02, float noundef 0x3F6D6390C0000000, float noundef 0xBF76690F00000000)
  %.061 = select i1 %.not1, float %170, float %169
  %171 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %171, 0
  %172 = call float @llvm.nvvm.fma.rn.ftz.f(float %.061, float %.02, float noundef 0xBFA6D95DE0000000) #139
  %173 = call float @llvm.fma.f32(float %.061, float %.02, float noundef 0xBFA6D95DE0000000)
  %.063 = select i1 %.not2, float %173, float %172
  %174 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %174, 0
  %175 = call float @llvm.nvvm.fma.rn.ftz.f(float %.063, float %.02, float noundef 0x3FC56AAAA0000000) #139
  %176 = call float @llvm.fma.f32(float %.063, float %.02, float noundef 0x3FC56AAAA0000000)
  %.065 = select i1 %.not3, float %176, float %175
  %177 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %177, 0
  %178 = call float @llvm.nvvm.fma.rn.ftz.f(float %.065, float %.02, float noundef 0xBFA592D8E0000000) #139
  %179 = call float @llvm.fma.f32(float %.065, float %.02, float noundef 0xBFA592D8E0000000)
  %.066 = select i1 %.not4, float %179, float %178
  %180 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %180, 0
  %181 = call float @llvm.nvvm.fma.rn.ftz.f(float %.066, float %.02, float noundef 0xBFE4FCDD60000000) #139
  %182 = call float @llvm.fma.f32(float %.066, float %.02, float noundef 0xBFE4FCDD60000000)
  %.067 = select i1 %.not5, float %182, float %181
  %183 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %183, 0
  %184 = call float @llvm.nvvm.fma.rn.ftz.f(float %.067, float %.02, float noundef 0x3FE2788C60000000) #139
  %185 = call float @llvm.fma.f32(float %.067, float %.02, float noundef 0x3FE2788C60000000)
  %.068 = select i1 %.not6, float %185, float %184
  %186 = fmul float %.02, %.068
  %187 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %187, 0
  %188 = call float @llvm.nvvm.fma.rn.ftz.f(float %186, float %.02, float %.02) #139
  %189 = call float @llvm.fma.f32(float %186, float %.02, float %.02)
  %.069 = select i1 %.not7, float %189, float %188
  %190 = fcmp olt float %.069, 0x3810000000000000
  %191 = fmul float %.069, 0x4160000000000000
  %.071 = select i1 %190, float %191, float %.069
  %i.i82.i.0 = select i1 %190, float -2.300000e+01, float 0.000000e+00
  %192 = bitcast float %.071 to i32
  %193 = add i32 %192, -1059760811
  %194 = and i32 %193, -8388608
  %195 = sub i32 %192, %194
  %196 = bitcast i32 %195 to float
  %197 = sitofp i32 %194 to float
  %198 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %198, 0
  %199 = call float @llvm.nvvm.fma.rn.ftz.f(float %197, float noundef 0x3E80000000000000, float noundef %i.i82.i.0) #139
  %200 = call float @llvm.fma.f32(float %197, float noundef 0x3E80000000000000, float noundef %i.i82.i.0)
  %.077 = select i1 %.not8, float %200, float %199
  %201 = fadd float %196, -1.000000e+00
  %202 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %202, 0
  %203 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %201, float noundef 0x3FC2073EC0000000) #139
  %204 = call float @llvm.fma.f32(float %201, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.078 = select i1 %.not9, float %204, float %203
  %205 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %205, 0
  %206 = call float @llvm.nvvm.fma.rn.ftz.f(float %.078, float %201, float noundef 0xBFBF19B980000000) #139
  %207 = call float @llvm.fma.f32(float %.078, float %201, float noundef 0xBFBF19B980000000)
  %.079 = select i1 %.not10, float %207, float %206
  %208 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %208, 0
  %209 = call float @llvm.nvvm.fma.rn.ftz.f(float %.079, float %201, float noundef 0x3FC1E52AA0000000) #139
  %210 = call float @llvm.fma.f32(float %.079, float %201, float noundef 0x3FC1E52AA0000000)
  %.080 = select i1 %.not11, float %210, float %209
  %211 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %211, 0
  %212 = call float @llvm.nvvm.fma.rn.ftz.f(float %.080, float %201, float noundef 0xBFC55B1720000000) #139
  %213 = call float @llvm.fma.f32(float %.080, float %201, float noundef 0xBFC55B1720000000)
  %.081 = select i1 %.not12, float %213, float %212
  %214 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %214, 0
  %215 = call float @llvm.nvvm.fma.rn.ftz.f(float %.081, float %201, float noundef 0x3FC99DA160000000) #139
  %216 = call float @llvm.fma.f32(float %.081, float %201, float noundef 0x3FC99DA160000000)
  %.083 = select i1 %.not13, float %216, float %215
  %217 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %217, 0
  %218 = call float @llvm.nvvm.fma.rn.ftz.f(float %.083, float %201, float noundef 0xBFCFFFE440000000) #139
  %219 = call float @llvm.fma.f32(float %.083, float %201, float noundef 0xBFCFFFE440000000)
  %.084 = select i1 %.not14, float %219, float %218
  %220 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %220, 0
  %221 = call float @llvm.nvvm.fma.rn.ftz.f(float %.084, float %201, float noundef 0x3FD5554F00000000) #139
  %222 = call float @llvm.fma.f32(float %.084, float %201, float noundef 0x3FD5554F00000000)
  %.085 = select i1 %.not15, float %222, float %221
  %223 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %223, 0
  %224 = call float @llvm.nvvm.fma.rn.ftz.f(float %.085, float %201, float noundef -5.000000e-01) #139
  %225 = call float @llvm.fma.f32(float %.085, float %201, float noundef -5.000000e-01)
  %.076 = select i1 %.not16, float %225, float %224
  %226 = fmul float %201, %.076
  %227 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %227, 0
  %228 = call float @llvm.nvvm.fma.rn.ftz.f(float %226, float %201, float %201) #139
  %229 = call float @llvm.fma.f32(float %226, float %201, float %201)
  %.074 = select i1 %.not17, float %229, float %228
  %230 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %230, 0
  %231 = call float @llvm.nvvm.fma.rn.ftz.f(float %.077, float noundef 0x3FE62E4300000000, float %.074) #139
  %232 = call float @llvm.fma.f32(float %.077, float noundef 0x3FE62E4300000000, float %.074)
  %.073 = select i1 %.not18, float %232, float %231
  %233 = icmp ugt i32 %192, 2139095039
  br i1 %233, label %__nv_fmaf_rn.exit.i95.i, label %__internal_accurate_logf.exit96.i

__nv_fmaf_rn.exit.i95.i:                          ; preds = %__internal_fmad.exit60.i
  %234 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not60 = icmp eq i32 %234, 0
  %235 = call float @llvm.nvvm.fma.rn.ftz.f(float %.071, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %236 = call float @llvm.fma.f32(float %.071, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.072 = select i1 %.not60, float %236, float %235
  br label %__internal_accurate_logf.exit96.i

__internal_accurate_logf.exit96.i:                ; preds = %__nv_fmaf_rn.exit.i95.i, %__internal_fmad.exit60.i
  %r.i81.i.0 = phi float [ %.072, %__nv_fmaf_rn.exit.i95.i ], [ %.073, %__internal_fmad.exit60.i ]
  %237 = fcmp oeq float %.071, 0.000000e+00
  %r.i81.i.1 = select i1 %237, float 0xFFF0000000000000, float %r.i81.i.0
  %238 = fneg float %r.i81.i.1
  br label %__internal_lgammaf_pos.exit

__internal_lgammaf_pos.exit:                      ; preds = %__internal_fmad.exit22.i, %__internal_accurate_logf.exit96.i, %__internal_fmad.exit40.i, %__internal_accurate_logf.exit.i, %__internal_fmad.exit5.i
  %t.i.4 = phi float [ %t.i.0, %__internal_accurate_logf.exit.i ], [ %.038, %__internal_fmad.exit5.i ], [ %133, %__internal_fmad.exit22.i ], [ %167, %__internal_fmad.exit40.i ], [ %238, %__internal_accurate_logf.exit96.i ]
  %239 = fcmp ult float %a, 0.000000e+00
  br i1 %239, label %__nv_fabsf.exit1, label %388

__nv_fabsf.exit1:                                 ; preds = %__internal_lgammaf_pos.exit
  %240 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %240, 0
  %.086 = select i1 %.not19, float %2, float %1
  %241 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %241, 0
  %242 = call float @llvm.nvvm.floor.ftz.f(float %.086) #139
  %243 = call float @llvm.floor.f32(float %.086)
  %.087 = select i1 %.not20, float %243, float %242
  %244 = fcmp oeq float %.086, %.087
  br i1 %244, label %388, label %245

245:                                              ; preds = %__nv_fabsf.exit1
  %246 = fcmp olt float %.086, 0x3BFD83C940000000
  br i1 %246, label %__nv_fmaf_rn.exit10.i, label %__internal_fmad.exit

__nv_fmaf_rn.exit10.i:                            ; preds = %245
  %247 = fcmp olt float %.086, 0x3810000000000000
  %248 = fmul float %.086, 0x4160000000000000
  %.089 = select i1 %247, float %248, float %.086
  %i.i.0 = select i1 %247, float -2.300000e+01, float 0.000000e+00
  %249 = bitcast float %.089 to i32
  %250 = add i32 %249, -1059760811
  %251 = and i32 %250, -8388608
  %252 = sub i32 %249, %251
  %253 = bitcast i32 %252 to float
  %254 = sitofp i32 %251 to float
  %255 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %255, 0
  %256 = call float @llvm.nvvm.fma.rn.ftz.f(float %254, float noundef 0x3E80000000000000, float noundef %i.i.0) #139
  %257 = call float @llvm.fma.f32(float %254, float noundef 0x3E80000000000000, float noundef %i.i.0)
  %.095 = select i1 %.not48, float %257, float %256
  %258 = fadd float %253, -1.000000e+00
  %259 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %259, 0
  %260 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %258, float noundef 0x3FC2073EC0000000) #139
  %261 = call float @llvm.fma.f32(float %258, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.096 = select i1 %.not49, float %261, float %260
  %262 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %262, 0
  %263 = call float @llvm.nvvm.fma.rn.ftz.f(float %.096, float %258, float noundef 0xBFBF19B980000000) #139
  %264 = call float @llvm.fma.f32(float %.096, float %258, float noundef 0xBFBF19B980000000)
  %.097 = select i1 %.not50, float %264, float %263
  %265 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not51 = icmp eq i32 %265, 0
  %266 = call float @llvm.nvvm.fma.rn.ftz.f(float %.097, float %258, float noundef 0x3FC1E52AA0000000) #139
  %267 = call float @llvm.fma.f32(float %.097, float %258, float noundef 0x3FC1E52AA0000000)
  %.098 = select i1 %.not51, float %267, float %266
  %268 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not52 = icmp eq i32 %268, 0
  %269 = call float @llvm.nvvm.fma.rn.ftz.f(float %.098, float %258, float noundef 0xBFC55B1720000000) #139
  %270 = call float @llvm.fma.f32(float %.098, float %258, float noundef 0xBFC55B1720000000)
  %.099 = select i1 %.not52, float %270, float %269
  %271 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not53 = icmp eq i32 %271, 0
  %272 = call float @llvm.nvvm.fma.rn.ftz.f(float %.099, float %258, float noundef 0x3FC99DA160000000) #139
  %273 = call float @llvm.fma.f32(float %.099, float %258, float noundef 0x3FC99DA160000000)
  %.0101 = select i1 %.not53, float %273, float %272
  %274 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not54 = icmp eq i32 %274, 0
  %275 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0101, float %258, float noundef 0xBFCFFFE440000000) #139
  %276 = call float @llvm.fma.f32(float %.0101, float %258, float noundef 0xBFCFFFE440000000)
  %.0102 = select i1 %.not54, float %276, float %275
  %277 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not55 = icmp eq i32 %277, 0
  %278 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0102, float %258, float noundef 0x3FD5554F00000000) #139
  %279 = call float @llvm.fma.f32(float %.0102, float %258, float noundef 0x3FD5554F00000000)
  %.0103 = select i1 %.not55, float %279, float %278
  %280 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not56 = icmp eq i32 %280, 0
  %281 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0103, float %258, float noundef -5.000000e-01) #139
  %282 = call float @llvm.fma.f32(float %.0103, float %258, float noundef -5.000000e-01)
  %.094 = select i1 %.not56, float %282, float %281
  %283 = fmul float %258, %.094
  %284 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not57 = icmp eq i32 %284, 0
  %285 = call float @llvm.nvvm.fma.rn.ftz.f(float %283, float %258, float %258) #139
  %286 = call float @llvm.fma.f32(float %283, float %258, float %258)
  %.092 = select i1 %.not57, float %286, float %285
  %287 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not58 = icmp eq i32 %287, 0
  %288 = call float @llvm.nvvm.fma.rn.ftz.f(float %.095, float noundef 0x3FE62E4300000000, float %.092) #139
  %289 = call float @llvm.fma.f32(float %.095, float noundef 0x3FE62E4300000000, float %.092)
  %.091 = select i1 %.not58, float %289, float %288
  %290 = icmp ugt i32 %249, 2139095039
  br i1 %290, label %__nv_fmaf_rn.exit.i, label %__internal_accurate_logf.exit

__nv_fmaf_rn.exit.i:                              ; preds = %__nv_fmaf_rn.exit10.i
  %291 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not59 = icmp eq i32 %291, 0
  %292 = call float @llvm.nvvm.fma.rn.ftz.f(float %.089, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %293 = call float @llvm.fma.f32(float %.089, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.090 = select i1 %.not59, float %293, float %292
  br label %__internal_accurate_logf.exit

__internal_accurate_logf.exit:                    ; preds = %__nv_fmaf_rn.exit.i, %__nv_fmaf_rn.exit10.i
  %r.i.0 = phi float [ %.090, %__nv_fmaf_rn.exit.i ], [ %.091, %__nv_fmaf_rn.exit10.i ]
  %294 = fcmp oeq float %.089, 0.000000e+00
  %r.i.1 = select i1 %294, float 0xFFF0000000000000, float %r.i.0
  %295 = fneg float %r.i.1
  br label %388

__internal_fmad.exit:                             ; preds = %245
  %296 = fmul float %.086, 2.000000e+00
  %297 = call float @llvm.round.f32(float %296)
  %298 = fptosi float %297 to i32
  %299 = fneg float %297
  %300 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %300, 0
  %301 = call float @llvm.nvvm.fma.rn.ftz.f(float %299, float noundef 5.000000e-01, float %.086) #139
  %302 = call float @llvm.fma.f32(float %299, float noundef 5.000000e-01, float %.086)
  %.0104 = select i1 %.not21, float %302, float %301
  %303 = fmul float %.0104, 0x400921FB60000000
  %304 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %304, 0
  %305 = call float @llvm.nvvm.mul.rn.ftz.f(float %303, float %303) #139
  %306 = fmul float %303, %303
  %.0100 = select i1 %.not22, float %306, float %305
  %307 = and i32 %298, 1
  %.not23 = icmp eq i32 %307, 0
  %308 = select i1 %.not23, float %303, float 1.000000e+00
  %309 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %309, 0
  %310 = call float @llvm.nvvm.fma.rn.ftz.f(float %.0100, float %308, float noundef 0.000000e+00) #139
  %311 = call float @llvm.fma.f32(float %.0100, float %308, float noundef 0.000000e+00)
  %.093 = select i1 %.not24, float %311, float %310
  br i1 %.not23, label %__internal_fmad.exit2.i9, label %__internal_fmad.exit1.i8

__internal_fmad.exit1.i8:                         ; preds = %__internal_fmad.exit
  %312 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %312, 0
  %313 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3EF9758000000000, float %.0100, float noundef 0xBF56C0FDA0000000) #139
  %314 = call float @llvm.fma.f32(float %.0100, float noundef 0x3EF9758000000000, float noundef 0xBF56C0FDA0000000)
  %.088 = select i1 %.not47, float %314, float %313
  br label %__internal_fmad.exit2.i9

__internal_fmad.exit2.i9:                         ; preds = %__internal_fmad.exit, %__internal_fmad.exit1.i8
  %315 = phi float [ 0xBFC5555500000000, %__internal_fmad.exit ], [ 0xBFDFFFFFE0000000, %__internal_fmad.exit1.i8 ]
  %316 = phi float [ 0x3F8110BC80000000, %__internal_fmad.exit ], [ 0x3FA5555760000000, %__internal_fmad.exit1.i8 ]
  %317 = phi float [ 0xBF29A82A60000000, %__internal_fmad.exit ], [ %.088, %__internal_fmad.exit1.i8 ]
  %318 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %318, 0
  %319 = call float @llvm.nvvm.fma.rn.ftz.f(float %317, float %.0100, float noundef %316) #139
  %320 = call float @llvm.fma.f32(float %317, float %.0100, float noundef %316)
  %.082 = select i1 %.not27, float %320, float %319
  %321 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not29 = icmp eq i32 %321, 0
  %322 = call float @llvm.nvvm.fma.rn.ftz.f(float %.082, float %.0100, float noundef %315) #139
  %323 = call float @llvm.fma.f32(float %.082, float %.0100, float noundef %315)
  %.075 = select i1 %.not29, float %323, float %322
  %324 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %324, 0
  %325 = call float @llvm.nvvm.fma.rn.ftz.f(float %.075, float %.093, float %308) #139
  %326 = call float @llvm.fma.f32(float %.075, float %.093, float %308)
  %.070 = select i1 %.not30, float %326, float %325
  %327 = and i32 %298, 2
  %.not31 = icmp eq i32 %327, 0
  br i1 %.not31, label %__internal_sin_cos_kernel.exit, label %__internal_fmad.exit5.i10

__internal_fmad.exit5.i10:                        ; preds = %__internal_fmad.exit2.i9
  %328 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %328, 0
  %329 = call float @llvm.nvvm.fma.rn.ftz.f(float %.070, float noundef -1.000000e+00, float noundef 0.000000e+00) #139
  %330 = call float @llvm.fma.f32(float %.070, float noundef -1.000000e+00, float noundef 0.000000e+00)
  %.064 = select i1 %.not46, float %330, float %329
  br label %__internal_sin_cos_kernel.exit

__internal_sin_cos_kernel.exit:                   ; preds = %__internal_fmad.exit5.i10, %__internal_fmad.exit2.i9
  %z.i.0 = phi float [ %.064, %__internal_fmad.exit5.i10 ], [ %.070, %__internal_fmad.exit2.i9 ]
  %331 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %331, 0
  %332 = call float @llvm.nvvm.fabs.ftz.f(float %z.i.0) #139
  %333 = call float @llvm.fabs.f32(float %z.i.0)
  %.062 = select i1 %.not32, float %333, float %332
  %334 = fmul float %.086, %.062
  %335 = fcmp olt float %334, 0x3810000000000000
  %336 = fmul float %334, 0x4160000000000000
  %i.i21.0 = select i1 %335, float -2.300000e+01, float 0.000000e+00
  %.060 = select i1 %335, float %336, float %334
  %337 = bitcast float %.060 to i32
  %338 = add i32 %337, -1059760811
  %339 = and i32 %338, -8388608
  %340 = sub i32 %337, %339
  %341 = bitcast i32 %340 to float
  %342 = sitofp i32 %339 to float
  %343 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not33 = icmp eq i32 %343, 0
  %344 = call float @llvm.nvvm.fma.rn.ftz.f(float %342, float noundef 0x3E80000000000000, float noundef %i.i21.0) #139
  %345 = call float @llvm.fma.f32(float %342, float noundef 0x3E80000000000000, float noundef %i.i21.0)
  %.025 = select i1 %.not33, float %345, float %344
  %346 = fadd float %341, -1.000000e+00
  %347 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not34 = icmp eq i32 %347, 0
  %348 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFC0AA04E0000000, float %346, float noundef 0x3FC2073EC0000000) #139
  %349 = call float @llvm.fma.f32(float %346, float noundef 0xBFC0AA04E0000000, float noundef 0x3FC2073EC0000000)
  %.021 = select i1 %.not34, float %349, float %348
  %350 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not35 = icmp eq i32 %350, 0
  %351 = call float @llvm.nvvm.fma.rn.ftz.f(float %.021, float %346, float noundef 0xBFBF19B980000000) #139
  %352 = call float @llvm.fma.f32(float %.021, float %346, float noundef 0xBFBF19B980000000)
  %.017 = select i1 %.not35, float %352, float %351
  %353 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not36 = icmp eq i32 %353, 0
  %354 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %346, float noundef 0x3FC1E52AA0000000) #139
  %355 = call float @llvm.fma.f32(float %.017, float %346, float noundef 0x3FC1E52AA0000000)
  %.013 = select i1 %.not36, float %355, float %354
  %356 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not37 = icmp eq i32 %356, 0
  %357 = call float @llvm.nvvm.fma.rn.ftz.f(float %.013, float %346, float noundef 0xBFC55B1720000000) #139
  %358 = call float @llvm.fma.f32(float %.013, float %346, float noundef 0xBFC55B1720000000)
  %.09 = select i1 %.not37, float %358, float %357
  %359 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not38 = icmp eq i32 %359, 0
  %360 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %346, float noundef 0x3FC99DA160000000) #139
  %361 = call float @llvm.fma.f32(float %.09, float %346, float noundef 0x3FC99DA160000000)
  %.08 = select i1 %.not38, float %361, float %360
  %362 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not39 = icmp eq i32 %362, 0
  %363 = call float @llvm.nvvm.fma.rn.ftz.f(float %.08, float %346, float noundef 0xBFCFFFE440000000) #139
  %364 = call float @llvm.fma.f32(float %.08, float %346, float noundef 0xBFCFFFE440000000)
  %.04 = select i1 %.not39, float %364, float %363
  %365 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not40 = icmp eq i32 %365, 0
  %366 = call float @llvm.nvvm.fma.rn.ftz.f(float %.04, float %346, float noundef 0x3FD5554F00000000) #139
  %367 = call float @llvm.fma.f32(float %.04, float %346, float noundef 0x3FD5554F00000000)
  %.03 = select i1 %.not40, float %367, float %366
  %368 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not41 = icmp eq i32 %368, 0
  %369 = call float @llvm.nvvm.fma.rn.ftz.f(float %.03, float %346, float noundef -5.000000e-01) #139
  %370 = call float @llvm.fma.f32(float %.03, float %346, float noundef -5.000000e-01)
  %.029 = select i1 %.not41, float %370, float %369
  %371 = fmul float %346, %.029
  %372 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %372, 0
  %373 = call float @llvm.nvvm.fma.rn.ftz.f(float %371, float %346, float %346) #139
  %374 = call float @llvm.fma.f32(float %371, float %346, float %346)
  %.034 = select i1 %.not42, float %374, float %373
  %375 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %375, 0
  %376 = call float @llvm.nvvm.fma.rn.ftz.f(float %.025, float noundef 0x3FE62E4300000000, float %.034) #139
  %377 = call float @llvm.fma.f32(float %.025, float noundef 0x3FE62E4300000000, float %.034)
  %.040 = select i1 %.not43, float %377, float %376
  %378 = icmp ugt i32 %337, 2139095039
  br i1 %378, label %__nv_fmaf_rn.exit.i34, label %__internal_accurate_logf.exit35

__nv_fmaf_rn.exit.i34:                            ; preds = %__internal_sin_cos_kernel.exit
  %379 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %379, 0
  %380 = call float @llvm.nvvm.fma.rn.ftz.f(float %.060, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000) #139
  %381 = call float @llvm.fma.f32(float %.060, float noundef 0x7FF0000000000000, float noundef 0x7FF0000000000000)
  %.052 = select i1 %.not45, float %381, float %380
  br label %__internal_accurate_logf.exit35

__internal_accurate_logf.exit35:                  ; preds = %__nv_fmaf_rn.exit.i34, %__internal_sin_cos_kernel.exit
  %r.i20.0 = phi float [ %.052, %__nv_fmaf_rn.exit.i34 ], [ %.040, %__internal_sin_cos_kernel.exit ]
  %382 = fcmp oeq float %.060, 0.000000e+00
  %r.i20.0.op = fsub float 0x3FF250D040000000, %r.i20.0
  %383 = select i1 %382, float 0x7FF0000000000000, float %r.i20.0.op
  %384 = fneg float %t.i.4
  %385 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %385, 0
  %386 = call float @llvm.nvvm.fma.rn.ftz.f(float %383, float noundef 1.000000e+00, float %384) #139
  %387 = fsub float %383, %t.i.4
  %.0 = select i1 %.not44, float %387, float %386
  br label %388

388:                                              ; preds = %__nv_fabsf.exit1, %__internal_lgammaf_pos.exit, %__internal_accurate_logf.exit35, %__internal_accurate_logf.exit
  %.01 = phi float [ %295, %__internal_accurate_logf.exit ], [ %.0, %__internal_accurate_logf.exit35 ], [ %t.i.4, %__internal_lgammaf_pos.exit ], [ 0x7FF0000000000000, %__nv_fabsf.exit1 ]
  ret float %.01
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_ldexpf(float %a, i32 %b) local_unnamed_addr #57 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp oeq float %.01, 0.000000e+00
  %4 = fcmp oeq float %.01, 0x7FF0000000000000
  %or.cond = or i1 %3, %4
  %5 = icmp eq i32 %b, 0
  %or.cond2 = select i1 %or.cond, i1 true, i1 %5
  br i1 %or.cond2, label %6, label %9

6:                                                ; preds = %__nv_fabsf.exit
  %7 = fcmp ogt float %.01, 0.000000e+00
  %8 = select i1 %7, float -0.000000e+00, float %a
  %.0 = fadd float %8, %a
  br label %38

9:                                                ; preds = %__nv_fabsf.exit
  %10 = call i32 @llvm.abs.i32(i32 %b, i1 noundef false)
  %11 = icmp slt i32 %10, 126
  br i1 %11, label %12, label %16

12:                                               ; preds = %9
  %13 = sitofp i32 %b to float
  %14 = call float @llvm.nvvm.ex2.approx.ftz.f(float %13) #139
  %15 = fmul float %14, %a
  br label %38

16:                                               ; preds = %9
  %17 = icmp ult i32 %10, 252
  br i1 %17, label %18, label %27

18:                                               ; preds = %16
  %19 = sdiv i32 %b, 2
  %20 = sub nsw i32 %b, %19
  %21 = sitofp i32 %19 to float
  %22 = call float @llvm.nvvm.ex2.approx.ftz.f(float %21) #139
  %23 = fmul float %22, %a
  %24 = sitofp i32 %20 to float
  %25 = call float @llvm.nvvm.ex2.approx.ftz.f(float %24) #139
  %26 = fmul float %23, %25
  br label %38

27:                                               ; preds = %16
  %28 = sdiv i32 %b, 4
  %29 = sitofp i32 %28 to float
  %30 = call float @llvm.nvvm.ex2.approx.ftz.f(float %29) #139
  %.neg = mul nsw i32 %28, -3
  %31 = add i32 %.neg, %b
  %32 = fmul float %30, %a
  %33 = fmul float %30, %32
  %34 = fmul float %30, %33
  %35 = sitofp i32 %31 to float
  %36 = call float @llvm.nvvm.ex2.approx.ftz.f(float %35) #139
  %37 = fmul float %36, %34
  br label %38

38:                                               ; preds = %12, %27, %18, %6
  %.3 = phi float [ %.0, %6 ], [ %15, %12 ], [ %26, %18 ], [ %37, %27 ]
  ret float %.3
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_scalbnf(float %a, i32 %b) local_unnamed_addr #58 {
__nv_fabsf.exit.i:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp oeq float %.01, 0.000000e+00
  %4 = fcmp oeq float %.01, 0x7FF0000000000000
  %or.cond = or i1 %3, %4
  %5 = icmp eq i32 %b, 0
  %or.cond2 = select i1 %or.cond, i1 true, i1 %5
  br i1 %or.cond2, label %6, label %9

6:                                                ; preds = %__nv_fabsf.exit.i
  %7 = fcmp ogt float %.01, 0.000000e+00
  %8 = select i1 %7, float -0.000000e+00, float %a
  %.0 = fadd float %8, %a
  br label %__nv_ldexpf.exit

9:                                                ; preds = %__nv_fabsf.exit.i
  %10 = call i32 @llvm.abs.i32(i32 %b, i1 noundef false)
  %11 = icmp slt i32 %10, 126
  br i1 %11, label %12, label %16

12:                                               ; preds = %9
  %13 = sitofp i32 %b to float
  %14 = call float @llvm.nvvm.ex2.approx.ftz.f(float %13) #139
  %15 = fmul float %14, %a
  br label %__nv_ldexpf.exit

16:                                               ; preds = %9
  %17 = icmp ult i32 %10, 252
  br i1 %17, label %18, label %27

18:                                               ; preds = %16
  %19 = sdiv i32 %b, 2
  %20 = sub nsw i32 %b, %19
  %21 = sitofp i32 %19 to float
  %22 = call float @llvm.nvvm.ex2.approx.ftz.f(float %21) #139
  %23 = fmul float %22, %a
  %24 = sitofp i32 %20 to float
  %25 = call float @llvm.nvvm.ex2.approx.ftz.f(float %24) #139
  %26 = fmul float %23, %25
  br label %__nv_ldexpf.exit

27:                                               ; preds = %16
  %28 = sdiv i32 %b, 4
  %29 = sitofp i32 %28 to float
  %30 = call float @llvm.nvvm.ex2.approx.ftz.f(float %29) #139
  %.neg = mul nsw i32 %28, -3
  %31 = add i32 %.neg, %b
  %32 = fmul float %30, %a
  %33 = fmul float %30, %32
  %34 = fmul float %30, %33
  %35 = sitofp i32 %31 to float
  %36 = call float @llvm.nvvm.ex2.approx.ftz.f(float %35) #139
  %37 = fmul float %36, %34
  br label %__nv_ldexpf.exit

__nv_ldexpf.exit:                                 ; preds = %12, %27, %18, %6
  %.3 = phi float [ %.0, %6 ], [ %15, %12 ], [ %26, %18 ], [ %37, %27 ]
  ret float %.3
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_frexpf(float %a, i32* nocapture nofree writeonly %b) local_unnamed_addr #59 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.01 = select i1 %.not, float %2, float %1
  %3 = fcmp olt float %.01, 0x3810000000000000
  %4 = fmul float %a, 0x4170000000000000
  %.0 = select i1 %3, float %4, float %a
  %5 = bitcast float %.0 to i32
  %6 = lshr i32 %5, 23
  %7 = and i32 %6, 255
  %8 = fcmp oeq float %.01, 0.000000e+00
  %9 = icmp eq i32 %7, 255
  %or.cond = select i1 %8, i1 true, i1 %9
  br i1 %or.cond, label %10, label %12

10:                                               ; preds = %__nv_fabsf.exit
  %11 = fadd float %.0, %.0
  br label %18

12:                                               ; preds = %__nv_fabsf.exit
  %13 = select i1 %3, i32 -150, i32 -126
  %14 = add nsw i32 %13, %7
  %15 = and i32 %5, -2139095041
  %16 = or i32 %15, 1056964608
  %17 = bitcast i32 %16 to float
  br label %18

18:                                               ; preds = %12, %10
  %.1 = phi float [ %11, %10 ], [ %17, %12 ]
  %expo.0 = phi i32 [ 0, %10 ], [ %14, %12 ]
  store i32 %expo.0, i32* %b, align 4, !noalias !68
  ret float %.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_modff(float %a, float* nocapture nofree writeonly %b) local_unnamed_addr #60 {
__nv_finitef.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = fcmp olt float %.0, 0x7FF0000000000000
  br i1 %3, label %4, label %__nv_isinff.exit

4:                                                ; preds = %__nv_finitef.exit
  %5 = call float @llvm.trunc.f32(float %a)
  store float %5, float* %b, align 4, !noalias !68
  %6 = fsub float %a, %5
  %7 = bitcast float %6 to i32
  %8 = bitcast float %a to i32
  %9 = and i32 %8, -2147483648
  %10 = or i32 %9, %7
  %11 = bitcast i32 %10 to float
  br label %20

__nv_isinff.exit:                                 ; preds = %__nv_finitef.exit
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %12, 0
  %.01 = select i1 %.not1, float %2, float %1
  %13 = fcmp oeq float %.01, 0x7FF0000000000000
  br i1 %13, label %14, label %18

14:                                               ; preds = %__nv_isinff.exit
  store float %a, float* %b, align 4, !noalias !68
  %15 = bitcast float %a to i32
  %16 = and i32 %15, -2147483648
  %17 = bitcast i32 %16 to float
  br label %20

18:                                               ; preds = %__nv_isinff.exit
  %19 = fadd float %a, %a
  store float %19, float* %b, align 4, !noalias !68
  br label %20

20:                                               ; preds = %14, %18, %4
  %t.1 = phi float [ %11, %4 ], [ %17, %14 ], [ %19, %18 ]
  ret float %t.1
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_fmodf(float %x, float %y) local_unnamed_addr #61 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %y) #139
  %2 = call float @llvm.fabs.f32(float %y)
  %.0 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %x) #139
  %5 = call float @llvm.fabs.f32(float %x)
  %.01 = select i1 %.not2, float %5, float %4
  %6 = fcmp olt float %.01, %.0
  br i1 %6, label %__internal_fmodf_kernel.exit, label %7

7:                                                ; preds = %__nv_fabsf.exit
  %8 = fmul float %.0, 0x4160000000000000
  %9 = fcmp ugt float %.01, %8
  br i1 %9, label %40, label %10

10:                                               ; preds = %7
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %11, 0
  br i1 %.not10, label %14, label %12

12:                                               ; preds = %10
  %13 = call float @llvm.nvvm.div.approx.ftz.f(float %.01, float %.0) #139
  br label %__nv_fast_fdividef.exit.i.i

14:                                               ; preds = %10
  %15 = call float @llvm.nvvm.div.approx.f(float %.01, float %.0) #139
  br label %__nv_fast_fdividef.exit.i.i

__nv_fast_fdividef.exit.i.i:                      ; preds = %14, %12
  %.02 = phi float [ %13, %12 ], [ %15, %14 ]
  %16 = call float @llvm.trunc.f32(float %.02)
  %17 = fneg float %.0
  %18 = call float @llvm.fma.f32(float %17, float %16, float %.01)
  %19 = bitcast float %18 to i32
  %20 = bitcast float %.0 to i32
  %.not11 = icmp ult i32 %19, %20
  br i1 %.not11, label %__internal_fmodf_fastpath_quot.exit.i, label %21

21:                                               ; preds = %__nv_fast_fdividef.exit.i.i
  %22 = icmp ugt i32 %19, -2147483648
  br i1 %22, label %23, label %27

23:                                               ; preds = %21
  %24 = fadd float %16, -1.000000e+00
  %25 = fcmp olt float %18, %17
  %26 = fadd float %24, -1.000000e+00
  %quotient.i.i.0 = select i1 %25, float %26, float %24
  br label %__internal_fmodf_fastpath_quot.exit.i

27:                                               ; preds = %21
  %28 = fadd float %16, 1.000000e+00
  %29 = fmul float %.0, 2.000000e+00
  %30 = fcmp ult float %18, %29
  br i1 %30, label %__internal_fmodf_fastpath_quot.exit.i, label %__nv_fmaf_rn.exit.i.i

__nv_fmaf_rn.exit.i.i:                            ; preds = %27
  %31 = fadd float %28, 1.000000e+00
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef -3.000000e+00, float %.0, float %18) #139
  %34 = call float @llvm.fma.f32(float %.0, float noundef -3.000000e+00, float %18)
  %.03 = select i1 %.not13, float %34, float %33
  %35 = fcmp oge float %.03, 0.000000e+00
  %36 = fadd float %31, 1.000000e+00
  %quotient.i.i.1 = select i1 %35, float %36, float %31
  br label %__internal_fmodf_fastpath_quot.exit.i

__internal_fmodf_fastpath_quot.exit.i:            ; preds = %23, %__nv_fmaf_rn.exit.i.i, %27, %__nv_fast_fdividef.exit.i.i
  %quotient.i.i.4 = phi float [ %16, %__nv_fast_fdividef.exit.i.i ], [ %quotient.i.i.0, %23 ], [ %quotient.i.i.1, %__nv_fmaf_rn.exit.i.i ], [ %28, %27 ]
  %37 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %37, 0
  %38 = call float @llvm.nvvm.fma.rn.ftz.f(float %17, float %quotient.i.i.4, float %.01) #139
  %39 = call float @llvm.fma.f32(float %17, float %quotient.i.i.4, float %.01)
  %.04 = select i1 %.not12, float %39, float %38
  br label %__internal_fmodf_kernel.exit

40:                                               ; preds = %7
  %41 = bitcast float %.01 to i32
  %42 = bitcast float %8 to i32
  %43 = and i32 %42, -8388608
  %44 = and i32 %41, 8388607
  %45 = or i32 %44, 1065353216
  %46 = and i32 %42, 8388607
  %47 = or i32 %46, 1065353216
  %48 = bitcast i32 %45 to float
  %49 = bitcast i32 %47 to float
  %50 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %49) #141, !srcloc !73
  %51 = add i32 %42, -192937984
  %52 = and i32 %51, -8388608
  %53 = sub i32 %41, %52
  %54 = and i32 %53, -8388608
  %.not17 = icmp eq i32 %54, 0
  br i1 %.not17, label %__internal_fmodf_slowpath_mod.exit.i, label %__nv_fmaf_rn.exit4.i.i.lr.ph

__nv_fmaf_rn.exit4.i.i.lr.ph:                     ; preds = %40
  %55 = fneg float %49
  br label %__nv_fmaf_rn.exit4.i.i

__nv_fmaf_rn.exit4.i.i:                           ; preds = %__nv_fmaf_rn.exit4.i.i.lr.ph, %__nv_fmaf_rn.exit4.i.i
  %ux.i.i.016 = phi i32 [ %45, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %79, %__nv_fmaf_rn.exit4.i.i ]
  %i.i.i.015 = phi i32 [ %54, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %78, %__nv_fmaf_rn.exit4.i.i ]
  %56 = icmp ult i32 %i.i.i.015, 192937984
  %min2 = select i1 %56, i32 %i.i.i.015, i32 192937984
  %57 = add i32 %min2, %ux.i.i.016
  %58 = bitcast i32 %57 to float
  %59 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %59, 0
  %60 = call float @llvm.nvvm.fma.rn.ftz.f(float %58, float %50, float noundef -0.000000e+00) #139
  %61 = fmul float %50, %58
  %.010 = select i1 %.not4, float %61, float %60
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %55, float %.010, float %58) #139
  %64 = call float @llvm.fma.f32(float %55, float %.010, float %58)
  %.011 = select i1 %.not5, float %64, float %63
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %50, float %.010) #139
  %67 = call float @llvm.fma.f32(float %.011, float %50, float %.010)
  %.012 = select i1 %.not6, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %55, float %.012, float %58) #139
  %70 = call float @llvm.fma.f32(float %55, float %.012, float %58)
  %.09 = select i1 %.not7, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rz.ftz.f(float %.09, float %50, float %.012) #139
  %73 = call float @llvm.nvvm.fma.rz.f(float %.09, float %50, float %.012) #139
  %.08 = select i1 %.not8, float %73, float %72
  %74 = call float @llvm.trunc.f32(float %.08)
  %75 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %75, 0
  %76 = call float @llvm.nvvm.fma.rn.ftz.f(float %55, float %74, float %58) #139
  %77 = call float @llvm.fma.f32(float %55, float %74, float %58)
  %.07 = select i1 %.not9, float %77, float %76
  %78 = sub i32 %i.i.i.015, %min2
  %79 = bitcast float %.07 to i32
  %80 = icmp ne i32 %78, 0
  %81 = icmp ne i32 %79, 0
  %spec.select = select i1 %80, i1 %81, i1 false
  br i1 %spec.select, label %__nv_fmaf_rn.exit4.i.i, label %__internal_fmodf_slowpath_mod.exit.i

__internal_fmodf_slowpath_mod.exit.i:             ; preds = %__nv_fmaf_rn.exit4.i.i, %40
  %.06.lcssa = phi float [ %48, %40 ], [ %.07, %__nv_fmaf_rn.exit4.i.i ]
  %82 = bitcast i32 %43 to float
  %83 = fcmp ule float %.0, 0.000000e+00
  %84 = icmp ugt i32 %41, 2139095039
  %or.cond = select i1 %83, i1 true, i1 %84
  %85 = select i1 %or.cond, float 0x7FFFFFFFE0000000, float %82
  %86 = fmul float %.06.lcssa, 0x3E80000000000000
  %87 = fmul float %85, %86
  br label %__internal_fmodf_kernel.exit

__internal_fmodf_kernel.exit:                     ; preds = %__internal_fmodf_fastpath_quot.exit.i, %__internal_fmodf_slowpath_mod.exit.i, %__nv_fabsf.exit
  %ans.i.1 = phi float [ %.01, %__nv_fabsf.exit ], [ %.04, %__internal_fmodf_fastpath_quot.exit.i ], [ %87, %__internal_fmodf_slowpath_mod.exit.i ]
  %88 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %88, 0
  %89 = call float @llvm.nvvm.fabs.ftz.f(float %ans.i.1) #139
  %90 = call float @llvm.fabs.f32(float %ans.i.1)
  %.05 = select i1 %.not3, float %90, float %89
  %91 = fcmp ugt float %.05, 0x7FF0000000000000
  br i1 %91, label %98, label %92

92:                                               ; preds = %__internal_fmodf_kernel.exit
  %93 = bitcast float %ans.i.1 to i32
  %94 = bitcast float %x to i32
  %95 = and i32 %94, -2147483648
  %96 = or i32 %95, %93
  %97 = bitcast i32 %96 to float
  br label %98

98:                                               ; preds = %__internal_fmodf_kernel.exit, %92
  %99 = phi float [ %97, %92 ], [ %ans.i.1, %__internal_fmodf_kernel.exit ]
  ret float %99
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_remainderf(float %x, float %y) local_unnamed_addr #62 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %y) #139
  %2 = call float @llvm.fabs.f32(float %y)
  %.01 = select i1 %.not, float %2, float %1
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %x) #139
  %5 = call float @llvm.fabs.f32(float %x)
  %.02 = select i1 %.not6, float %5, float %4
  %6 = fmul float %.02, 2.000000e+00
  %7 = fcmp olt float %.02, %.01
  br i1 %7, label %__internal_fmodf_kernel.exit, label %8

8:                                                ; preds = %__nv_fabsf.exit
  %9 = fmul float %.01, 0x4160000000000000
  %10 = fcmp ugt float %.02, %9
  br i1 %10, label %43, label %11

11:                                               ; preds = %8
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %12, 0
  br i1 %.not17, label %15, label %13

13:                                               ; preds = %11
  %14 = call float @llvm.nvvm.div.approx.ftz.f(float %.02, float %.01) #139
  br label %__nv_fast_fdividef.exit.i.i

15:                                               ; preds = %11
  %16 = call float @llvm.nvvm.div.approx.f(float %.02, float %.01) #139
  br label %__nv_fast_fdividef.exit.i.i

__nv_fast_fdividef.exit.i.i:                      ; preds = %15, %13
  %.04 = phi float [ %14, %13 ], [ %16, %15 ]
  %17 = call float @llvm.trunc.f32(float %.04)
  %18 = fneg float %.01
  %19 = call float @llvm.fma.f32(float %18, float %17, float %.02)
  %20 = bitcast float %19 to i32
  %21 = bitcast float %.01 to i32
  %.not18 = icmp ult i32 %20, %21
  br i1 %.not18, label %__internal_fmodf_fastpath_quot.exit.i, label %22

22:                                               ; preds = %__nv_fast_fdividef.exit.i.i
  %23 = icmp ugt i32 %20, -2147483648
  br i1 %23, label %24, label %28

24:                                               ; preds = %22
  %25 = fadd float %17, -1.000000e+00
  %26 = fcmp olt float %19, %18
  %27 = fadd float %25, -1.000000e+00
  %quotient.i.i.0 = select i1 %26, float %27, float %25
  br label %__internal_fmodf_fastpath_quot.exit.i

28:                                               ; preds = %22
  %29 = fadd float %17, 1.000000e+00
  %30 = fmul float %.01, 2.000000e+00
  %31 = fcmp ult float %19, %30
  br i1 %31, label %__internal_fmodf_fastpath_quot.exit.i, label %__nv_fmaf_rn.exit.i.i

__nv_fmaf_rn.exit.i.i:                            ; preds = %28
  %32 = fadd float %29, 1.000000e+00
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef -3.000000e+00, float %.01, float %19) #139
  %35 = call float @llvm.fma.f32(float %.01, float noundef -3.000000e+00, float %19)
  %.05 = select i1 %.not20, float %35, float %34
  %36 = fcmp oge float %.05, 0.000000e+00
  %37 = fadd float %32, 1.000000e+00
  %quotient.i.i.1 = select i1 %36, float %37, float %32
  br label %__internal_fmodf_fastpath_quot.exit.i

__internal_fmodf_fastpath_quot.exit.i:            ; preds = %24, %__nv_fmaf_rn.exit.i.i, %28, %__nv_fast_fdividef.exit.i.i
  %quotient.i.i.4 = phi float [ %17, %__nv_fast_fdividef.exit.i.i ], [ %quotient.i.i.0, %24 ], [ %quotient.i.i.1, %__nv_fmaf_rn.exit.i.i ], [ %29, %28 ]
  %38 = fptoui float %quotient.i.i.4 to i32
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float %18, float %quotient.i.i.4, float %.02) #139
  %41 = call float @llvm.fma.f32(float %18, float %quotient.i.i.4, float %.02)
  %.06 = select i1 %.not19, float %41, float %40
  %42 = call float @llvm.fma.f32(float %41, float noundef 2.000000e+00, float noundef 0.000000e+00)
  br label %__internal_fmodf_kernel.exit

43:                                               ; preds = %8
  %44 = bitcast float %.02 to i32
  %45 = bitcast float %9 to i32
  %46 = and i32 %45, -8388608
  %47 = and i32 %44, 8388607
  %48 = or i32 %47, 1065353216
  %49 = and i32 %45, 8388607
  %50 = or i32 %49, 1065353216
  %51 = bitcast i32 %48 to float
  %52 = bitcast i32 %50 to float
  %53 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %52) #141, !srcloc !73
  %54 = add i32 %45, -192937984
  %55 = and i32 %54, -8388608
  %56 = sub i32 %44, %55
  %57 = and i32 %56, -8388608
  %.not26 = icmp eq i32 %57, 0
  br i1 %.not26, label %__internal_fmodf_slowpath_mod.exit.i, label %__nv_fmaf_rn.exit4.i.i.lr.ph

__nv_fmaf_rn.exit4.i.i.lr.ph:                     ; preds = %43
  %58 = fneg float %52
  br label %__nv_fmaf_rn.exit4.i.i

__nv_fmaf_rn.exit4.i.i:                           ; preds = %__nv_fmaf_rn.exit4.i.i.lr.ph, %__nv_fmaf_rn.exit4.i.i
  %ux.i.i.023 = phi i32 [ %48, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %82, %__nv_fmaf_rn.exit4.i.i ]
  %i.i.i.022 = phi i32 [ %57, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %81, %__nv_fmaf_rn.exit4.i.i ]
  %quotient.i8.i.021 = phi i32 [ 0, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %86, %__nv_fmaf_rn.exit4.i.i ]
  %59 = icmp ult i32 %i.i.i.022, 192937984
  %min2 = select i1 %59, i32 %i.i.i.022, i32 192937984
  %60 = add i32 %min2, %ux.i.i.023
  %61 = bitcast i32 %60 to float
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %62, 0
  %63 = call float @llvm.nvvm.fma.rn.ftz.f(float %61, float %53, float noundef -0.000000e+00) #139
  %64 = fmul float %53, %61
  %.013 = select i1 %.not11, float %64, float %63
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %58, float %.013, float %61) #139
  %67 = call float @llvm.fma.f32(float %58, float %.013, float %61)
  %.014 = select i1 %.not12, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %.014, float %53, float %.013) #139
  %70 = call float @llvm.fma.f32(float %.014, float %53, float %.013)
  %.015 = select i1 %.not13, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rn.ftz.f(float %58, float %.015, float %61) #139
  %73 = call float @llvm.fma.f32(float %58, float %.015, float %61)
  %.012 = select i1 %.not14, float %73, float %72
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %74, 0
  %75 = call float @llvm.nvvm.fma.rz.ftz.f(float %.012, float %53, float %.015) #139
  %76 = call float @llvm.nvvm.fma.rz.f(float %.012, float %53, float %.015) #139
  %.011 = select i1 %.not15, float %76, float %75
  %77 = call float @llvm.trunc.f32(float %.011)
  %78 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %78, 0
  %79 = call float @llvm.nvvm.fma.rn.ftz.f(float %58, float %77, float %61) #139
  %80 = call float @llvm.fma.f32(float %58, float %77, float %61)
  %.010 = select i1 %.not16, float %80, float %79
  %81 = sub i32 %i.i.i.022, %min2
  %82 = bitcast float %.010 to i32
  %83 = lshr i32 %min2, 23
  %84 = shl i32 %quotient.i8.i.021, %83
  %85 = fptoui float %77 to i32
  %86 = add i32 %84, %85
  %87 = icmp ne i32 %81, 0
  %88 = icmp ne i32 %82, 0
  %spec.select = select i1 %87, i1 %88, i1 false
  br i1 %spec.select, label %__nv_fmaf_rn.exit4.i.i, label %__internal_fmodf_slowpath_mod.exit.i

__internal_fmodf_slowpath_mod.exit.i:             ; preds = %__nv_fmaf_rn.exit4.i.i, %43
  %.09.lcssa = phi float [ %51, %43 ], [ %.010, %__nv_fmaf_rn.exit4.i.i ]
  %quotient.i8.i.0.lcssa = phi i32 [ 0, %43 ], [ %86, %__nv_fmaf_rn.exit4.i.i ]
  %i.i.i.0.lcssa = phi i32 [ 0, %43 ], [ %81, %__nv_fmaf_rn.exit4.i.i ]
  %89 = lshr i32 %i.i.i.0.lcssa, 23
  %min.cond = icmp ult i32 %i.i.i.0.lcssa, 33554432
  %min = select i1 %min.cond, i32 %89, i32 3
  %90 = shl i32 %quotient.i8.i.0.lcssa, %min
  %91 = bitcast i32 %46 to float
  %92 = fcmp ule float %.01, 0.000000e+00
  %93 = icmp ugt i32 %44, 2139095039
  %or.cond = select i1 %92, i1 true, i1 %93
  %94 = select i1 %or.cond, float 0x7FFFFFFFE0000000, float %91
  %95 = fmul float %.09.lcssa, 0x3E80000000000000
  %96 = fmul float %94, %95
  %97 = fmul float %.09.lcssa, 0x3E90000000000000
  %98 = fmul float %94, %97
  br label %__internal_fmodf_kernel.exit

__internal_fmodf_kernel.exit:                     ; preds = %__internal_fmodf_fastpath_quot.exit.i, %__internal_fmodf_slowpath_mod.exit.i, %__nv_fabsf.exit
  %ignore_quot.1 = phi i32 [ 0, %__nv_fabsf.exit ], [ %38, %__internal_fmodf_fastpath_quot.exit.i ], [ %90, %__internal_fmodf_slowpath_mod.exit.i ]
  %twoans.1 = phi float [ %6, %__nv_fabsf.exit ], [ %42, %__internal_fmodf_fastpath_quot.exit.i ], [ %98, %__internal_fmodf_slowpath_mod.exit.i ]
  %ans.i.1 = phi float [ %.02, %__nv_fabsf.exit ], [ %.06, %__internal_fmodf_fastpath_quot.exit.i ], [ %96, %__internal_fmodf_slowpath_mod.exit.i ]
  %99 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %100 = icmp eq i32 %99, 0
  %101 = fmul float %ans.i.1, 2.000000e+00
  %spec.select2 = select i1 %100, float %101, float %twoans.1
  %102 = fcmp ogt float %spec.select2, %.01
  br i1 %102, label %107, label %103

103:                                              ; preds = %__internal_fmodf_kernel.exit
  %104 = fcmp oeq float %spec.select2, %.01
  %105 = and i32 %ignore_quot.1, 1
  %106 = icmp ne i32 %105, 0
  %or.cond4 = select i1 %104, i1 %106, i1 false
  br i1 %or.cond4, label %107, label %__internal_remainderf_round_to_nearest_even.exit

107:                                              ; preds = %103, %__internal_fmodf_kernel.exit
  %108 = fcmp olt float %ans.i.1, 0x3810000000000000
  br i1 %108, label %109, label %115

109:                                              ; preds = %107
  %110 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %110, 0
  br i1 %.not9, label %115, label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %109
  %111 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %111, 0
  %112 = call float @llvm.nvvm.fma.rn.ftz.f(float %.01, float noundef -2.000000e+00, float %spec.select2) #139
  %113 = call float @llvm.fma.f32(float %.01, float noundef -2.000000e+00, float %spec.select2)
  %.03 = select i1 %.not10, float %113, float %112
  %114 = fmul float %.03, 5.000000e-01
  br label %__internal_remainderf_round_to_nearest_even.exit

115:                                              ; preds = %109, %107
  %116 = fsub float %ans.i.1, %.01
  br label %__internal_remainderf_round_to_nearest_even.exit

__internal_remainderf_round_to_nearest_even.exit: ; preds = %__internal_fmad.exit.i, %115, %103
  %.1 = phi float [ %ans.i.1, %103 ], [ %114, %__internal_fmad.exit.i ], [ %116, %115 ]
  %117 = bitcast float %x to i32
  %.not7 = icmp sgt i32 %117, -1
  br i1 %.not7, label %123, label %__nv_isnanf.exit

__nv_isnanf.exit:                                 ; preds = %__internal_remainderf_round_to_nearest_even.exit
  %118 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %118, 0
  %119 = call float @llvm.nvvm.fabs.ftz.f(float %.1) #139
  %120 = call float @llvm.fabs.f32(float %.1)
  %.0 = select i1 %.not8, float %120, float %119
  %121 = fcmp ole float %.0, 0x7FF0000000000000
  %122 = fneg float %.1
  %spec.select5 = select i1 %121, float %122, float %.1
  br label %123

123:                                              ; preds = %__nv_isnanf.exit, %__internal_remainderf_round_to_nearest_even.exit
  %ans.0 = phi float [ %.1, %__internal_remainderf_round_to_nearest_even.exit ], [ %spec.select5, %__nv_isnanf.exit ]
  ret float %ans.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_remquof(float %x, float %y, i32* nocapture nofree writeonly %quo) local_unnamed_addr #63 {
__nv_fabsf.exit:
  %0 = bitcast float %x to i32
  %1 = bitcast float %y to i32
  %2 = xor i32 %1, %0
  %.lobit = ashr i32 %2, 31
  %3 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %3, 0
  %4 = call float @llvm.nvvm.fabs.ftz.f(float %y) #139
  %5 = call float @llvm.fabs.f32(float %y)
  %.05 = select i1 %.not, float %5, float %4
  %6 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %6, 0
  %7 = call float @llvm.nvvm.fabs.ftz.f(float %x) #139
  %8 = call float @llvm.fabs.f32(float %x)
  %.06 = select i1 %.not6, float %8, float %7
  %9 = fmul float %.06, 2.000000e+00
  %10 = fcmp olt float %.06, %.05
  br i1 %10, label %__internal_fmodf_kernel.exit, label %11

11:                                               ; preds = %__nv_fabsf.exit
  %12 = fmul float %.05, 0x4160000000000000
  %13 = fcmp ugt float %.06, %12
  br i1 %13, label %46, label %14

14:                                               ; preds = %11
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %15, 0
  br i1 %.not18, label %18, label %16

16:                                               ; preds = %14
  %17 = call float @llvm.nvvm.div.approx.ftz.f(float %.06, float %.05) #139
  br label %__nv_fast_fdividef.exit.i.i

18:                                               ; preds = %14
  %19 = call float @llvm.nvvm.div.approx.f(float %.06, float %.05) #139
  br label %__nv_fast_fdividef.exit.i.i

__nv_fast_fdividef.exit.i.i:                      ; preds = %18, %16
  %.08 = phi float [ %17, %16 ], [ %19, %18 ]
  %20 = call float @llvm.trunc.f32(float %.08)
  %21 = fneg float %.05
  %22 = call float @llvm.fma.f32(float %21, float %20, float %.06)
  %23 = bitcast float %22 to i32
  %24 = bitcast float %.05 to i32
  %.not19 = icmp ult i32 %23, %24
  br i1 %.not19, label %__internal_fmodf_fastpath_quot.exit.i, label %25

25:                                               ; preds = %__nv_fast_fdividef.exit.i.i
  %26 = icmp ugt i32 %23, -2147483648
  br i1 %26, label %27, label %31

27:                                               ; preds = %25
  %28 = fadd float %20, -1.000000e+00
  %29 = fcmp olt float %22, %21
  %30 = fadd float %28, -1.000000e+00
  %quotient.i.i.0 = select i1 %29, float %30, float %28
  br label %__internal_fmodf_fastpath_quot.exit.i

31:                                               ; preds = %25
  %32 = fadd float %20, 1.000000e+00
  %33 = fmul float %.05, 2.000000e+00
  %34 = fcmp ult float %22, %33
  br i1 %34, label %__internal_fmodf_fastpath_quot.exit.i, label %__nv_fmaf_rn.exit.i.i

__nv_fmaf_rn.exit.i.i:                            ; preds = %31
  %35 = fadd float %32, 1.000000e+00
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef -3.000000e+00, float %.05, float %22) #139
  %38 = call float @llvm.fma.f32(float %.05, float noundef -3.000000e+00, float %22)
  %.09 = select i1 %.not21, float %38, float %37
  %39 = fcmp oge float %.09, 0.000000e+00
  %40 = fadd float %35, 1.000000e+00
  %quotient.i.i.1 = select i1 %39, float %40, float %35
  br label %__internal_fmodf_fastpath_quot.exit.i

__internal_fmodf_fastpath_quot.exit.i:            ; preds = %27, %__nv_fmaf_rn.exit.i.i, %31, %__nv_fast_fdividef.exit.i.i
  %quotient.i.i.4 = phi float [ %20, %__nv_fast_fdividef.exit.i.i ], [ %quotient.i.i.0, %27 ], [ %quotient.i.i.1, %__nv_fmaf_rn.exit.i.i ], [ %32, %31 ]
  %41 = fptoui float %quotient.i.i.4 to i32
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %21, float %quotient.i.i.4, float %.06) #139
  %44 = call float @llvm.fma.f32(float %21, float %quotient.i.i.4, float %.06)
  %.010 = select i1 %.not20, float %44, float %43
  %45 = call float @llvm.fma.f32(float %44, float noundef 2.000000e+00, float noundef 0.000000e+00)
  br label %__internal_fmodf_kernel.exit

46:                                               ; preds = %11
  %47 = bitcast float %.06 to i32
  %48 = bitcast float %12 to i32
  %49 = and i32 %48, -8388608
  %50 = and i32 %47, 8388607
  %51 = or i32 %50, 1065353216
  %52 = and i32 %48, 8388607
  %53 = or i32 %52, 1065353216
  %54 = bitcast i32 %51 to float
  %55 = bitcast i32 %53 to float
  %56 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %55) #141, !srcloc !73
  %57 = add i32 %48, -192937984
  %58 = and i32 %57, -8388608
  %59 = sub i32 %47, %58
  %60 = and i32 %59, -8388608
  %.not27 = icmp eq i32 %60, 0
  br i1 %.not27, label %__internal_fmodf_slowpath_mod.exit.i, label %__nv_fmaf_rn.exit4.i.i.lr.ph

__nv_fmaf_rn.exit4.i.i.lr.ph:                     ; preds = %46
  %61 = fneg float %55
  br label %__nv_fmaf_rn.exit4.i.i

__nv_fmaf_rn.exit4.i.i:                           ; preds = %__nv_fmaf_rn.exit4.i.i.lr.ph, %__nv_fmaf_rn.exit4.i.i
  %ux.i.i.024 = phi i32 [ %51, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %85, %__nv_fmaf_rn.exit4.i.i ]
  %i.i.i.023 = phi i32 [ %60, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %84, %__nv_fmaf_rn.exit4.i.i ]
  %quotient.i8.i.022 = phi i32 [ 0, %__nv_fmaf_rn.exit4.i.i.lr.ph ], [ %89, %__nv_fmaf_rn.exit4.i.i ]
  %62 = icmp ult i32 %i.i.i.023, 192937984
  %min2 = select i1 %62, i32 %i.i.i.023, i32 192937984
  %63 = add i32 %min2, %ux.i.i.024
  %64 = bitcast i32 %63 to float
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %64, float %56, float noundef -0.000000e+00) #139
  %67 = fmul float %56, %64
  %.012 = select i1 %.not12, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %61, float %.012, float %64) #139
  %70 = call float @llvm.fma.f32(float %61, float %.012, float %64)
  %.011 = select i1 %.not13, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.fma.rn.ftz.f(float %.011, float %56, float %.012) #139
  %73 = call float @llvm.fma.f32(float %.011, float %56, float %.012)
  %.07 = select i1 %.not14, float %73, float %72
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %74, 0
  %75 = call float @llvm.nvvm.fma.rn.ftz.f(float %61, float %.07, float %64) #139
  %76 = call float @llvm.fma.f32(float %61, float %.07, float %64)
  %.014 = select i1 %.not15, float %76, float %75
  %77 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %77, 0
  %78 = call float @llvm.nvvm.fma.rz.ftz.f(float %.014, float %56, float %.07) #139
  %79 = call float @llvm.nvvm.fma.rz.f(float %.014, float %56, float %.07) #139
  %.016 = select i1 %.not16, float %79, float %78
  %80 = call float @llvm.trunc.f32(float %.016)
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %81, 0
  %82 = call float @llvm.nvvm.fma.rn.ftz.f(float %61, float %80, float %64) #139
  %83 = call float @llvm.fma.f32(float %61, float %80, float %64)
  %.015 = select i1 %.not17, float %83, float %82
  %84 = sub i32 %i.i.i.023, %min2
  %85 = bitcast float %.015 to i32
  %86 = lshr i32 %min2, 23
  %87 = shl i32 %quotient.i8.i.022, %86
  %88 = fptoui float %80 to i32
  %89 = add i32 %87, %88
  %90 = icmp ne i32 %84, 0
  %91 = icmp ne i32 %85, 0
  %spec.select = select i1 %90, i1 %91, i1 false
  br i1 %spec.select, label %__nv_fmaf_rn.exit4.i.i, label %__internal_fmodf_slowpath_mod.exit.i

__internal_fmodf_slowpath_mod.exit.i:             ; preds = %__nv_fmaf_rn.exit4.i.i, %46
  %.013.lcssa = phi float [ %54, %46 ], [ %.015, %__nv_fmaf_rn.exit4.i.i ]
  %quotient.i8.i.0.lcssa = phi i32 [ 0, %46 ], [ %89, %__nv_fmaf_rn.exit4.i.i ]
  %i.i.i.0.lcssa = phi i32 [ 0, %46 ], [ %84, %__nv_fmaf_rn.exit4.i.i ]
  %92 = lshr i32 %i.i.i.0.lcssa, 23
  %min.cond = icmp ult i32 %i.i.i.0.lcssa, 33554432
  %min = select i1 %min.cond, i32 %92, i32 3
  %93 = shl i32 %quotient.i8.i.0.lcssa, %min
  %94 = bitcast i32 %49 to float
  %95 = fcmp ule float %.05, 0.000000e+00
  %96 = icmp ugt i32 %47, 2139095039
  %or.cond = select i1 %95, i1 true, i1 %96
  %97 = select i1 %or.cond, float 0x7FFFFFFFE0000000, float %94
  %98 = fmul float %.013.lcssa, 0x3E80000000000000
  %99 = fmul float %97, %98
  %100 = fmul float %.013.lcssa, 0x3E90000000000000
  %101 = fmul float %97, %100
  br label %__internal_fmodf_kernel.exit

__internal_fmodf_kernel.exit:                     ; preds = %__internal_fmodf_fastpath_quot.exit.i, %__internal_fmodf_slowpath_mod.exit.i, %__nv_fabsf.exit
  %quot.1 = phi i32 [ 0, %__nv_fabsf.exit ], [ %41, %__internal_fmodf_fastpath_quot.exit.i ], [ %93, %__internal_fmodf_slowpath_mod.exit.i ]
  %twoans.1 = phi float [ %9, %__nv_fabsf.exit ], [ %45, %__internal_fmodf_fastpath_quot.exit.i ], [ %101, %__internal_fmodf_slowpath_mod.exit.i ]
  %ans.i.1 = phi float [ %.06, %__nv_fabsf.exit ], [ %.010, %__internal_fmodf_fastpath_quot.exit.i ], [ %99, %__internal_fmodf_slowpath_mod.exit.i ]
  %102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %103 = icmp eq i32 %102, 0
  %104 = fmul float %ans.i.1, 2.000000e+00
  %spec.select2 = select i1 %103, float %104, float %twoans.1
  %105 = fcmp ogt float %spec.select2, %.05
  br i1 %105, label %110, label %106

106:                                              ; preds = %__internal_fmodf_kernel.exit
  %107 = fcmp oeq float %spec.select2, %.05
  %108 = and i32 %quot.1, 1
  %109 = icmp ne i32 %108, 0
  %or.cond4 = select i1 %107, i1 %109, i1 false
  br i1 %or.cond4, label %110, label %__internal_remainderf_round_to_nearest_even.exit

110:                                              ; preds = %106, %__internal_fmodf_kernel.exit
  %111 = add i32 %quot.1, 1
  %112 = fcmp olt float %ans.i.1, 0x3810000000000000
  br i1 %112, label %113, label %119

113:                                              ; preds = %110
  %114 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %114, 0
  br i1 %.not10, label %119, label %__internal_fmad.exit.i

__internal_fmad.exit.i:                           ; preds = %113
  %115 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %115, 0
  %116 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float noundef -2.000000e+00, float %spec.select2) #139
  %117 = call float @llvm.fma.f32(float %.05, float noundef -2.000000e+00, float %spec.select2)
  %.04 = select i1 %.not11, float %117, float %116
  %118 = fmul float %.04, 5.000000e-01
  br label %__internal_remainderf_round_to_nearest_even.exit

119:                                              ; preds = %113, %110
  %120 = fsub float %ans.i.1, %.05
  br label %__internal_remainderf_round_to_nearest_even.exit

__internal_remainderf_round_to_nearest_even.exit: ; preds = %__internal_fmad.exit.i, %119, %106
  %quot.2 = phi i32 [ %quot.1, %106 ], [ %111, %119 ], [ %111, %__internal_fmad.exit.i ]
  %.1 = phi float [ %ans.i.1, %106 ], [ %120, %119 ], [ %118, %__internal_fmad.exit.i ]
  %.not7 = icmp sgt i32 %0, -1
  br i1 %.not7, label %__nv_isnanf.exit, label %__nv_isnanf.exit3

__nv_isnanf.exit3:                                ; preds = %__internal_remainderf_round_to_nearest_even.exit
  %121 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %121, 0
  %122 = call float @llvm.nvvm.fabs.ftz.f(float %.1) #139
  %123 = call float @llvm.fabs.f32(float %.1)
  %.01 = select i1 %.not9, float %123, float %122
  %124 = fcmp ole float %.01, 0x7FF0000000000000
  %125 = fneg float %.1
  %spec.select5 = select i1 %124, float %125, float %.1
  br label %__nv_isnanf.exit

__nv_isnanf.exit:                                 ; preds = %__nv_isnanf.exit3, %__internal_remainderf_round_to_nearest_even.exit
  %ans.0 = phi float [ %.1, %__internal_remainderf_round_to_nearest_even.exit ], [ %spec.select5, %__nv_isnanf.exit3 ]
  %126 = and i32 %quot.2, 7
  %127 = xor i32 %126, %.lobit
  %128 = sub nsw i32 %127, %.lobit
  %129 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %129, 0
  %130 = call float @llvm.nvvm.fabs.ftz.f(float %ans.0) #139
  %131 = call float @llvm.fabs.f32(float %ans.0)
  %.0 = select i1 %.not8, float %131, float %130
  %132 = fcmp ugt float %.0, 0x7FF0000000000000
  %133 = select i1 %132, i32 0, i32 %128
  store i32 %133, i32* %quo, align 4, !noalias !68
  ret float %ans.0
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_powif(float %a, i32 %b) local_unnamed_addr #64 {
  %1 = call i32 @llvm.abs.i32(i32 %b, i1 noundef false)
  br label %2

2:                                                ; preds = %__nv_fmul_rn.exit, %0
  %.01 = phi float [ %a, %0 ], [ %.02, %__nv_fmul_rn.exit ]
  %e.0 = phi i32 [ %1, %0 ], [ %13, %__nv_fmul_rn.exit ]
  %r.0 = phi float [ 1.000000e+00, %0 ], [ %r.1, %__nv_fmul_rn.exit ]
  %3 = and i32 %e.0, 1
  %.not = icmp eq i32 %3, 0
  br i1 %.not, label %7, label %__nv_fmul_rn.exit1

__nv_fmul_rn.exit1:                               ; preds = %2
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %4, 0
  %5 = call float @llvm.nvvm.mul.rn.ftz.f(float %r.0, float %.01) #139
  %6 = fmul float %.01, %r.0
  %.0 = select i1 %.not2, float %6, float %5
  br label %7

7:                                                ; preds = %__nv_fmul_rn.exit1, %2
  %r.1 = phi float [ %.0, %__nv_fmul_rn.exit1 ], [ %r.0, %2 ]
  %8 = icmp ult i32 %e.0, 2
  br i1 %8, label %9, label %__nv_fmul_rn.exit

9:                                                ; preds = %7
  %10 = icmp slt i32 %b, 0
  %11 = fdiv float 1.000000e+00, %r.1
  %12 = select i1 %10, float %11, float %r.1
  ret float %12

__nv_fmul_rn.exit:                                ; preds = %7
  %13 = lshr i32 %e.0, 1
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %14, 0
  %15 = call float @llvm.nvvm.mul.rn.ftz.f(float %.01, float %.01) #139
  %16 = fmul float %.01, %.01
  %.02 = select i1 %.not1, float %16, float %15
  br label %2
}

; Function Attrs: nofree noinline nosync nounwind readnone
define dso_local double @__nv_powi(double %a, i32 %b) local_unnamed_addr #65 {
  %1 = call i32 @llvm.abs.i32(i32 %b, i1 noundef false)
  %2 = and i32 %1, 1
  %.not1 = icmp eq i32 %2, 0
  %3 = select i1 %.not1, double 1.000000e+00, double %a
  %4 = icmp ult i32 %1, 2
  br i1 %4, label %._crit_edge, label %.lr.ph

._crit_edge:                                      ; preds = %.lr.ph, %0
  %spec.select.lcssa = phi double [ %3, %0 ], [ %spec.select, %.lr.ph ]
  %5 = icmp slt i32 %b, 0
  %6 = fdiv double 1.000000e+00, %spec.select.lcssa
  %7 = select i1 %5, double %6, double %spec.select.lcssa
  ret double %7

.lr.ph:                                           ; preds = %0, %.lr.ph
  %spec.select4 = phi double [ %spec.select, %.lr.ph ], [ %3, %0 ]
  %e.03 = phi i32 [ %8, %.lr.ph ], [ %1, %0 ]
  %.02 = phi double [ %9, %.lr.ph ], [ %a, %0 ]
  %8 = lshr i32 %e.03, 1
  %9 = fmul double %.02, %.02
  %10 = and i32 %e.03, 2
  %.not = icmp eq i32 %10, 0
  %11 = select i1 %.not, double 1.000000e+00, double %9
  %spec.select = fmul double %spec.select4, %11
  %12 = icmp ult i32 %e.03, 4
  br i1 %12, label %._crit_edge, label %.lr.ph
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_powf(float %a, float %b) local_unnamed_addr #66 {
__nv_fabsf.exit:
  %0 = fmul float %b, 5.000000e-01
  %1 = call float @llvm.trunc.f32(float %0) #140
  %2 = fmul float %1, 2.000000e+00
  %3 = fsub float %b, %2
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %4, 0
  %5 = call float @llvm.nvvm.fabs.ftz.f(float %3) #139
  %6 = call float @llvm.fabs.f32(float %3)
  %.010 = select i1 %.not, float %6, float %5
  %7 = fcmp oeq float %.010, 1.000000e+00
  %8 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %8, 0
  %9 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %10 = call float @llvm.fabs.f32(float %a)
  %.011 = select i1 %.not1, float %10, float %9
  %11 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %11, 0
  %12 = fcmp olt float %.011, 0x3810000000000000
  %13 = fmul float %.011, 0x4170000000000000
  %14 = select i1 %.not2, i1 %12, i1 false
  %.114 = select i1 %14, float %13, float %.011
  %expo.i.i.1 = select i1 %14, float -2.400000e+01, float 0.000000e+00
  %15 = bitcast float %.114 to i32
  %16 = add i32 %15, -1060439283
  %17 = and i32 %16, -8388608
  %18 = sub i32 %15, %17
  %19 = bitcast i32 %18 to float
  %20 = sitofp i32 %17 to float
  %21 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %21, 0
  %22 = call float @llvm.nvvm.fma.rn.ftz.f(float %20, float noundef 0x3E80000000000000, float noundef %expo.i.i.1) #139
  %23 = call float @llvm.fma.f32(float %20, float noundef 0x3E80000000000000, float noundef %expo.i.i.1)
  %.021 = select i1 %.not3, float %23, float %22
  %24 = fadd float %19, -1.000000e+00
  %25 = fadd float %19, 1.000000e+00
  %26 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %25) #141, !srcloc !73
  %27 = fmul float %24, 2.000000e+00
  %28 = fmul float %26, %27
  %29 = fmul float %28, %28
  %30 = fsub float %24, %28
  %31 = fmul float %30, 2.000000e+00
  %32 = fneg float %28
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %33, 0
  %34 = call float @llvm.nvvm.fma.rn.ftz.f(float %32, float %24, float %31) #139
  %35 = call float @llvm.fma.f32(float %32, float %24, float %31)
  %.022 = select i1 %.not4, float %35, float %34
  %36 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %36, 0
  %37 = call float @llvm.nvvm.mul.rn.ftz.f(float %26, float %.022) #139
  %38 = fmul float %26, %.022
  %.023 = select i1 %.not5, float %38, float %37
  %39 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %39, 0
  %40 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F45865C80000000, float %29, float noundef 0x3F6A5CFB60000000) #139
  %41 = call float @llvm.fma.f32(float %29, float noundef 0x3F45865C80000000, float noundef 0x3F6A5CFB60000000)
  %.024 = select i1 %.not6, float %41, float %40
  %42 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %42, 0
  %43 = call float @llvm.nvvm.fma.rn.ftz.f(float %.024, float %29, float noundef 0x3F92776E60000000) #139
  %44 = call float @llvm.fma.f32(float %.024, float %29, float noundef 0x3F92776E60000000)
  %.025 = select i1 %.not7, float %44, float %43
  %45 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %45, 0
  %46 = call float @llvm.nvvm.fma.rn.ftz.f(float %.025, float %29, float noundef 0x3FBEC709E0000000) #139
  %47 = call float @llvm.fma.f32(float %.025, float %29, float noundef 0x3FBEC709E0000000)
  %.027 = select i1 %.not8, float %47, float %46
  %48 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %48, 0
  %49 = call float @llvm.nvvm.mul.rn.ftz.f(float %.027, float %29) #139
  %50 = fmul float %29, %.027
  %.028 = select i1 %.not9, float %50, float %49
  %51 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %51, 0
  %52 = call float @llvm.nvvm.fma.rn.ftz.f(float %28, float noundef 0x3FF7154760000000, float %.021) #139
  %53 = call float @llvm.fma.f32(float %28, float noundef 0x3FF7154760000000, float %.021)
  %.029 = select i1 %.not10, float %53, float %52
  %54 = fsub float %.021, %.029
  %55 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not11 = icmp eq i32 %55, 0
  %56 = call float @llvm.nvvm.fma.rn.ftz.f(float %28, float noundef 0x3FF7154760000000, float %54) #139
  %57 = call float @llvm.fma.f32(float %28, float noundef 0x3FF7154760000000, float %54)
  %.030 = select i1 %.not11, float %57, float %56
  %58 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %58, 0
  %59 = call float @llvm.nvvm.fma.rn.ftz.f(float %.023, float noundef 0x3FF7154760000000, float %.030) #139
  %60 = call float @llvm.fma.f32(float %.023, float noundef 0x3FF7154760000000, float %.030)
  %.031 = select i1 %.not12, float %60, float %59
  %61 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %61, 0
  %62 = call float @llvm.nvvm.fma.rn.ftz.f(float %28, float noundef 0x3E54ABC680000000, float %.031) #139
  %63 = call float @llvm.fma.f32(float %28, float noundef 0x3E54ABC680000000, float %.031)
  %.032 = select i1 %.not13, float %63, float %62
  %64 = fmul float %.028, 3.000000e+00
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %65, 0
  %66 = call float @llvm.nvvm.fma.rn.ftz.f(float %64, float %.023, float %.032) #139
  %67 = call float @llvm.fma.f32(float %64, float %.023, float %.032)
  %.020 = select i1 %.not14, float %67, float %66
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %68, 0
  %69 = call float @llvm.nvvm.fma.rn.ftz.f(float %.028, float %28, float %.020) #139
  %70 = call float @llvm.fma.f32(float %.028, float %28, float %.020)
  %.018 = select i1 %.not15, float %70, float %69
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %71, 0
  %72 = call float @llvm.nvvm.add.rn.ftz.f(float %.029, float %.018) #139
  %73 = fadd float %.029, %.018
  %.015 = select i1 %.not16, float %73, float %72
  %74 = fneg float %.029
  %75 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %75, 0
  %76 = call float @llvm.nvvm.add.rn.ftz.f(float %.015, float %74) #139
  %77 = fsub float %.015, %.029
  %.016 = select i1 %.not17, float %77, float %76
  %78 = fneg float %.016
  %79 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %79, 0
  %80 = call float @llvm.nvvm.add.rn.ftz.f(float %.018, float %78) #139
  %81 = fsub float %.018, %.016
  %.017 = select i1 %.not18, float %81, float %80
  %82 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %82, 0
  %83 = call float @llvm.nvvm.mul.rn.ftz.f(float %.015, float %b) #139
  %84 = fmul float %.015, %b
  %.034 = select i1 %.not19, float %84, float %83
  %85 = fneg float %.034
  %86 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %86, 0
  %87 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float %b, float %85) #139
  %88 = call float @llvm.fma.f32(float %.015, float %b, float %85)
  %.035 = select i1 %.not20, float %88, float %87
  %89 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %89, 0
  %90 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %b, float %.035) #139
  %91 = call float @llvm.fma.f32(float %.017, float %b, float %.035)
  %.036 = select i1 %.not21, float %91, float %90
  %92 = call float @llvm.round.f32(float %.034)
  %93 = fsub float %.034, %92
  %94 = fadd float %93, %.036
  %95 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %95, 0
  %96 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F23F971C0000000, float %94, float noundef 0x3F55F0BDA0000000) #139
  %97 = call float @llvm.fma.f32(float %94, float noundef 0x3F23F971C0000000, float noundef 0x3F55F0BDA0000000)
  %.026 = select i1 %.not22, float %97, float %96
  %98 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %98, 0
  %99 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %94, float noundef 0x3F83B30AC0000000) #139
  %100 = call float @llvm.fma.f32(float %.026, float %94, float noundef 0x3F83B30AC0000000)
  %.019 = select i1 %.not23, float %100, float %99
  %101 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %101, 0
  %102 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %94, float noundef 0x3FAC6AF760000000) #139
  %103 = call float @llvm.fma.f32(float %.019, float %94, float noundef 0x3FAC6AF760000000)
  %.012 = select i1 %.not24, float %103, float %102
  %104 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %104, 0
  %105 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %94, float noundef 0x3FCEBFBD80000000) #139
  %106 = call float @llvm.fma.f32(float %.012, float %94, float noundef 0x3FCEBFBD80000000)
  %.09 = select i1 %.not25, float %106, float %105
  %107 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not26 = icmp eq i32 %107, 0
  %108 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %94, float noundef 0x3FE62E4300000000) #139
  %109 = call float @llvm.fma.f32(float %.09, float %94, float noundef 0x3FE62E4300000000)
  %.02 = select i1 %.not26, float %109, float %108
  %110 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %110, 0
  %111 = call float @llvm.nvvm.fma.rn.ftz.f(float %.02, float %94, float noundef 1.000000e+00) #139
  %112 = call float @llvm.fma.f32(float %.02, float %94, float noundef 1.000000e+00)
  %.033 = select i1 %.not27, float %112, float %111
  %113 = fptosi float %92 to i32
  %114 = fcmp ogt float %92, 0.000000e+00
  %115 = select i1 %114, i32 0, i32 -2097152000
  %116 = add nsw i32 %115, 2130706432
  %117 = bitcast i32 %116 to float
  %118 = fmul float %.033, %117
  %119 = shl i32 %113, 23
  %120 = sub i32 %119, %115
  %121 = bitcast i32 %120 to float
  %122 = fmul float %118, %121
  %123 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %123, 0
  %124 = call float @llvm.nvvm.fabs.ftz.f(float %.034) #139
  %125 = call float @llvm.fabs.f32(float %.034)
  %.0 = select i1 %.not28, float %125, float %124
  %126 = fcmp ogt float %.0, 1.520000e+02
  %127 = fcmp olt float %.034, 0.000000e+00
  %128 = select i1 %127, float 0.000000e+00, float 0x7FF0000000000000
  %t.i.0 = select i1 %126, float %128, float %122
  %129 = fcmp oeq float %a, 1.000000e+00
  %130 = fcmp oeq float %b, 0.000000e+00
  %or.cond = select i1 %129, i1 true, i1 %130
  br i1 %or.cond, label %__internal_powf_special_cases.exit, label %__nv_isnanf.exit.i

__nv_isnanf.exit.i:                               ; preds = %__nv_fabsf.exit
  %131 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not29 = icmp eq i32 %131, 0
  %.03 = select i1 %.not29, float %10, float %9
  %132 = fcmp ugt float %.03, 0x7FF0000000000000
  br i1 %132, label %__nv_fadd_rn.exit.i, label %__nv_isnanf.exit2.i

__nv_isnanf.exit2.i:                              ; preds = %__nv_isnanf.exit.i
  %133 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %133, 0
  %134 = call float @llvm.nvvm.fabs.ftz.f(float %b) #139
  %135 = call float @llvm.fabs.f32(float %b)
  %.04 = select i1 %.not30, float %135, float %134
  %136 = fcmp ugt float %.04, 0x7FF0000000000000
  br i1 %136, label %__nv_fadd_rn.exit.i, label %140

__nv_fadd_rn.exit.i:                              ; preds = %__nv_isnanf.exit2.i, %__nv_isnanf.exit.i
  %137 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not34 = icmp eq i32 %137, 0
  %138 = call float @llvm.nvvm.add.rn.ftz.f(float %a, float %b) #139
  %139 = fadd float %a, %b
  %.05 = select i1 %.not34, float %139, float %138
  br label %__internal_powf_special_cases.exit

140:                                              ; preds = %__nv_isnanf.exit2.i
  %141 = fcmp oeq float %a, 0.000000e+00
  br i1 %141, label %144, label %__nv_isinff.exit.i

__nv_isinff.exit.i:                               ; preds = %140
  %142 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %142, 0
  %.06 = select i1 %.not31, float %10, float %9
  %143 = fcmp oeq float %.06, 0x7FF0000000000000
  br i1 %143, label %144, label %151

144:                                              ; preds = %__nv_isinff.exit.i, %140
  %145 = fadd float %a, %a
  %146 = bitcast float %145 to i32
  %147 = fcmp olt float %b, 0.000000e+00
  %148 = xor i32 %146, 2139095040
  %ti.i.0 = select i1 %147, i32 %148, i32 %146
  %149 = and i32 %ti.i.0, 2147483647
  %ti.i.1 = select i1 %7, i32 %ti.i.0, i32 %149
  %150 = bitcast i32 %ti.i.1 to float
  br label %__internal_powf_special_cases.exit

151:                                              ; preds = %__nv_isinff.exit.i
  %152 = fcmp oeq float %a, -1.000000e+00
  br i1 %152, label %__nv_isinff.exit7.i, label %155

__nv_isinff.exit7.i:                              ; preds = %151
  %153 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not33 = icmp eq i32 %153, 0
  %.08 = select i1 %.not33, float %135, float %134
  %154 = fcmp oeq float %.08, 0x7FF0000000000000
  br i1 %154, label %__internal_powf_special_cases.exit, label %155

155:                                              ; preds = %__nv_isinff.exit7.i, %151
  %156 = fcmp olt float %a, 0.000000e+00
  br i1 %156, label %__nv_floorf.exit.i, label %__internal_powf_special_cases.exit

__nv_floorf.exit.i:                               ; preds = %155
  %157 = fneg float %t.i.0
  %spec.select = select i1 %7, float %157, float %t.i.0
  %158 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %158, 0
  %159 = call float @llvm.nvvm.floor.ftz.f(float %b) #139
  %160 = call float @llvm.floor.f32(float %b)
  %.07 = select i1 %.not32, float %160, float %159
  %161 = fcmp une float %.07, %b
  %.1 = select i1 %161, float 0x7FFFFFFFE0000000, float %spec.select
  br label %__internal_powf_special_cases.exit

__internal_powf_special_cases.exit:               ; preds = %__nv_fadd_rn.exit.i, %__nv_isinff.exit7.i, %__nv_floorf.exit.i, %155, %144, %__nv_fabsf.exit
  %.6 = phi float [ 1.000000e+00, %__nv_fabsf.exit ], [ %.05, %__nv_fadd_rn.exit.i ], [ %150, %144 ], [ 1.000000e+00, %__nv_isinff.exit7.i ], [ %.1, %__nv_floorf.exit.i ], [ %t.i.0, %155 ]
  ret float %.6
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_tgammaf(float %a) local_unnamed_addr #67 {
__nv_fabsf.exit:
  %0 = fcmp olt float %a, 0.000000e+00
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  %2 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %3 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %3, float %2
  %4 = fcmp ult float %.0, 1.500000e+00
  br i1 %4, label %__internal_fmad.exit.i1, label %__nv_fabsf.exit.i

__nv_fabsf.exit.i:                                ; preds = %__nv_fabsf.exit
  %5 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not12 = icmp eq i32 %5, 0
  %.03 = select i1 %.not12, float %3, float %2
  %6 = fcmp ogt float %.03, 0x40448CCCC0000000
  %x.i.0 = select i1 %6, float 0x40448CCCC0000000, float %a
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not13 = icmp eq i32 %7, 0
  %8 = call float @llvm.nvvm.fabs.ftz.f(float %x.i.0) #139
  %9 = call float @llvm.fabs.f32(float %x.i.0)
  %.04 = select i1 %.not13, float %9, float %8
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not14 = icmp eq i32 %10, 0
  %11 = fcmp olt float %.04, 0x3810000000000000
  %12 = fmul float %.04, 0x4170000000000000
  %13 = select i1 %.not14, i1 %11, i1 false
  %.1 = select i1 %13, float %12, float %.04
  %expo.i.i.1 = select i1 %13, float -2.400000e+01, float 0.000000e+00
  %14 = bitcast float %.1 to i32
  %15 = add i32 %14, -1060439283
  %16 = and i32 %15, -8388608
  %17 = sub i32 %14, %16
  %18 = bitcast i32 %17 to float
  %19 = sitofp i32 %16 to float
  %20 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not15 = icmp eq i32 %20, 0
  %21 = call float @llvm.nvvm.fma.rn.ftz.f(float %19, float noundef 0x3E80000000000000, float noundef %expo.i.i.1) #139
  %22 = call float @llvm.fma.f32(float %19, float noundef 0x3E80000000000000, float noundef %expo.i.i.1)
  %.013 = select i1 %.not15, float %22, float %21
  %23 = fadd float %18, -1.000000e+00
  %24 = fadd float %18, 1.000000e+00
  %25 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %24) #141, !srcloc !73
  %26 = fmul float %23, 2.000000e+00
  %27 = fmul float %25, %26
  %28 = fmul float %27, %27
  %29 = fsub float %23, %27
  %30 = fmul float %29, 2.000000e+00
  %31 = fneg float %27
  %32 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not16 = icmp eq i32 %32, 0
  %33 = call float @llvm.nvvm.fma.rn.ftz.f(float %31, float %23, float %30) #139
  %34 = call float @llvm.fma.f32(float %31, float %23, float %30)
  %.014 = select i1 %.not16, float %34, float %33
  %35 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not17 = icmp eq i32 %35, 0
  %36 = call float @llvm.nvvm.mul.rn.ftz.f(float %25, float %.014) #139
  %37 = fmul float %25, %.014
  %.015 = select i1 %.not17, float %37, float %36
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not18 = icmp eq i32 %38, 0
  %39 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F45865C80000000, float %28, float noundef 0x3F6A5CFB60000000) #139
  %40 = call float @llvm.fma.f32(float %28, float noundef 0x3F45865C80000000, float noundef 0x3F6A5CFB60000000)
  %.016 = select i1 %.not18, float %40, float %39
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not19 = icmp eq i32 %41, 0
  %42 = call float @llvm.nvvm.fma.rn.ftz.f(float %.016, float %28, float noundef 0x3F92776E60000000) #139
  %43 = call float @llvm.fma.f32(float %.016, float %28, float noundef 0x3F92776E60000000)
  %.017 = select i1 %.not19, float %43, float %42
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not20 = icmp eq i32 %44, 0
  %45 = call float @llvm.nvvm.fma.rn.ftz.f(float %.017, float %28, float noundef 0x3FBEC709E0000000) #139
  %46 = call float @llvm.fma.f32(float %.017, float %28, float noundef 0x3FBEC709E0000000)
  %.018 = select i1 %.not20, float %46, float %45
  %47 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not21 = icmp eq i32 %47, 0
  %48 = call float @llvm.nvvm.mul.rn.ftz.f(float %.018, float %28) #139
  %49 = fmul float %28, %.018
  %.020 = select i1 %.not21, float %49, float %48
  %50 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not22 = icmp eq i32 %50, 0
  %51 = call float @llvm.nvvm.fma.rn.ftz.f(float %27, float noundef 0x3FF7154760000000, float %.013) #139
  %52 = call float @llvm.fma.f32(float %27, float noundef 0x3FF7154760000000, float %.013)
  %.021 = select i1 %.not22, float %52, float %51
  %53 = fsub float %.013, %.021
  %54 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %54, 0
  %55 = call float @llvm.nvvm.fma.rn.ftz.f(float %27, float noundef 0x3FF7154760000000, float %53) #139
  %56 = call float @llvm.fma.f32(float %27, float noundef 0x3FF7154760000000, float %53)
  %.022 = select i1 %.not23, float %56, float %55
  %57 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not24 = icmp eq i32 %57, 0
  %58 = call float @llvm.nvvm.fma.rn.ftz.f(float %.015, float noundef 0x3FF7154760000000, float %.022) #139
  %59 = call float @llvm.fma.f32(float %.015, float noundef 0x3FF7154760000000, float %.022)
  %.023 = select i1 %.not24, float %59, float %58
  %60 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not25 = icmp eq i32 %60, 0
  %61 = call float @llvm.nvvm.fma.rn.ftz.f(float %27, float noundef 0x3E54ABC680000000, float %.023) #139
  %62 = call float @llvm.fma.f32(float %27, float noundef 0x3E54ABC680000000, float %.023)
  %.024 = select i1 %.not25, float %62, float %61
  %63 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not26 = icmp eq i32 %63, 0
  %64 = call float @llvm.nvvm.fma.rn.ftz.f(float %.020, float %27, float %.024) #139
  %65 = call float @llvm.fma.f32(float %.020, float %27, float %.024)
  %.010 = select i1 %.not26, float %65, float %64
  %66 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not27 = icmp eq i32 %66, 0
  %67 = call float @llvm.nvvm.add.rn.ftz.f(float %.021, float %.010) #139
  %68 = fadd float %.021, %.010
  %.07 = select i1 %.not27, float %68, float %67
  %69 = fneg float %.021
  %70 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not28 = icmp eq i32 %70, 0
  %71 = call float @llvm.nvvm.add.rn.ftz.f(float %.07, float %69) #139
  %72 = fsub float %.07, %.021
  %.08 = select i1 %.not28, float %72, float %71
  %73 = fneg float %.08
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not29 = icmp eq i32 %74, 0
  %75 = call float @llvm.nvvm.add.rn.ftz.f(float %.010, float %73) #139
  %76 = fsub float %.010, %.08
  %.09 = select i1 %.not29, float %76, float %75
  %77 = fadd float %.04, -5.000000e-01
  %78 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not30 = icmp eq i32 %78, 0
  %79 = call float @llvm.nvvm.mul.rn.ftz.f(float %.07, float %77) #139
  %80 = fmul float %77, %.07
  %.025 = select i1 %.not30, float %80, float %79
  %81 = fneg float %.025
  %82 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not31 = icmp eq i32 %82, 0
  %83 = call float @llvm.nvvm.fma.rn.ftz.f(float %.07, float %77, float %81) #139
  %84 = call float @llvm.fma.f32(float %.07, float %77, float %81)
  %.027 = select i1 %.not31, float %84, float %83
  %85 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not32 = icmp eq i32 %85, 0
  %86 = call float @llvm.nvvm.fma.rn.ftz.f(float %.09, float %77, float %.027) #139
  %87 = call float @llvm.fma.f32(float %.09, float %77, float %.027)
  %.028 = select i1 %.not32, float %87, float %86
  %88 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not33 = icmp eq i32 %88, 0
  %89 = call float @llvm.nvvm.mul.rn.ftz.f(float noundef 0x3FF7154760000000, float %.04) #139
  %90 = fmul float %.04, 0x3FF7154760000000
  %.029 = select i1 %.not33, float %90, float %89
  %91 = fneg float %.029
  %92 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not34 = icmp eq i32 %92, 0
  %93 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FF7154760000000, float %.04, float %91) #139
  %94 = call float @llvm.fma.f32(float %.04, float noundef 0x3FF7154760000000, float %91)
  %.030 = select i1 %.not34, float %94, float %93
  %95 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not35 = icmp eq i32 %95, 0
  %96 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3E54AE0C00000000, float %.04, float %.030) #139
  %97 = call float @llvm.fma.f32(float %.04, float noundef 0x3E54AE0C00000000, float %.030)
  %.031 = select i1 %.not35, float %97, float %96
  %98 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not36 = icmp eq i32 %98, 0
  %99 = call float @llvm.nvvm.add.rn.ftz.f(float %.025, float %91) #139
  %100 = fsub float %.025, %.029
  %.032 = select i1 %.not36, float %100, float %99
  %101 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not37 = icmp eq i32 %101, 0
  %102 = call float @llvm.nvvm.add.rn.ftz.f(float %.032, float %.029) #139
  %103 = fadd float %.029, %.032
  %.034 = select i1 %.not37, float %103, float %102
  %104 = fneg float %.034
  %105 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not38 = icmp eq i32 %105, 0
  %106 = call float @llvm.nvvm.add.rn.ftz.f(float %.032, float %104) #139
  %107 = fsub float %.032, %.034
  %.035 = select i1 %.not38, float %107, float %106
  %108 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not39 = icmp eq i32 %108, 0
  %109 = call float @llvm.nvvm.add.rn.ftz.f(float %.025, float %104) #139
  %110 = fsub float %.025, %.034
  %.036 = select i1 %.not39, float %110, float %109
  %111 = fneg float %.035
  %112 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not40 = icmp eq i32 %112, 0
  %113 = call float @llvm.nvvm.add.rn.ftz.f(float %91, float %111) #139
  %114 = fsub float %111, %.029
  %.037 = select i1 %.not40, float %114, float %113
  %115 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not41 = icmp eq i32 %115, 0
  %116 = call float @llvm.nvvm.add.rn.ftz.f(float %.036, float %.037) #139
  %117 = fadd float %.036, %.037
  %.038 = select i1 %.not41, float %117, float %116
  %118 = fsub float %.028, %.031
  %119 = fadd float %118, %.038
  br i1 %0, label %120, label %__nv_fmaf.exit2.i.i

120:                                              ; preds = %__nv_fabsf.exit.i
  %121 = fneg float %.032
  %122 = fneg float %119
  %123 = fcmp ogt float %.04, 3.300000e+01
  %124 = fsub float 4.800000e+01, %.032
  %arg.i.0.0 = select i1 %123, float %124, float %121
  br label %__nv_fmaf.exit2.i.i

__nv_fmaf.exit2.i.i:                              ; preds = %120, %__nv_fabsf.exit.i
  %arg.i.0.1 = phi float [ %arg.i.0.0, %120 ], [ %.032, %__nv_fabsf.exit.i ]
  %arg.i.1.0 = phi float [ %122, %120 ], [ %119, %__nv_fabsf.exit.i ]
  %125 = call float @llvm.round.f32(float %arg.i.0.1)
  %126 = fsub float %arg.i.0.1, %125
  %127 = fadd float %arg.i.1.0, %126
  %128 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not42 = icmp eq i32 %128, 0
  %129 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F23F971C0000000, float %127, float noundef 0x3F55F0BDA0000000) #139
  %130 = call float @llvm.fma.f32(float %127, float noundef 0x3F23F971C0000000, float noundef 0x3F55F0BDA0000000)
  %.041 = select i1 %.not42, float %130, float %129
  %131 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not43 = icmp eq i32 %131, 0
  %132 = call float @llvm.nvvm.fma.rn.ftz.f(float %.041, float %127, float noundef 0x3F83B30AC0000000) #139
  %133 = call float @llvm.fma.f32(float %.041, float %127, float noundef 0x3F83B30AC0000000)
  %.042 = select i1 %.not43, float %133, float %132
  %134 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not44 = icmp eq i32 %134, 0
  %135 = call float @llvm.nvvm.fma.rn.ftz.f(float %.042, float %127, float noundef 0x3FAC6AF760000000) #139
  %136 = call float @llvm.fma.f32(float %.042, float %127, float noundef 0x3FAC6AF760000000)
  %.043 = select i1 %.not44, float %136, float %135
  %137 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not45 = icmp eq i32 %137, 0
  %138 = call float @llvm.nvvm.fma.rn.ftz.f(float %.043, float %127, float noundef 0x3FCEBFBD80000000) #139
  %139 = call float @llvm.fma.f32(float %.043, float %127, float noundef 0x3FCEBFBD80000000)
  %.044 = select i1 %.not45, float %139, float %138
  %140 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not46 = icmp eq i32 %140, 0
  %141 = call float @llvm.nvvm.fma.rn.ftz.f(float %.044, float %127, float noundef 0x3FE62E4300000000) #139
  %142 = call float @llvm.fma.f32(float %.044, float %127, float noundef 0x3FE62E4300000000)
  %.046 = select i1 %.not46, float %142, float %141
  %143 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not47 = icmp eq i32 %143, 0
  %144 = call float @llvm.nvvm.fma.rn.ftz.f(float %.046, float %127, float noundef 1.000000e+00) #139
  %145 = call float @llvm.fma.f32(float %.046, float %127, float noundef 1.000000e+00)
  %.040 = select i1 %.not47, float %145, float %144
  %146 = fptosi float %125 to i32
  %147 = fcmp ogt float %125, 0.000000e+00
  %148 = select i1 %147, i32 0, i32 -2097152000
  %149 = add nsw i32 %148, 2130706432
  %150 = bitcast i32 %149 to float
  %151 = fmul float %.040, %150
  %152 = shl i32 %146, 23
  %153 = sub i32 %152, %148
  %154 = bitcast i32 %153 to float
  %155 = fmul float %151, %154
  %156 = fmul float %155, 0x40040D9320000000
  %157 = call float asm "rcp.approx.ftz.f32 $0,$1;", "=f,f"(float %.04) #141, !srcloc !73
  %158 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not48 = icmp eq i32 %158, 0
  %159 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3F11EF2E20000000, float %157, float noundef 0xBF0A87DB00000000) #139
  %160 = call float @llvm.fma.f32(float %157, float noundef 0x3F11EF2E20000000, float noundef 0xBF0A87DB00000000)
  %.047 = select i1 %.not48, float %160, float %159
  %161 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not49 = icmp eq i32 %161, 0
  %162 = call float @llvm.nvvm.fma.rn.ftz.f(float %.047, float %157, float noundef 0xBF3BB4D7C0000000) #139
  %163 = call float @llvm.fma.f32(float %.047, float %157, float noundef 0xBF3BB4D7C0000000)
  %.048 = select i1 %.not49, float %163, float %162
  %164 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not50 = icmp eq i32 %164, 0
  %165 = call float @llvm.nvvm.fma.rn.ftz.f(float %.048, float %157, float noundef 0x3F504157C0000000) #139
  %166 = call float @llvm.fma.f32(float %.048, float %157, float noundef 0x3F504157C0000000)
  %.049 = select i1 %.not50, float %166, float %165
  %167 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not51 = icmp eq i32 %167, 0
  %168 = call float @llvm.nvvm.fma.rn.ftz.f(float %.049, float %157, float noundef 0xBF32415FA0000000) #139
  %169 = call float @llvm.fma.f32(float %.049, float %157, float noundef 0xBF32415FA0000000)
  %.050 = select i1 %.not51, float %169, float %168
  %170 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not52 = icmp eq i32 %170, 0
  %171 = call float @llvm.nvvm.fma.rn.ftz.f(float %.050, float %157, float noundef 0xBF65E9AC80000000) #139
  %172 = call float @llvm.fma.f32(float %.050, float %157, float noundef 0xBF65E9AC80000000)
  %.051 = select i1 %.not52, float %172, float %171
  %173 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not53 = icmp eq i32 %173, 0
  %174 = call float @llvm.nvvm.fma.rn.ftz.f(float %.051, float %157, float noundef 0x3F6C70E640000000) #139
  %175 = call float @llvm.fma.f32(float %.051, float %157, float noundef 0x3F6C70E640000000)
  %.053 = select i1 %.not53, float %175, float %174
  %176 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not54 = icmp eq i32 %176, 0
  %177 = call float @llvm.nvvm.fma.rn.ftz.f(float %.053, float %157, float noundef 0x3FB5555580000000) #139
  %178 = call float @llvm.fma.f32(float %.053, float %157, float noundef 0x3FB5555580000000)
  %.054 = select i1 %.not54, float %178, float %177
  br i1 %0, label %__internal_fmad.exit35.i, label %__internal_fmad.exit45.i

__internal_fmad.exit35.i:                         ; preds = %__nv_fmaf.exit2.i.i
  %179 = fmul float %157, %.054
  %180 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not56 = icmp eq i32 %180, 0
  %181 = call float @llvm.nvvm.fma.rn.ftz.f(float %179, float %x.i.0, float %x.i.0) #139
  %182 = call float @llvm.fma.f32(float %179, float %x.i.0, float %x.i.0)
  %.055 = select i1 %.not56, float %182, float %181
  %183 = fadd float %.04, %.04
  %184 = call float @llvm.round.f32(float %183)
  %185 = fptosi float %184 to i32
  %186 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not57 = icmp eq i32 %186, 0
  %187 = call float @llvm.nvvm.fma.rn.ftz.f(float %184, float noundef -5.000000e-01, float %.04) #139
  %188 = call float @llvm.fma.f32(float %184, float noundef -5.000000e-01, float %.04)
  %.075 = select i1 %.not57, float %188, float %187
  %189 = and i32 %185, 1
  %190 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not58 = icmp eq i32 %190, 0
  %191 = call float @llvm.nvvm.mul.rn.ftz.f(float %.075, float %.075) #139
  %192 = fmul float %.075, %.075
  %.056 = select i1 %.not58, float %192, float %191
  %193 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %194 = icmp sgt i32 %193, 820
  br i1 %194, label %__internal_fmad.exit.i.i.i, label %__internal_fmad.exit9.i.i.i

__internal_fmad.exit.i.i.i:                       ; preds = %__internal_fmad.exit35.i
  %195 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not76 = icmp eq i32 %195, 0
  %196 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBFE2F59920000000, float %.056, float noundef 0x400466B200000000) #139
  %197 = call float @llvm.fma.f32(float %.056, float noundef 0xBFE2F59920000000, float noundef 0x400466B200000000)
  %.057 = select i1 %.not76, float %197, float %196
  %198 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not77 = icmp eq i32 %198, 0
  %199 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FCD09C240000000, float %.056, float noundef 0xBFF55A5C00000000) #139
  %200 = call float @llvm.fma.f32(float %.056, float noundef 0x3FCD09C240000000, float noundef 0xBFF55A5C00000000)
  %.058 = select i1 %.not77, float %200, float %199
  %201 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not78 = icmp eq i32 %201, 0
  %202 = call float @llvm.nvvm.fma.rn.ftz.f(float %.057, float %.056, float noundef 0xC014ABBEC0000000) #139
  %203 = call float @llvm.fma.f32(float %.057, float %.056, float noundef 0xC014ABBEC0000000)
  %.060 = select i1 %.not78, float %203, float %202
  %204 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not79 = icmp eq i32 %204, 0
  %205 = call float @llvm.nvvm.fma.rn.ftz.f(float %.058, float %.056, float noundef 0x40103C19E0000000) #139
  %206 = call float @llvm.fma.f32(float %.058, float %.056, float noundef 0x40103C19E0000000)
  %.061 = select i1 %.not79, float %206, float %205
  %207 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not80 = icmp eq i32 %207, 0
  %208 = call float @llvm.nvvm.fma.rn.ftz.f(float %.056, float %.075, float noundef 0.000000e+00) #139
  %209 = call float @llvm.fma.f32(float %.056, float %.075, float noundef 0.000000e+00)
  %.062 = select i1 %.not80, float %209, float %208
  %210 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not81 = icmp eq i32 %210, 0
  %211 = call float @llvm.nvvm.fma.rn.ftz.f(float %.061, float %.056, float noundef 0xC013BD3CC0000000) #139
  %212 = call float @llvm.fma.f32(float %.061, float %.056, float noundef 0xC013BD3CC0000000)
  %.063 = select i1 %.not81, float %212, float %211
  %213 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not82 = icmp eq i32 %213, 0
  %214 = call float @llvm.nvvm.fma.rn.ftz.f(float %.060, float %.062, float noundef 0.000000e+00) #139
  %215 = call float @llvm.fma.f32(float %.060, float %.062, float noundef 0.000000e+00)
  %.064 = select i1 %.not82, float %215, float %214
  %216 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not83 = icmp eq i32 %216, 0
  %217 = call float @llvm.nvvm.fma.rn.ftz.f(float %.063, float %.056, float noundef 1.000000e+00) #139
  %218 = call float @llvm.fma.f32(float %.063, float %.056, float noundef 1.000000e+00)
  %.065 = select i1 %.not83, float %218, float %217
  %219 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not84 = icmp eq i32 %219, 0
  %220 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400921FB60000000, float %.075, float %.064) #139
  %221 = call float @llvm.fma.f32(float %.075, float noundef 0x400921FB60000000, float %.064)
  %.067 = select i1 %.not84, float %221, float %220
  br label %__internal_tgammaf_sinpi.exit.i

__internal_fmad.exit9.i.i.i:                      ; preds = %__internal_fmad.exit35.i
  %.not59 = icmp eq i32 %189, 0
  %..075 = select i1 %.not59, float %.075, float 1.000000e+00
  %222 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not60 = icmp eq i32 %222, 0
  %223 = call float @llvm.nvvm.fma.rn.ftz.f(float %.056, float %..075, float noundef 0.000000e+00) #139
  %224 = call float @llvm.fma.f32(float %.056, float %..075, float noundef 0.000000e+00)
  %.068 = select i1 %.not60, float %224, float %223
  br i1 %.not59, label %__internal_fmad.exit11.i.i.i, label %__internal_fmad.exit10.i.i.i

__internal_fmad.exit10.i.i.i:                     ; preds = %__internal_fmad.exit9.i.i.i
  %225 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not75 = icmp eq i32 %225, 0
  %226 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x3FCD09C240000000, float %.056, float noundef 0xBFF55A5C00000000) #139
  %227 = call float @llvm.fma.f32(float %.056, float noundef 0x3FCD09C240000000, float noundef 0xBFF55A5C00000000)
  %.069 = select i1 %.not75, float %227, float %226
  br label %__internal_fmad.exit11.i.i.i

__internal_fmad.exit11.i.i.i:                     ; preds = %__internal_fmad.exit9.i.i.i, %__internal_fmad.exit10.i.i.i
  %228 = phi float [ 0.000000e+00, %__internal_fmad.exit9.i.i.i ], [ 1.000000e+00, %__internal_fmad.exit10.i.i.i ]
  %229 = phi float [ 0xC014ABBEC0000000, %__internal_fmad.exit9.i.i.i ], [ 0xC013BD3CC0000000, %__internal_fmad.exit10.i.i.i ]
  %230 = phi float [ 0x400466B200000000, %__internal_fmad.exit9.i.i.i ], [ 0x40103C19E0000000, %__internal_fmad.exit10.i.i.i ]
  %231 = phi float [ 0xBFE2F59920000000, %__internal_fmad.exit9.i.i.i ], [ %.069, %__internal_fmad.exit10.i.i.i ]
  %232 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not63 = icmp eq i32 %232, 0
  %233 = call float @llvm.nvvm.fma.rn.ftz.f(float %231, float %.056, float noundef %230) #139
  %234 = call float @llvm.fma.f32(float %231, float %.056, float noundef %230)
  %.070 = select i1 %.not63, float %234, float %233
  %235 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not65 = icmp eq i32 %235, 0
  %236 = call float @llvm.nvvm.fma.rn.ftz.f(float %.070, float %.056, float noundef %229) #139
  %237 = call float @llvm.fma.f32(float %.070, float %.056, float noundef %229)
  %.071 = select i1 %.not65, float %237, float %236
  %238 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not67 = icmp eq i32 %238, 0
  %239 = call float @llvm.nvvm.fma.rn.ftz.f(float %.071, float %.068, float noundef %228) #139
  %240 = call float @llvm.fma.f32(float %.071, float %.068, float noundef %228)
  %.072 = select i1 %.not67, float %240, float %239
  %241 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not68 = icmp eq i32 %241, 0
  %242 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0x400921FB60000000, float %.075, float %.072) #139
  %243 = call float @llvm.fma.f32(float %.075, float noundef 0x400921FB60000000, float %.072)
  %.074 = select i1 %.not68, float %243, float %242
  br label %__internal_tgammaf_sinpi.exit.i

__internal_tgammaf_sinpi.exit.i:                  ; preds = %__internal_fmad.exit11.i.i.i, %__internal_fmad.exit.i.i.i
  %rc.i.i.i.0 = phi float [ %.065, %__internal_fmad.exit.i.i.i ], [ %.072, %__internal_fmad.exit11.i.i.i ]
  %rs.i.i.i.0 = phi float [ %.067, %__internal_fmad.exit.i.i.i ], [ %.074, %__internal_fmad.exit11.i.i.i ]
  %.not69 = icmp eq i32 %189, 0
  %rc.i.i.i.0.rs.i.i.i.0 = select i1 %.not69, float %rs.i.i.i.0, float %rc.i.i.i.0
  %244 = and i32 %185, 2
  %.not70 = icmp eq i32 %244, 0
  %245 = fsub float 0.000000e+00, %rc.i.i.i.0.rs.i.i.i.0
  %246 = select i1 %.not70, float %rc.i.i.i.0.rs.i.i.i.0, float %245
  %247 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not71 = icmp eq i32 %247, 0
  %248 = call float @llvm.nvvm.mul.rn.ftz.f(float %.055, float %246) #139
  %249 = fmul float %.055, %246
  %.077 = select i1 %.not71, float %249, float %248
  %250 = fneg float %.077
  %251 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not72 = icmp eq i32 %251, 0
  %252 = call float @llvm.nvvm.fma.rn.ftz.f(float %.055, float %246, float %250) #139
  %253 = call float @llvm.fma.f32(float %.055, float %246, float %250)
  %.076 = select i1 %.not72, float %253, float %252
  %254 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not73 = icmp eq i32 %254, 0
  br i1 %.not73, label %257, label %255

255:                                              ; preds = %__internal_tgammaf_sinpi.exit.i
  %256 = call float @llvm.nvvm.div.approx.ftz.f(float noundef 1.000000e+00, float %.077) #139
  br label %__nv_fast_fdividef.exit.i

257:                                              ; preds = %__internal_tgammaf_sinpi.exit.i
  %258 = call float @llvm.nvvm.div.approx.f(float noundef 1.000000e+00, float %.077) #139
  br label %__nv_fast_fdividef.exit.i

__nv_fast_fdividef.exit.i:                        ; preds = %257, %255
  %.073 = phi float [ %256, %255 ], [ %258, %257 ]
  %259 = fneg float %.073
  %260 = fmul float %.073, %259
  %261 = fmul float %.076, %260
  %262 = fmul float %156, %261
  %263 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not74 = icmp eq i32 %263, 0
  %264 = call float @llvm.nvvm.fma.rn.ftz.f(float %156, float %.073, float %262) #139
  %265 = call float @llvm.fma.f32(float %156, float %.073, float %262)
  %.066 = select i1 %.not74, float %265, float %264
  %266 = fmul float %.066, 5.000000e-01
  %267 = fcmp ogt float %.04, 3.300000e+01
  %268 = fmul float %266, 0x3CF0000000000000
  %s.i.0 = select i1 %267, float %268, float %266
  br label %__internal_tgammaf_stirling.exit

__internal_fmad.exit45.i:                         ; preds = %__nv_fmaf.exit2.i.i
  %269 = fmul float %157, %.054
  %270 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not55 = icmp eq i32 %270, 0
  %271 = call float @llvm.nvvm.fma.rn.ftz.f(float %269, float %156, float %156) #139
  %272 = call float @llvm.fma.f32(float %269, float %156, float %156)
  %.059 = select i1 %.not55, float %272, float %271
  br label %__internal_tgammaf_stirling.exit

__internal_fmad.exit.i1:                          ; preds = %__nv_fabsf.exit
  %273 = fcmp olt float %a, -5.000000e-01
  %a. = select i1 %273, float %a, float 1.000000e+00
  %274 = fcmp oeq float %a, 0.000000e+00
  %275 = call float @llvm.round.f32(float %a)
  %276 = select i1 %274, float 0.000000e+00, float %275
  %277 = fsub float %a, %276
  %278 = fcmp ole float %a, 5.000000e-01
  %279 = select i1 %278, float %277, float 1.000000e+00
  %fy.0 = fmul float %a., %279
  %280 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %280, 0
  %281 = call float @llvm.nvvm.fma.rn.ftz.f(float noundef 0xBF5193ECC0000000, float %277, float noundef 0x3F7D207AE0000000) #139
  %282 = call float @llvm.fma.f32(float %277, float noundef 0xBF5193ECC0000000, float noundef 0x3F7D207AE0000000)
  %.052 = select i1 %.not1, float %282, float %281
  %283 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %283, 0
  %284 = call float @llvm.nvvm.fma.rn.ftz.f(float %.052, float %277, float noundef 0xBF83C01600000000) #139
  %285 = call float @llvm.fma.f32(float %.052, float %277, float noundef 0xBF83C01600000000)
  %.045 = select i1 %.not2, float %285, float %284
  %286 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not3 = icmp eq i32 %286, 0
  %287 = call float @llvm.nvvm.fma.rn.ftz.f(float %.045, float %277, float noundef 0xBFA598A440000000) #139
  %288 = call float @llvm.fma.f32(float %.045, float %277, float noundef 0xBFA598A440000000)
  %.039 = select i1 %.not3, float %288, float %287
  %289 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not4 = icmp eq i32 %289, 0
  %290 = call float @llvm.nvvm.fma.rn.ftz.f(float %.039, float %277, float noundef 0x3FC5513660000000) #139
  %291 = call float @llvm.fma.f32(float %.039, float %277, float noundef 0x3FC5513660000000)
  %.033 = select i1 %.not4, float %291, float %290
  %292 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not5 = icmp eq i32 %292, 0
  %293 = call float @llvm.nvvm.fma.rn.ftz.f(float %.033, float %277, float noundef 0xBFA58180A0000000) #139
  %294 = call float @llvm.fma.f32(float %.033, float %277, float noundef 0xBFA58180A0000000)
  %.026 = select i1 %.not5, float %294, float %293
  %295 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %295, 0
  %296 = call float @llvm.nvvm.fma.rn.ftz.f(float %.026, float %277, float noundef 0xBFE4FCF440000000) #139
  %297 = call float @llvm.fma.f32(float %.026, float %277, float noundef 0xBFE4FCF440000000)
  %.019 = select i1 %.not6, float %297, float %296
  %298 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %298, 0
  %299 = call float @llvm.nvvm.fma.rn.ftz.f(float %.019, float %277, float noundef 0x3FE2788D00000000) #139
  %300 = call float @llvm.fma.f32(float %.019, float %277, float noundef 0x3FE2788D00000000)
  %.012 = select i1 %.not7, float %300, float %299
  %301 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not8 = icmp eq i32 %301, 0
  %302 = call float @llvm.nvvm.fma.rn.ftz.f(float %.012, float %fy.0, float noundef 0.000000e+00) #139
  %303 = call float @llvm.fma.f32(float %.012, float %fy.0, float noundef 0.000000e+00)
  %.05 = select i1 %.not8, float %303, float %302
  %304 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not9 = icmp eq i32 %304, 0
  %305 = call float @llvm.nvvm.fma.rn.ftz.f(float %.05, float %277, float %fy.0) #139
  %306 = call float @llvm.fma.f32(float %.05, float %277, float %fy.0)
  %.02 = select i1 %.not9, float %306, float %305
  %307 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not10 = icmp eq i32 %307, 0
  br i1 %.not10, label %310, label %308

308:                                              ; preds = %__internal_fmad.exit.i1
  %309 = call float @llvm.nvvm.div.approx.ftz.f(float noundef 1.000000e+00, float %.02) #139
  br label %__internal_tgammaf_stirling.exit

310:                                              ; preds = %__internal_fmad.exit.i1
  %311 = call float @llvm.nvvm.div.approx.f(float noundef 1.000000e+00, float %.02) #139
  br label %__internal_tgammaf_stirling.exit

__internal_tgammaf_stirling.exit:                 ; preds = %308, %310, %__nv_fast_fdividef.exit.i, %__internal_fmad.exit45.i
  %s.0 = phi float [ %s.i.0, %__nv_fast_fdividef.exit.i ], [ %.059, %__internal_fmad.exit45.i ], [ %309, %308 ], [ %311, %310 ]
  br i1 %0, label %312, label %320

312:                                              ; preds = %__internal_tgammaf_stirling.exit
  %313 = call float @llvm.trunc.f32(float %a)
  %314 = fcmp oeq float %313, %a
  br i1 %314, label %320, label %315

315:                                              ; preds = %312
  %316 = fcmp olt float %a, 0xC0448CCCC0000000
  br i1 %316, label %317, label %320

317:                                              ; preds = %315
  %318 = fptosi float %a to i32
  %319 = and i32 %318, 1
  %.not11 = icmp eq i32 %319, 0
  %spec.select = select i1 %.not11, float %s.0, float 0.000000e+00
  br label %320

320:                                              ; preds = %312, %317, %315, %__internal_tgammaf_stirling.exit
  %s.3 = phi float [ %s.0, %__internal_tgammaf_stirling.exit ], [ %s.0, %315 ], [ %spec.select, %317 ], [ 0x7FFFFFFFE0000000, %312 ]
  ret float %s.3
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_roundf(float %a) local_unnamed_addr #68 {
__nv_fadd_rz.exit:
  %0 = bitcast float %a to i32
  %1 = and i32 %0, -2147483648
  %2 = or i32 %1, 1056964608
  %3 = bitcast i32 %2 to float
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %4, 0
  %5 = call float @llvm.nvvm.add.rz.ftz.f(float %a, float %3) #139
  %6 = call float @llvm.nvvm.add.rz.f(float %a, float %3) #139
  %.0 = select i1 %.not, float %6, float %5
  %7 = call float @llvm.trunc.f32(float %.0)
  ret float %7
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_fdimf(float %a, float %b) local_unnamed_addr #69 {
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %1, 0
  br i1 %.not, label %4, label %2

2:                                                ; preds = %0
  %3 = call float asm "sub.rn.ftz.f32 $0, $1, $2;", "=f,f,f"(float %a, float %b) #141, !srcloc !74
  br label %__nv_fsub_rn.exit

4:                                                ; preds = %0
  %5 = call float asm "sub.rn.f32 $0, $1, $2;", "=f,f,f"(float %a, float %b) #141, !srcloc !75
  br label %__nv_fsub_rn.exit

__nv_fsub_rn.exit:                                ; preds = %4, %2
  %res.i.0 = phi float [ %3, %2 ], [ %5, %4 ]
  %6 = fcmp ole float %a, %b
  %t.0 = select i1 %6, float 0.000000e+00, float %res.i.0
  ret float %t.0
}

; Function Attrs: noinline nounwind
define dso_local i32 @__nv_ilogbf(float %a) local_unnamed_addr #70 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = bitcast float %.0 to i32
  %4 = icmp ult i32 %3, 8388608
  br i1 %4, label %5, label %9

5:                                                ; preds = %__nv_fabsf.exit
  %6 = call i32 @llvm.ctlz.i32(i32 %3, i1 noundef false), !range !76
  %7 = sub nuw nsw i32 -118, %6
  %8 = icmp eq i32 %3, 0
  %expo.0 = select i1 %8, i32 -2147483648, i32 %7
  br label %14

9:                                                ; preds = %__nv_fabsf.exit
  %10 = lshr i32 %3, 23
  %11 = add nsw i32 %10, -127
  %12 = icmp eq i32 %3, 2139095040
  %spec.select = select i1 %12, i32 2147483647, i32 %11
  %13 = icmp ugt i32 %3, 2139095040
  %expo.2 = select i1 %13, i32 -2147483648, i32 %spec.select
  br label %14

14:                                               ; preds = %9, %5
  %expo.3 = phi i32 [ %expo.0, %5 ], [ %expo.2, %9 ]
  ret i32 %expo.3
}

; Function Attrs: noinline nounwind
define dso_local float @__nv_logbf(float %a) local_unnamed_addr #71 {
__nv_fabsf.exit:
  %0 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %0, 0
  %1 = call float @llvm.nvvm.fabs.ftz.f(float %a) #139
  %2 = call float @llvm.fabs.f32(float %a)
  %.0 = select i1 %.not, float %2, float %1
  %3 = bitcast float %.0 to i32
  %4 = icmp ult i32 %3, 8388608
  br i1 %4, label %5, label %10

5:                                                ; preds = %__nv_fabsf.exit
  %6 = call i32 @llvm.ctlz.i32(i32 %3, i1 noundef false), !range !76
  %7 = sub nuw nsw i32 -118, %6
  %8 = sitofp i32 %7 to float
  %9 = fcmp oeq float %a, 0.000000e+00
  %res.0 = select i1 %9, float 0xFFF0000000000000, float %8
  br label %16

10:                                               ; preds = %__nv_fabsf.exit
  %11 = lshr i32 %3, 23
  %12 = add nsw i32 %11, -127
  %13 = sitofp i32 %12 to float
  %14 = icmp ugt i32 %3, 2139095039
  %15 = fmul float %a, %a
  %spec.select = select i1 %14, float %15, float %13
  br label %16

16:                                               ; preds = %10, %5
  %res.2 = phi float [ %res.0, %5 ], [ %spec.select, %10 ]
  ret float %res.2
}

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.nvvm.round.d(double) #6

; Function Attrs: argmemonly noinline nosync nounwind
define internal fastcc double @__internal_trig_reduction_slowpathd(double %a, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %quadrant) unnamed_addr #72 {
  %result = alloca [5 x i64], align 8
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %2 = and i32 %1, -2147483648
  %3 = lshr i32 %1, 20
  %4 = and i32 %3, 2047
  %5 = icmp eq i32 %4, 2047
  br i1 %5, label %86, label %6

6:                                                ; preds = %0
  %7 = add nsw i32 %4, -1024
  %8 = bitcast double %a to i64
  %9 = shl i64 %8, 11
  %10 = or i64 %9, -9223372036854775808
  %11 = lshr i32 %7, 6
  %12 = sub nsw i32 15, %11
  %13 = sub nsw i32 19, %11
  %min.cond = icmp ult i32 %7, 128
  %min = select i1 %min.cond, i32 18, i32 %13
  %14 = icmp slt i32 %12, %min
  br i1 %14, label %.lr.ph, label %._crit_edge24

._crit_edge24:                                    ; preds = %6
  %.pre = add nsw i32 %11, -15
  br label %._crit_edge

.lr.ph:                                           ; preds = %6
  %.neg13 = add nsw i32 %11, -15
  br label %15

15:                                               ; preds = %.lr.ph, %15
  %q.019 = phi i32 [ %12, %.lr.ph ], [ %25, %15 ]
  %p.122.018 = phi i64 [ 0, %.lr.ph ], [ %21, %15 ]
  %16 = sext i32 %q.019 to i64
  %17 = getelementptr inbounds [18 x i64], [18 x i64] addrspace(1)* @__cudart_i2opi_d, i64 0, i64 %16
  %18 = load i64, i64 addrspace(1)* %17, align 8
  %19 = call { i64, i64 } asm "{\0A\09.reg .u32 r0, r1, r2, r3, alo, ahi, blo, bhi, clo, chi;\0A\09mov.b64         {alo,ahi}, $2;    \0A\09mov.b64         {blo,bhi}, $3;    \0A\09mov.b64         {clo,chi}, $4;    \0A\09mad.lo.cc.u32   r0, alo, blo, clo;\0A\09madc.hi.cc.u32  r1, alo, blo, chi;\0A\09madc.hi.u32     r2, alo, bhi,   0;\0A\09mad.lo.cc.u32   r1, alo, bhi,  r1;\0A\09madc.hi.cc.u32  r2, ahi, blo,  r2;\0A\09madc.hi.u32     r3, ahi, bhi,   0;\0A\09mad.lo.cc.u32   r1, ahi, blo,  r1;\0A\09madc.lo.cc.u32  r2, ahi, bhi,  r2;\0A\09addc.u32        r3,  r3,   0;     \0A\09mov.b64         $0, {r0,r1};      \0A\09mov.b64         $1, {r2,r3};      \0A\09}", "=l,=l,l,l,l"(i64 %18, i64 %10, i64 %p.122.018) #141, !srcloc !77
  %20 = extractvalue { i64, i64 } %19, 0
  %21 = extractvalue { i64, i64 } %19, 1
  %22 = add i32 %.neg13, %q.019
  %23 = sext i32 %22 to i64
  %24 = getelementptr inbounds [5 x i64], [5 x i64]* %result, i64 0, i64 %23
  store i64 %20, i64* %24, align 8, !noalias !68
  %25 = add nsw i32 %q.019, 1
  %exitcond.not = icmp eq i32 %25, %min
  br i1 %exitcond.not, label %._crit_edge, label %15, !llvm.loop !78

._crit_edge:                                      ; preds = %15, %._crit_edge24
  %.neg.pre-phi = phi i32 [ %.pre, %._crit_edge24 ], [ %.neg13, %15 ]
  %p.122.0.lcssa = phi i64 [ 0, %._crit_edge24 ], [ %21, %15 ]
  %q.0.lcssa = phi i32 [ %12, %._crit_edge24 ], [ %min, %15 ]
  %26 = add nsw i32 %.neg.pre-phi, %q.0.lcssa
  %27 = sext i32 %26 to i64
  %28 = getelementptr inbounds [5 x i64], [5 x i64]* %result, i64 0, i64 %27
  store i64 %p.122.0.lcssa, i64* %28, align 8, !noalias !68
  %29 = and i32 %3, 63
  %30 = getelementptr inbounds [5 x i64], [5 x i64]* %result, i64 0, i64 2
  %31 = load i64, i64* %30, align 8
  %32 = getelementptr inbounds [5 x i64], [5 x i64]* %result, i64 0, i64 3
  %33 = load i64, i64* %32, align 8
  %.not = icmp eq i32 %29, 0
  br i1 %.not, label %46, label %34

34:                                               ; preds = %._crit_edge
  %35 = sub nuw nsw i32 64, %29
  %36 = zext i32 %29 to i64
  %37 = shl i64 %33, %36
  %38 = zext i32 %35 to i64
  %39 = lshr i64 %31, %38
  %40 = or i64 %37, %39
  %41 = shl i64 %31, %36
  %42 = getelementptr inbounds [5 x i64], [5 x i64]* %result, i64 0, i64 1
  %43 = load i64, i64* %42, align 8
  %44 = lshr i64 %43, %38
  %45 = or i64 %44, %41
  br label %46

46:                                               ; preds = %34, %._crit_edge
  %hi.0 = phi i64 [ %40, %34 ], [ %33, %._crit_edge ]
  %lo.0 = phi i64 [ %45, %34 ], [ %31, %._crit_edge ]
  %47 = lshr i64 %hi.0, 62
  %48 = trunc i64 %47 to i32
  %49 = call i64 @llvm.fshl.i64(i64 %hi.0, i64 %lo.0, i64 noundef 2)
  %50 = shl i64 %lo.0, 2
  %51 = lshr i64 %49, 63
  %52 = trunc i64 %51 to i32
  %53 = add nuw nsw i32 %52, %48
  %.not1 = icmp eq i32 %2, 0
  %54 = sub nsw i32 0, %53
  %spec.select = select i1 %.not1, i32 %53, i32 %54
  store i32 %spec.select, i32* %quadrant, align 4, !noalias !68
  %.not2 = icmp sgt i64 %49, -1
  br i1 %.not2, label %60, label %55

55:                                               ; preds = %46
  %56 = call { i64, i64 } asm "{\0A\09.reg .u32 r0, r1, r2, r3, a0, a1, a2, a3, b0, b1, b2, b3;\0A\09mov.b64         {a0,a1}, $2;\0A\09mov.b64         {a2,a3}, $3;\0A\09mov.b64         {b0,b1}, $4;\0A\09mov.b64         {b2,b3}, $5;\0A\09sub.cc.u32      r0, a0, b0; \0A\09subc.cc.u32     r1, a1, b1; \0A\09subc.cc.u32     r2, a2, b2; \0A\09subc.u32        r3, a3, b3; \0A\09mov.b64         $0, {r0,r1};\0A\09mov.b64         $1, {r2,r3};\0A\09}", "=l,=l,l,l,l,l"(i64 0, i64 0, i64 %50, i64 %49) #141, !srcloc !79
  %57 = extractvalue { i64, i64 } %56, 0
  %58 = extractvalue { i64, i64 } %56, 1
  %59 = xor i32 %2, -2147483648
  br label %60

60:                                               ; preds = %55, %46
  %hi.1 = phi i64 [ %58, %55 ], [ %49, %46 ]
  %lo.1 = phi i64 [ %57, %55 ], [ %50, %46 ]
  %s.0 = phi i32 [ %59, %55 ], [ %2, %46 ]
  %ctlz = call i64 @llvm.ctlz.i64(i64 %hi.1, i1 noundef false), !range !80
  %.not3 = icmp eq i64 %ctlz, 0
  br i1 %.not3, label %66, label %61

61:                                               ; preds = %60
  %62 = shl i64 %hi.1, %ctlz
  %63 = sub nuw nsw i64 64, %ctlz
  %64 = lshr i64 %lo.1, %63
  %65 = or i64 %64, %62
  br label %66

66:                                               ; preds = %61, %60
  %hi.2 = phi i64 [ %65, %61 ], [ %hi.1, %60 ]
  %67 = call { i64, i64 } asm "{\0A\09.reg .u32 r0, r1, r2, r3, alo, ahi, blo, bhi;\0A\09mov.b64         {alo,ahi}, $2;   \0A\09mov.b64         {blo,bhi}, $3;   \0A\09mul.lo.u32      r0, alo, blo;    \0A\09mul.hi.u32      r1, alo, blo;    \0A\09mad.lo.cc.u32   r1, alo, bhi, r1;\0A\09madc.hi.u32     r2, alo, bhi,  0;\0A\09mad.lo.cc.u32   r1, ahi, blo, r1;\0A\09madc.hi.cc.u32  r2, ahi, blo, r2;\0A\09madc.hi.u32     r3, ahi, bhi,  0;\0A\09mad.lo.cc.u32   r2, ahi, bhi, r2;\0A\09addc.u32        r3, r3,  0;      \0A\09mov.b64         $0, {r0,r1};     \0A\09mov.b64         $1, {r2,r3};     \0A\09}", "=l,=l,l,l"(i64 %hi.2, i64 -3958705157555305931) #141, !srcloc !81
  %68 = extractvalue { i64, i64 } %67, 1
  %69 = icmp sgt i64 %68, 0
  br i1 %69, label %70, label %75

70:                                               ; preds = %66
  %71 = extractvalue { i64, i64 } %67, 0
  %72 = call { i64, i64 } asm "{\0A\09.reg .u32 r0, r1, r2, r3, a0, a1, a2, a3, b0, b1, b2, b3;\0A\09mov.b64         {a0,a1}, $2;\0A\09mov.b64         {a2,a3}, $3;\0A\09mov.b64         {b0,b1}, $4;\0A\09mov.b64         {b2,b3}, $5;\0A\09add.cc.u32      r0, a0, b0; \0A\09addc.cc.u32     r1, a1, b1; \0A\09addc.cc.u32     r2, a2, b2; \0A\09addc.u32        r3, a3, b3; \0A\09mov.b64         $0, {r0,r1};\0A\09mov.b64         $1, {r2,r3};\0A\09}", "=l,=l,l,l,l,l"(i64 %71, i64 %68, i64 %71, i64 %68) #141, !srcloc !82
  %73 = extractvalue { i64, i64 } %72, 1
  %74 = add nuw nsw i64 %ctlz, 1
  br label %75

75:                                               ; preds = %70, %66
  %hi.3 = phi i64 [ %73, %70 ], [ %68, %66 ]
  %e.0 = phi i64 [ %74, %70 ], [ %ctlz, %66 ]
  %76 = zext i32 %s.0 to i64
  %77 = shl nuw i64 %76, 32
  %.neg10 = mul nsw i64 %e.0, -4503599627370496
  %78 = add nsw i64 %.neg10, 4602678819172646912
  %79 = add i64 %hi.3, 1
  %80 = lshr i64 %79, 10
  %81 = add nuw nsw i64 %80, 1
  %82 = lshr i64 %81, 1
  %83 = add nuw nsw i64 %78, %82
  %84 = or i64 %83, %77
  %85 = bitcast i64 %84 to double
  br label %86

86:                                               ; preds = %0, %75
  %.0 = phi double [ %85, %75 ], [ %a, %0 ]
  ret double %.0
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_sin(double %a) local_unnamed_addr #73 {
  %q.i = alloca i32, align 4
  %q.i.i = alloca i32, align 4
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2 = icmp eq i32 %1, 350
  br i1 %2, label %6, label %3

3:                                                ; preds = %0
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %5 = icmp eq i32 %4, 370
  br i1 %5, label %6, label %61

6:                                                ; preds = %3, %0
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %8 = icmp eq i32 %7, 200
  br i1 %8, label %15, label %9

9:                                                ; preds = %6
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %11 = icmp eq i32 %10, 350
  br i1 %11, label %15, label %12

12:                                               ; preds = %9
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %14 = icmp eq i32 %13, 370
  br i1 %14, label %15, label %18

15:                                               ; preds = %12, %9, %6
  %16 = call double @llvm.fabs.f64(double %a)
  %17 = fcmp oeq double %16, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i

18:                                               ; preds = %12
  %19 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %20 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %21 = and i32 %20, 2147483647
  %22 = icmp eq i32 %21, 2146435072
  %23 = icmp eq i32 %19, 0
  %24 = select i1 %22, i1 %23, i1 false
  br label %__nv_isinfd.exit.i

__nv_isinfd.exit.i:                               ; preds = %18, %15
  %.02.in = phi i1 [ %17, %15 ], [ %24, %18 ]
  br i1 %.02.in, label %25, label %27

25:                                               ; preds = %__nv_isinfd.exit.i
  %26 = fmul double %a, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i

27:                                               ; preds = %__nv_isinfd.exit.i
  %28 = fmul double %a, 0x3FE45F306DC9C883
  %29 = call i32 @llvm.nvvm.d2i.rn(double %28) #139
  store i32 %29, i32* %q.i.i, align 4, !noalias !68
  %30 = sitofp i32 %29 to double
  %31 = fneg double %30
  %32 = call double @llvm.fma.f64(double %31, double noundef 0x3FF921FB54442D18, double %a)
  %33 = call double @llvm.fma.f64(double %31, double noundef 0x3C91A62633145C00, double %32)
  %34 = call double @llvm.fma.f64(double %31, double noundef 0x397B839A252049C0, double %33)
  %35 = call double @llvm.fabs.f64(double %a)
  %36 = fcmp ult double %35, 0x41E0000000000000
  br i1 %36, label %__internal_trig_reduction_kerneld.exit.i, label %37

37:                                               ; preds = %27
  %38 = call fastcc double @__internal_trig_reduction_slowpathd(double %a, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i) #139
  %.pre7 = load i32, i32* %q.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i

__internal_trig_reduction_kerneld.exit.i:         ; preds = %27, %37, %25
  %i.i.0 = phi i32 [ 0, %25 ], [ %.pre7, %37 ], [ %29, %27 ]
  %.01 = phi double [ %26, %25 ], [ %38, %37 ], [ %34, %27 ]
  %39 = fmul double %.01, %.01
  %40 = call double @llvm.fma.f64(double %39, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %41 = call double @llvm.fma.f64(double %40, double %39, double noundef 0xBE927E4F8E06E6D9)
  %42 = call double @llvm.fma.f64(double %41, double %39, double noundef 0x3EFA01A019DDBCE9)
  %43 = call double @llvm.fma.f64(double %42, double %39, double noundef 0xBF56C16C16C15D47)
  %44 = call double @llvm.fma.f64(double %43, double %39, double noundef 0x3FA5555555555551)
  %45 = call double @llvm.fma.f64(double %44, double %39, double noundef -5.000000e-01)
  %46 = call double @llvm.fma.f64(double %45, double %39, double noundef 1.000000e+00)
  %47 = call double @llvm.fma.f64(double %39, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %48 = call double @llvm.fma.f64(double %47, double %39, double noundef 0x3EC71DE369ACE392)
  %49 = call double @llvm.fma.f64(double %48, double %39, double noundef 0xBF2A01A019DB62A1)
  %50 = call double @llvm.fma.f64(double %49, double %39, double noundef 0x3F81111111110818)
  %51 = call double @llvm.fma.f64(double %50, double %39, double noundef 0xBFC5555555555554)
  %52 = call double @llvm.fma.f64(double %51, double %39, double noundef 0.000000e+00)
  %53 = call double @llvm.fma.f64(double %52, double %.01, double %.01)
  %54 = and i32 %i.i.0, 1
  %.not5 = icmp eq i32 %54, 0
  %spec.select = select i1 %.not5, double %53, double %46
  %55 = and i32 %i.i.0, 2
  %.not6 = icmp eq i32 %55, 0
  br i1 %.not6, label %__nv_sincos.exit, label %56

56:                                               ; preds = %__internal_trig_reduction_kerneld.exit.i
  %57 = call i32 @llvm.nvvm.d2i.hi(double %spec.select) #139
  %58 = call i32 @llvm.nvvm.d2i.lo(double %spec.select) #139
  %59 = xor i32 %57, -2147483648
  %60 = call double @llvm.nvvm.lohi.i2d(i32 %58, i32 %59) #139
  br label %__nv_sincos.exit

61:                                               ; preds = %3
  %62 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %63 = icmp eq i32 %62, 200
  br i1 %63, label %70, label %64

64:                                               ; preds = %61
  %65 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %66 = icmp eq i32 %65, 350
  br i1 %66, label %70, label %67

67:                                               ; preds = %64
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %69 = icmp eq i32 %68, 370
  br i1 %69, label %70, label %73

70:                                               ; preds = %67, %64, %61
  %71 = call double @llvm.fabs.f64(double %a)
  %72 = fcmp oeq double %71, 0x7FF0000000000000
  br label %__nv_isinfd.exit

73:                                               ; preds = %67
  %74 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %75 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %76 = and i32 %75, 2147483647
  %77 = icmp eq i32 %76, 2146435072
  %78 = icmp eq i32 %74, 0
  %79 = select i1 %77, i1 %78, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %73, %70
  %.03.in = phi i1 [ %72, %70 ], [ %79, %73 ]
  br i1 %.03.in, label %80, label %82

80:                                               ; preds = %__nv_isinfd.exit
  %81 = fmul double %a, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit

82:                                               ; preds = %__nv_isinfd.exit
  %83 = fmul double %a, 0x3FE45F306DC9C883
  %84 = call i32 @llvm.nvvm.d2i.rn(double %83) #139
  store i32 %84, i32* %q.i, align 4, !noalias !68
  %85 = sitofp i32 %84 to double
  %86 = fneg double %85
  %87 = call double @llvm.fma.f64(double %86, double noundef 0x3FF921FB54442D18, double %a)
  %88 = call double @llvm.fma.f64(double %86, double noundef 0x3C91A62633145C00, double %87)
  %89 = call double @llvm.fma.f64(double %86, double noundef 0x397B839A252049C0, double %88)
  %90 = call double @llvm.fabs.f64(double %a)
  %91 = fcmp ult double %90, 0x41E0000000000000
  br i1 %91, label %__internal_sin_cos_kerneld.exit, label %92

92:                                               ; preds = %82
  %93 = call fastcc double @__internal_trig_reduction_slowpathd(double %a, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i) #139
  %.pre = load i32, i32* %q.i, align 4
  br label %__internal_sin_cos_kerneld.exit

__internal_sin_cos_kerneld.exit:                  ; preds = %82, %92, %80
  %i.0 = phi i32 [ 0, %80 ], [ %.pre, %92 ], [ %84, %82 ]
  %z.0 = phi double [ %81, %80 ], [ %93, %92 ], [ %89, %82 ]
  %94 = shl i32 %i.0, 3
  %95 = and i32 %94, 8
  %96 = zext i32 %95 to i64
  %97 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %96
  %98 = fmul double %z.0, %z.0
  %99 = and i32 %i.0, 1
  %.not = icmp eq i32 %99, 0
  %100 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %101 = getelementptr inbounds double, double addrspace(1)* %97, i64 1
  %102 = load double, double addrspace(1)* %101, align 8
  %103 = call double @llvm.fma.f64(double noundef %100, double %98, double %102)
  %104 = getelementptr inbounds double, double addrspace(1)* %97, i64 2
  %105 = load double, double addrspace(1)* %104, align 8
  %106 = call double @llvm.fma.f64(double %103, double %98, double %105)
  %107 = getelementptr inbounds double, double addrspace(1)* %97, i64 3
  %108 = load double, double addrspace(1)* %107, align 8
  %109 = call double @llvm.fma.f64(double %106, double %98, double %108)
  %110 = getelementptr inbounds double, double addrspace(1)* %97, i64 4
  %111 = load double, double addrspace(1)* %110, align 8
  %112 = call double @llvm.fma.f64(double %109, double %98, double %111)
  %113 = getelementptr inbounds double, double addrspace(1)* %97, i64 5
  %114 = load double, double addrspace(1)* %113, align 8
  %115 = call double @llvm.fma.f64(double %112, double %98, double %114)
  %116 = getelementptr inbounds double, double addrspace(1)* %97, i64 6
  %117 = load double, double addrspace(1)* %116, align 8
  %118 = call double @llvm.fma.f64(double %115, double %98, double %117)
  %119 = call double @llvm.fma.f64(double %118, double %z.0, double %z.0)
  %120 = call double @llvm.fma.f64(double %118, double %98, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not, double %119, double %120
  %121 = and i32 %i.0, 2
  %.not4 = icmp eq i32 %121, 0
  %122 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not4, double %spec.select2, double %122
  br label %__nv_sincos.exit

__nv_sincos.exit:                                 ; preds = %__internal_trig_reduction_kerneld.exit.i, %56, %__internal_sin_cos_kerneld.exit
  %.0 = phi double [ %.1, %__internal_sin_cos_kerneld.exit ], [ %60, %56 ], [ %spec.select, %__internal_trig_reduction_kerneld.exit.i ]
  ret double %.0
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_cos(double %a) local_unnamed_addr #74 {
  %q.i = alloca i32, align 4
  %q.i.i = alloca i32, align 4
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2 = icmp eq i32 %1, 350
  br i1 %2, label %6, label %3

3:                                                ; preds = %0
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %5 = icmp eq i32 %4, 370
  br i1 %5, label %6, label %65

6:                                                ; preds = %3, %0
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %8 = icmp eq i32 %7, 200
  br i1 %8, label %15, label %9

9:                                                ; preds = %6
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %11 = icmp eq i32 %10, 350
  br i1 %11, label %15, label %12

12:                                               ; preds = %9
  %13 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %14 = icmp eq i32 %13, 370
  br i1 %14, label %15, label %18

15:                                               ; preds = %12, %9, %6
  %16 = call double @llvm.fabs.f64(double %a)
  %17 = fcmp oeq double %16, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i

18:                                               ; preds = %12
  %19 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %20 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %21 = and i32 %20, 2147483647
  %22 = icmp eq i32 %21, 2146435072
  %23 = icmp eq i32 %19, 0
  %24 = select i1 %22, i1 %23, i1 false
  br label %__nv_isinfd.exit.i

__nv_isinfd.exit.i:                               ; preds = %18, %15
  %.02.in = phi i1 [ %17, %15 ], [ %24, %18 ]
  br i1 %.02.in, label %25, label %27

25:                                               ; preds = %__nv_isinfd.exit.i
  %26 = fmul double %a, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i

27:                                               ; preds = %__nv_isinfd.exit.i
  %28 = fmul double %a, 0x3FE45F306DC9C883
  %29 = call i32 @llvm.nvvm.d2i.rn(double %28) #139
  store i32 %29, i32* %q.i.i, align 4, !noalias !68
  %30 = sitofp i32 %29 to double
  %31 = fneg double %30
  %32 = call double @llvm.fma.f64(double %31, double noundef 0x3FF921FB54442D18, double %a)
  %33 = call double @llvm.fma.f64(double %31, double noundef 0x3C91A62633145C00, double %32)
  %34 = call double @llvm.fma.f64(double %31, double noundef 0x397B839A252049C0, double %33)
  %35 = call double @llvm.fabs.f64(double %a)
  %36 = fcmp ult double %35, 0x41E0000000000000
  br i1 %36, label %__internal_trig_reduction_kerneld.exit.i, label %37

37:                                               ; preds = %27
  %38 = call fastcc double @__internal_trig_reduction_slowpathd(double %a, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i) #139
  %.pre7 = load i32, i32* %q.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i

__internal_trig_reduction_kerneld.exit.i:         ; preds = %27, %37, %25
  %i.i.0 = phi i32 [ 0, %25 ], [ %.pre7, %37 ], [ %29, %27 ]
  %.01 = phi double [ %26, %25 ], [ %38, %37 ], [ %34, %27 ]
  %39 = fmul double %.01, %.01
  %40 = call double @llvm.fma.f64(double %39, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %41 = call double @llvm.fma.f64(double %40, double %39, double noundef 0xBE927E4F8E06E6D9)
  %42 = call double @llvm.fma.f64(double %41, double %39, double noundef 0x3EFA01A019DDBCE9)
  %43 = call double @llvm.fma.f64(double %42, double %39, double noundef 0xBF56C16C16C15D47)
  %44 = call double @llvm.fma.f64(double %43, double %39, double noundef 0x3FA5555555555551)
  %45 = call double @llvm.fma.f64(double %44, double %39, double noundef -5.000000e-01)
  %46 = call double @llvm.fma.f64(double %45, double %39, double noundef 1.000000e+00)
  %47 = call double @llvm.fma.f64(double %39, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %48 = call double @llvm.fma.f64(double %47, double %39, double noundef 0x3EC71DE369ACE392)
  %49 = call double @llvm.fma.f64(double %48, double %39, double noundef 0xBF2A01A019DB62A1)
  %50 = call double @llvm.fma.f64(double %49, double %39, double noundef 0x3F81111111110818)
  %51 = call double @llvm.fma.f64(double %50, double %39, double noundef 0xBFC5555555555554)
  %52 = call double @llvm.fma.f64(double %51, double %39, double noundef 0.000000e+00)
  %53 = call double @llvm.fma.f64(double %52, double %.01, double %.01)
  %54 = call i32 @llvm.nvvm.d2i.hi(double %53) #139
  %55 = call i32 @llvm.nvvm.d2i.lo(double %53) #139
  %56 = xor i32 %54, -2147483648
  %57 = call double @llvm.nvvm.lohi.i2d(i32 %55, i32 %56) #139
  %58 = and i32 %i.i.0, 1
  %.not5 = icmp eq i32 %58, 0
  %spec.select1 = select i1 %.not5, double %46, double %57
  %59 = and i32 %i.i.0, 2
  %.not6 = icmp eq i32 %59, 0
  br i1 %.not6, label %__nv_sincos.exit, label %60

60:                                               ; preds = %__internal_trig_reduction_kerneld.exit.i
  %61 = call i32 @llvm.nvvm.d2i.hi(double %spec.select1) #139
  %62 = call i32 @llvm.nvvm.d2i.lo(double %spec.select1) #139
  %63 = xor i32 %61, -2147483648
  %64 = call double @llvm.nvvm.lohi.i2d(i32 %62, i32 %63) #139
  br label %__nv_sincos.exit

65:                                               ; preds = %3
  %66 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %67 = icmp eq i32 %66, 200
  br i1 %67, label %74, label %68

68:                                               ; preds = %65
  %69 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %70 = icmp eq i32 %69, 350
  br i1 %70, label %74, label %71

71:                                               ; preds = %68
  %72 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %73 = icmp eq i32 %72, 370
  br i1 %73, label %74, label %77

74:                                               ; preds = %71, %68, %65
  %75 = call double @llvm.fabs.f64(double %a)
  %76 = fcmp oeq double %75, 0x7FF0000000000000
  br label %__nv_isinfd.exit

77:                                               ; preds = %71
  %78 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %79 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %80 = and i32 %79, 2147483647
  %81 = icmp eq i32 %80, 2146435072
  %82 = icmp eq i32 %78, 0
  %83 = select i1 %81, i1 %82, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %77, %74
  %.03.in = phi i1 [ %76, %74 ], [ %83, %77 ]
  br i1 %.03.in, label %84, label %86

84:                                               ; preds = %__nv_isinfd.exit
  %85 = fmul double %a, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit

86:                                               ; preds = %__nv_isinfd.exit
  %87 = fmul double %a, 0x3FE45F306DC9C883
  %88 = call i32 @llvm.nvvm.d2i.rn(double %87) #139
  store i32 %88, i32* %q.i, align 4, !noalias !68
  %89 = sitofp i32 %88 to double
  %90 = fneg double %89
  %91 = call double @llvm.fma.f64(double %90, double noundef 0x3FF921FB54442D18, double %a)
  %92 = call double @llvm.fma.f64(double %90, double noundef 0x3C91A62633145C00, double %91)
  %93 = call double @llvm.fma.f64(double %90, double noundef 0x397B839A252049C0, double %92)
  %94 = call double @llvm.fabs.f64(double %a)
  %95 = fcmp ult double %94, 0x41E0000000000000
  br i1 %95, label %__internal_trig_reduction_kerneld.exit, label %96

96:                                               ; preds = %86
  %97 = call fastcc double @__internal_trig_reduction_slowpathd(double %a, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i) #139
  %.pre = load i32, i32* %q.i, align 4
  br label %__internal_trig_reduction_kerneld.exit

__internal_trig_reduction_kerneld.exit:           ; preds = %96, %86
  %98 = phi i32 [ %.pre, %96 ], [ %88, %86 ]
  %t.i1.0 = phi double [ %97, %96 ], [ %93, %86 ]
  %phi.bo = add i32 %98, 1
  br label %__internal_sin_cos_kerneld.exit

__internal_sin_cos_kerneld.exit:                  ; preds = %__internal_trig_reduction_kerneld.exit, %84
  %i.0 = phi i32 [ 1, %84 ], [ %phi.bo, %__internal_trig_reduction_kerneld.exit ]
  %z.0 = phi double [ %85, %84 ], [ %t.i1.0, %__internal_trig_reduction_kerneld.exit ]
  %99 = shl i32 %i.0, 3
  %100 = and i32 %99, 8
  %101 = zext i32 %100 to i64
  %102 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %101
  %103 = fmul double %z.0, %z.0
  %104 = and i32 %i.0, 1
  %.not = icmp eq i32 %104, 0
  %105 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %106 = getelementptr inbounds double, double addrspace(1)* %102, i64 1
  %107 = load double, double addrspace(1)* %106, align 8
  %108 = call double @llvm.fma.f64(double noundef %105, double %103, double %107)
  %109 = getelementptr inbounds double, double addrspace(1)* %102, i64 2
  %110 = load double, double addrspace(1)* %109, align 8
  %111 = call double @llvm.fma.f64(double %108, double %103, double %110)
  %112 = getelementptr inbounds double, double addrspace(1)* %102, i64 3
  %113 = load double, double addrspace(1)* %112, align 8
  %114 = call double @llvm.fma.f64(double %111, double %103, double %113)
  %115 = getelementptr inbounds double, double addrspace(1)* %102, i64 4
  %116 = load double, double addrspace(1)* %115, align 8
  %117 = call double @llvm.fma.f64(double %114, double %103, double %116)
  %118 = getelementptr inbounds double, double addrspace(1)* %102, i64 5
  %119 = load double, double addrspace(1)* %118, align 8
  %120 = call double @llvm.fma.f64(double %117, double %103, double %119)
  %121 = getelementptr inbounds double, double addrspace(1)* %102, i64 6
  %122 = load double, double addrspace(1)* %121, align 8
  %123 = call double @llvm.fma.f64(double %120, double %103, double %122)
  %124 = call double @llvm.fma.f64(double %123, double %z.0, double %z.0)
  %125 = call double @llvm.fma.f64(double %123, double %103, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not, double %124, double %125
  %126 = and i32 %i.0, 2
  %.not4 = icmp eq i32 %126, 0
  %127 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not4, double %spec.select2, double %127
  br label %__nv_sincos.exit

__nv_sincos.exit:                                 ; preds = %__internal_trig_reduction_kerneld.exit.i, %60, %__internal_sin_cos_kerneld.exit
  %.0 = phi double [ %.1, %__internal_sin_cos_kerneld.exit ], [ %64, %60 ], [ %spec.select1, %__internal_trig_reduction_kerneld.exit.i ]
  ret double %.0
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_tan(double %a) local_unnamed_addr #75 {
  %q.i = alloca i32, align 4
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2 = icmp eq i32 %1, 200
  br i1 %2, label %9, label %3

3:                                                ; preds = %0
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %5 = icmp eq i32 %4, 350
  br i1 %5, label %9, label %6

6:                                                ; preds = %3
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %8 = icmp eq i32 %7, 370
  br i1 %8, label %9, label %12

9:                                                ; preds = %6, %3, %0
  %10 = call double @llvm.fabs.f64(double %a)
  %11 = fcmp oeq double %10, 0x7FF0000000000000
  br label %__nv_isinfd.exit

12:                                               ; preds = %6
  %13 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %14 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %15 = and i32 %14, 2147483647
  %16 = icmp eq i32 %15, 2146435072
  %17 = icmp eq i32 %13, 0
  %18 = select i1 %16, i1 %17, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %12, %9
  %.0.in = phi i1 [ %11, %9 ], [ %18, %12 ]
  br i1 %.0.in, label %19, label %21

19:                                               ; preds = %__nv_isinfd.exit
  %20 = fmul double %a, 0.000000e+00
  br label %34

21:                                               ; preds = %__nv_isinfd.exit
  %22 = fmul double %a, 0x3FE45F306DC9C883
  %23 = call i32 @llvm.nvvm.d2i.rn(double %22) #139
  store i32 %23, i32* %q.i, align 4, !noalias !68
  %24 = sitofp i32 %23 to double
  %25 = fneg double %24
  %26 = call double @llvm.fma.f64(double %25, double noundef 0x3FF921FB54442D18, double %a)
  %27 = call double @llvm.fma.f64(double %25, double noundef 0x3C91A62633145C00, double %26)
  %28 = call double @llvm.fma.f64(double %25, double noundef 0x397B839A252049C0, double %27)
  %29 = call double @llvm.fabs.f64(double %a)
  %30 = fcmp ult double %29, 0x41E0000000000000
  br i1 %30, label %__internal_trig_reduction_kerneld.exit, label %31

31:                                               ; preds = %21
  %32 = call fastcc double @__internal_trig_reduction_slowpathd(double %a, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i) #139
  %.pre = load i32, i32* %q.i, align 4
  br label %__internal_trig_reduction_kerneld.exit

__internal_trig_reduction_kerneld.exit:           ; preds = %31, %21
  %33 = phi i32 [ %.pre, %31 ], [ %23, %21 ]
  %t.i.0 = phi double [ %32, %31 ], [ %28, %21 ]
  %phi.bo = and i32 %33, 1
  %phi.cmp = icmp eq i32 %phi.bo, 0
  br label %34

34:                                               ; preds = %__internal_trig_reduction_kerneld.exit, %19
  %i.0 = phi i1 [ true, %19 ], [ %phi.cmp, %__internal_trig_reduction_kerneld.exit ]
  %z.0 = phi double [ %20, %19 ], [ %t.i.0, %__internal_trig_reduction_kerneld.exit ]
  %35 = fmul double %z.0, %z.0
  %36 = call double @llvm.fma.f64(double %35, double noundef 0x3EE48DAC2799BCB9, double noundef 0xBEF9757C5B27EBB1)
  %37 = call double @llvm.fma.f64(double %36, double %35, double noundef 0x3F0980E90FD91E04)
  %38 = call double @llvm.fma.f64(double %37, double %35, double noundef 0xBEFAE2B0417D7E1D)
  %39 = call double @llvm.fma.f64(double %38, double %35, double noundef 0x3F119F5341BFBA57)
  %40 = call double @llvm.fma.f64(double %39, double %35, double noundef 0x3F15E791A00F6919)
  %41 = call double @llvm.fma.f64(double %40, double %35, double noundef 0x3F2FF2E7FADEC73A)
  %42 = call double @llvm.fma.f64(double %41, double %35, double noundef 0x3F434BC1B206DA62)
  %43 = call double @llvm.fma.f64(double %42, double %35, double noundef 0x3F57DB18EF2F83F9)
  %44 = call double @llvm.fma.f64(double %43, double %35, double noundef 0x3F6D6D2E7AE49FBC)
  %45 = call double @llvm.fma.f64(double %44, double %35, double noundef 0x3F8226E3A816A776)
  %46 = call double @llvm.fma.f64(double %45, double %35, double noundef 0x3F9664F485D25660)
  %47 = call double @llvm.fma.f64(double %46, double %35, double noundef 0x3FABA1BA1BABF31D)
  %48 = call double @llvm.fma.f64(double %47, double %35, double noundef 0x3FC11111111105D2)
  %49 = call double @llvm.fma.f64(double %48, double %35, double noundef 0x3FD555555555555E)
  %50 = fmul double %35, %49
  %51 = call double @llvm.fma.f64(double %50, double %z.0, double %z.0)
  br i1 %i.0, label %__internal_tan_kerneld.exit, label %52

52:                                               ; preds = %34
  %53 = fsub double %51, %z.0
  %54 = fneg double %53
  %55 = call double @llvm.fma.f64(double %50, double %z.0, double %54)
  %56 = call double @llvm.nvvm.rcp.approx.ftz.d(double %51) #139
  %57 = fneg double %51
  %58 = call double @llvm.fma.f64(double %57, double %56, double noundef 1.000000e+00)
  %59 = call double @llvm.fma.f64(double %58, double %58, double %58)
  %60 = call double @llvm.fma.f64(double %59, double %56, double %56)
  %61 = fneg double %60
  %62 = call double @llvm.fma.f64(double %51, double %61, double noundef 1.000000e+00)
  %63 = call double @llvm.fma.f64(double %61, double %55, double %62)
  %64 = call double @llvm.fma.f64(double %63, double %61, double %61)
  br label %__internal_tan_kerneld.exit

__internal_tan_kerneld.exit:                      ; preds = %52, %34
  %q.i1.0 = phi double [ %64, %52 ], [ %51, %34 ]
  ret double %q.i1.0
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_log(double %a) local_unnamed_addr #76 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %3 = icmp slt i32 %1, 1048576
  br i1 %3, label %4, label %8

4:                                                ; preds = %0
  %5 = fmul double %a, 0x4350000000000000
  %6 = call i32 @llvm.nvvm.d2i.hi(double %5) #142
  %7 = call i32 @llvm.nvvm.d2i.lo(double %5) #142
  br label %8

8:                                                ; preds = %4, %0
  %.0 = phi double [ %5, %4 ], [ %a, %0 ]
  %ihi.0 = phi i32 [ %6, %4 ], [ %1, %0 ]
  %ilo.0 = phi i32 [ %7, %4 ], [ %2, %0 ]
  %e.0 = phi i32 [ -1077, %4 ], [ -1023, %0 ]
  %9 = add i32 %ihi.0, -1
  %10 = icmp ult i32 %9, 2146435071
  br i1 %10, label %11, label %60

11:                                               ; preds = %8
  %12 = lshr i32 %ihi.0, 20
  %13 = add nsw i32 %e.0, %12
  %14 = and i32 %ihi.0, -2146435073
  %15 = or i32 %14, 1072693248
  %16 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.0, i32 %15) #142
  %17 = icmp sgt i32 %15, 1073127582
  br i1 %17, label %18, label %24

18:                                               ; preds = %11
  %19 = call i32 @llvm.nvvm.d2i.lo(double %16) #142
  %20 = call i32 @llvm.nvvm.d2i.hi(double %16) #142
  %21 = add i32 %20, -1048576
  %22 = call double @llvm.nvvm.lohi.i2d(i32 %19, i32 %21) #142
  %23 = add nsw i32 %13, 1
  br label %24

24:                                               ; preds = %18, %11
  %m.0 = phi double [ %22, %18 ], [ %16, %11 ]
  %e.1 = phi i32 [ %23, %18 ], [ %13, %11 ]
  %25 = fadd double %m.0, -1.000000e+00
  %26 = fadd double %m.0, 1.000000e+00
  %27 = call double @llvm.nvvm.rcp.approx.ftz.d(double %26) #142
  %28 = fneg double %26
  %29 = call double @llvm.fma.f64(double %28, double %27, double noundef 1.000000e+00) #140
  %30 = call double @llvm.fma.f64(double %29, double %29, double %29) #140
  %31 = call double @llvm.fma.f64(double %30, double %27, double %27) #140
  %32 = fmul double %25, %31
  %33 = fadd double %32, %32
  %34 = fmul double %33, %33
  %35 = call double @llvm.fma.f64(double %34, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %36 = call double @llvm.fma.f64(double %35, double %34, double noundef 0x3EF3B2669F02676F) #140
  %37 = call double @llvm.fma.f64(double %36, double %34, double noundef 0x3F1745CBA9AB0956) #140
  %38 = call double @llvm.fma.f64(double %37, double %34, double noundef 0x3F3C71C72D1B5154) #140
  %39 = call double @llvm.fma.f64(double %38, double %34, double noundef 0x3F624924923BE72D) #140
  %40 = call double @llvm.fma.f64(double %39, double %34, double noundef 0x3F8999999999A3C4) #140
  %41 = call double @llvm.fma.f64(double %40, double %34, double noundef 0x3FB5555555555554) #140
  %42 = fsub double %25, %33
  %43 = fmul double %42, 2.000000e+00
  %44 = fneg double %33
  %45 = call double @llvm.fma.f64(double %44, double %25, double %43) #140
  %46 = fmul double %31, %45
  %47 = fmul double %34, %41
  %48 = call double @llvm.fma.f64(double %47, double %33, double %46) #140
  %49 = xor i32 %e.1, -2147483648
  %50 = call double @llvm.nvvm.lohi.i2d(i32 %49, i32 noundef 1127219200) #142
  %51 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %52 = fsub double %50, %51
  %53 = call double @llvm.fma.f64(double %52, double noundef 0x3FE62E42FEFA39EF, double %33) #140
  %54 = fneg double %52
  %55 = call double @llvm.fma.f64(double %54, double noundef 0x3FE62E42FEFA39EF, double %53) #140
  %56 = fsub double %55, %33
  %57 = fsub double %48, %56
  %58 = call double @llvm.fma.f64(double %52, double noundef 0x3C7ABC9E3B39803F, double %57) #140
  %59 = fadd double %53, %58
  br label %65

60:                                               ; preds = %8
  %61 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %62 = call i32 @llvm.nvvm.d2i.hi(double %.0) #142
  %63 = bitcast i32 %62 to float
  %64 = fcmp oeq float %63, 0.000000e+00
  %q.0 = select i1 %64, double 0xFFF0000000000000, double %61
  br label %65

65:                                               ; preds = %60, %24
  %q.1 = phi double [ %59, %24 ], [ %q.0, %60 ]
  ret double %q.1
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_log2(double %a) local_unnamed_addr #77 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %3 = icmp slt i32 %1, 1048576
  br i1 %3, label %4, label %8

4:                                                ; preds = %0
  %5 = fmul double %a, 0x4350000000000000
  %6 = call i32 @llvm.nvvm.d2i.hi(double %5) #142
  %7 = call i32 @llvm.nvvm.d2i.lo(double %5) #142
  br label %8

8:                                                ; preds = %4, %0
  %.0 = phi double [ %5, %4 ], [ %a, %0 ]
  %ihi.i.0 = phi i32 [ %6, %4 ], [ %1, %0 ]
  %ilo.i.0 = phi i32 [ %7, %4 ], [ %2, %0 ]
  %e.i.0 = phi i32 [ -1077, %4 ], [ -1023, %0 ]
  %9 = add i32 %ihi.i.0, -1
  %10 = icmp ult i32 %9, 2146435071
  br i1 %10, label %11, label %60

11:                                               ; preds = %8
  %12 = lshr i32 %ihi.i.0, 20
  %13 = add nsw i32 %e.i.0, %12
  %14 = and i32 %ihi.i.0, -2146435073
  %15 = or i32 %14, 1072693248
  %16 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %15) #142
  %17 = icmp sgt i32 %15, 1073127582
  br i1 %17, label %18, label %24

18:                                               ; preds = %11
  %19 = call i32 @llvm.nvvm.d2i.lo(double %16) #142
  %20 = call i32 @llvm.nvvm.d2i.hi(double %16) #142
  %21 = add i32 %20, -1048576
  %22 = call double @llvm.nvvm.lohi.i2d(i32 %19, i32 %21) #142
  %23 = add nsw i32 %13, 1
  br label %24

24:                                               ; preds = %18, %11
  %m.i.0 = phi double [ %22, %18 ], [ %16, %11 ]
  %e.i.1 = phi i32 [ %23, %18 ], [ %13, %11 ]
  %25 = fadd double %m.i.0, -1.000000e+00
  %26 = fadd double %m.i.0, 1.000000e+00
  %27 = call double @llvm.nvvm.rcp.approx.ftz.d(double %26) #142
  %28 = fneg double %26
  %29 = call double @llvm.fma.f64(double %28, double %27, double noundef 1.000000e+00) #140
  %30 = call double @llvm.fma.f64(double %29, double %29, double %29) #140
  %31 = call double @llvm.fma.f64(double %30, double %27, double %27) #140
  %32 = fmul double %25, %31
  %33 = fadd double %32, %32
  %34 = fmul double %33, %33
  %35 = call double @llvm.fma.f64(double %34, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %36 = call double @llvm.fma.f64(double %35, double %34, double noundef 0x3EF3B2669F02676F) #140
  %37 = call double @llvm.fma.f64(double %36, double %34, double noundef 0x3F1745CBA9AB0956) #140
  %38 = call double @llvm.fma.f64(double %37, double %34, double noundef 0x3F3C71C72D1B5154) #140
  %39 = call double @llvm.fma.f64(double %38, double %34, double noundef 0x3F624924923BE72D) #140
  %40 = call double @llvm.fma.f64(double %39, double %34, double noundef 0x3F8999999999A3C4) #140
  %41 = call double @llvm.fma.f64(double %40, double %34, double noundef 0x3FB5555555555554) #140
  %42 = fsub double %25, %33
  %43 = fmul double %42, 2.000000e+00
  %44 = fneg double %33
  %45 = call double @llvm.fma.f64(double %44, double %25, double %43) #140
  %46 = fmul double %31, %45
  %47 = fmul double %34, %41
  %48 = call double @llvm.fma.f64(double %47, double %33, double %46) #140
  %49 = xor i32 %e.i.1, -2147483648
  %50 = call double @llvm.nvvm.lohi.i2d(i32 %49, i32 noundef 1127219200) #142
  %51 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %52 = fsub double %50, %51
  %53 = call double @llvm.fma.f64(double %52, double noundef 0x3FE62E42FEFA39EF, double %33) #140
  %54 = fneg double %52
  %55 = call double @llvm.fma.f64(double %54, double noundef 0x3FE62E42FEFA39EF, double %53) #140
  %56 = fsub double %55, %33
  %57 = fsub double %48, %56
  %58 = call double @llvm.fma.f64(double %52, double noundef 0x3C7ABC9E3B39803F, double %57) #140
  %59 = fadd double %53, %58
  br label %__nv_log.exit

60:                                               ; preds = %8
  %61 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %62 = call i32 @llvm.nvvm.d2i.hi(double %.0) #142
  %63 = bitcast i32 %62 to float
  %64 = fcmp oeq float %63, 0.000000e+00
  %q.i.0 = select i1 %64, double 0xFFF0000000000000, double %61
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %60, %24
  %q.i.1 = phi double [ %59, %24 ], [ %q.i.0, %60 ]
  %65 = fmul double %q.i.1, 0x3C7777D0FFDA0D24
  %66 = call double @llvm.fma.f64(double %q.i.1, double noundef 0x3FF71547652B82FE, double %65) #140
  ret double %66
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_log10(double %a) local_unnamed_addr #78 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %3 = icmp slt i32 %1, 1048576
  br i1 %3, label %4, label %8

4:                                                ; preds = %0
  %5 = fmul double %a, 0x4350000000000000
  %6 = call i32 @llvm.nvvm.d2i.hi(double %5) #142
  %7 = call i32 @llvm.nvvm.d2i.lo(double %5) #142
  br label %8

8:                                                ; preds = %4, %0
  %.0 = phi double [ %5, %4 ], [ %a, %0 ]
  %ihi.i.0 = phi i32 [ %6, %4 ], [ %1, %0 ]
  %ilo.i.0 = phi i32 [ %7, %4 ], [ %2, %0 ]
  %e.i.0 = phi i32 [ -1077, %4 ], [ -1023, %0 ]
  %9 = add i32 %ihi.i.0, -1
  %10 = icmp ult i32 %9, 2146435071
  br i1 %10, label %11, label %60

11:                                               ; preds = %8
  %12 = lshr i32 %ihi.i.0, 20
  %13 = add nsw i32 %e.i.0, %12
  %14 = and i32 %ihi.i.0, -2146435073
  %15 = or i32 %14, 1072693248
  %16 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %15) #142
  %17 = icmp sgt i32 %15, 1073127582
  br i1 %17, label %18, label %24

18:                                               ; preds = %11
  %19 = call i32 @llvm.nvvm.d2i.lo(double %16) #142
  %20 = call i32 @llvm.nvvm.d2i.hi(double %16) #142
  %21 = add i32 %20, -1048576
  %22 = call double @llvm.nvvm.lohi.i2d(i32 %19, i32 %21) #142
  %23 = add nsw i32 %13, 1
  br label %24

24:                                               ; preds = %18, %11
  %m.i.0 = phi double [ %22, %18 ], [ %16, %11 ]
  %e.i.1 = phi i32 [ %23, %18 ], [ %13, %11 ]
  %25 = fadd double %m.i.0, -1.000000e+00
  %26 = fadd double %m.i.0, 1.000000e+00
  %27 = call double @llvm.nvvm.rcp.approx.ftz.d(double %26) #142
  %28 = fneg double %26
  %29 = call double @llvm.fma.f64(double %28, double %27, double noundef 1.000000e+00) #140
  %30 = call double @llvm.fma.f64(double %29, double %29, double %29) #140
  %31 = call double @llvm.fma.f64(double %30, double %27, double %27) #140
  %32 = fmul double %25, %31
  %33 = fadd double %32, %32
  %34 = fmul double %33, %33
  %35 = call double @llvm.fma.f64(double %34, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %36 = call double @llvm.fma.f64(double %35, double %34, double noundef 0x3EF3B2669F02676F) #140
  %37 = call double @llvm.fma.f64(double %36, double %34, double noundef 0x3F1745CBA9AB0956) #140
  %38 = call double @llvm.fma.f64(double %37, double %34, double noundef 0x3F3C71C72D1B5154) #140
  %39 = call double @llvm.fma.f64(double %38, double %34, double noundef 0x3F624924923BE72D) #140
  %40 = call double @llvm.fma.f64(double %39, double %34, double noundef 0x3F8999999999A3C4) #140
  %41 = call double @llvm.fma.f64(double %40, double %34, double noundef 0x3FB5555555555554) #140
  %42 = fsub double %25, %33
  %43 = fmul double %42, 2.000000e+00
  %44 = fneg double %33
  %45 = call double @llvm.fma.f64(double %44, double %25, double %43) #140
  %46 = fmul double %31, %45
  %47 = fmul double %34, %41
  %48 = call double @llvm.fma.f64(double %47, double %33, double %46) #140
  %49 = xor i32 %e.i.1, -2147483648
  %50 = call double @llvm.nvvm.lohi.i2d(i32 %49, i32 noundef 1127219200) #142
  %51 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %52 = fsub double %50, %51
  %53 = call double @llvm.fma.f64(double %52, double noundef 0x3FE62E42FEFA39EF, double %33) #140
  %54 = fneg double %52
  %55 = call double @llvm.fma.f64(double %54, double noundef 0x3FE62E42FEFA39EF, double %53) #140
  %56 = fsub double %55, %33
  %57 = fsub double %48, %56
  %58 = call double @llvm.fma.f64(double %52, double noundef 0x3C7ABC9E3B39803F, double %57) #140
  %59 = fadd double %53, %58
  br label %__nv_log.exit

60:                                               ; preds = %8
  %61 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %62 = call i32 @llvm.nvvm.d2i.hi(double %.0) #142
  %63 = bitcast i32 %62 to float
  %64 = fcmp oeq float %63, 0.000000e+00
  %q.i.0 = select i1 %64, double 0xFFF0000000000000, double %61
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %60, %24
  %q.i.1 = phi double [ %59, %24 ], [ %q.i.0, %60 ]
  %65 = fmul double %q.i.1, 0x3C695355BAAAFAD3
  %66 = call double @llvm.fma.f64(double %q.i.1, double noundef 0x3FDBCB7B1526E50E, double %65) #140
  ret double %66
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_log1p(double %a) local_unnamed_addr #79 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = bitcast i32 %1 to float
  %3 = fcmp olt float %2, 0x3FFCAAAAA0000000
  %4 = fcmp ogt float %2, 0xBFFB333320000000
  %or.cond = and i1 %3, %4
  br i1 %or.cond, label %5, label %22

5:                                                ; preds = %0
  %6 = fadd double %a, 2.000000e+00
  %7 = fdiv double %a, %6
  %8 = fneg double %a
  %9 = fmul double %7, %8
  %10 = fadd double %9, %a
  %11 = fmul double %10, %10
  %12 = call double @llvm.fma.f64(double %11, double noundef 0x3EB372FB2FBE14B5, double noundef 0x3ED087FFCEB2DC44) #140
  %13 = call double @llvm.fma.f64(double %12, double %11, double noundef 0x3EF3B9FF890F468C) #140
  %14 = call double @llvm.fma.f64(double %13, double %11, double noundef 0x3F17457EFD51BAF8) #140
  %15 = call double @llvm.fma.f64(double %14, double %11, double noundef 0x3F3C71C8DE3CE825) #140
  %16 = call double @llvm.fma.f64(double %15, double %11, double noundef 0x3F6249248FA4661F) #140
  %17 = call double @llvm.fma.f64(double %16, double %11, double noundef 0x3F899999999D70C4) #140
  %18 = call double @llvm.fma.f64(double %17, double %11, double noundef 0x3FB5555555555462) #140
  %19 = fmul double %11, %18
  %20 = call double @llvm.fma.f64(double %19, double %10, double %9) #140
  %21 = fadd double %20, %a
  br label %__nv_log.exit

22:                                               ; preds = %0
  %23 = fadd double %a, 1.000000e+00
  %24 = call i32 @llvm.nvvm.d2i.hi(double %23) #142
  %25 = call i32 @llvm.nvvm.d2i.lo(double %23) #142
  %26 = icmp slt i32 %24, 1048576
  br i1 %26, label %27, label %31

27:                                               ; preds = %22
  %28 = fmul double %23, 0x4350000000000000
  %29 = call i32 @llvm.nvvm.d2i.hi(double %28) #142
  %30 = call i32 @llvm.nvvm.d2i.lo(double %28) #142
  br label %31

31:                                               ; preds = %27, %22
  %.0 = phi double [ %28, %27 ], [ %23, %22 ]
  %ihi.i.0 = phi i32 [ %29, %27 ], [ %24, %22 ]
  %ilo.i.0 = phi i32 [ %30, %27 ], [ %25, %22 ]
  %e.i.0 = phi i32 [ -1077, %27 ], [ -1023, %22 ]
  %32 = add i32 %ihi.i.0, -1
  %33 = icmp ult i32 %32, 2146435071
  br i1 %33, label %34, label %83

34:                                               ; preds = %31
  %35 = lshr i32 %ihi.i.0, 20
  %36 = add nsw i32 %e.i.0, %35
  %37 = and i32 %ihi.i.0, -2146435073
  %38 = or i32 %37, 1072693248
  %39 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %38) #142
  %40 = icmp sgt i32 %38, 1073127582
  br i1 %40, label %41, label %47

41:                                               ; preds = %34
  %42 = call i32 @llvm.nvvm.d2i.lo(double %39) #142
  %43 = call i32 @llvm.nvvm.d2i.hi(double %39) #142
  %44 = add i32 %43, -1048576
  %45 = call double @llvm.nvvm.lohi.i2d(i32 %42, i32 %44) #142
  %46 = add nsw i32 %36, 1
  br label %47

47:                                               ; preds = %41, %34
  %m.i.0 = phi double [ %45, %41 ], [ %39, %34 ]
  %e.i.1 = phi i32 [ %46, %41 ], [ %36, %34 ]
  %48 = fadd double %m.i.0, -1.000000e+00
  %49 = fadd double %m.i.0, 1.000000e+00
  %50 = call double @llvm.nvvm.rcp.approx.ftz.d(double %49) #142
  %51 = fneg double %49
  %52 = call double @llvm.fma.f64(double %51, double %50, double noundef 1.000000e+00) #140
  %53 = call double @llvm.fma.f64(double %52, double %52, double %52) #140
  %54 = call double @llvm.fma.f64(double %53, double %50, double %50) #140
  %55 = fmul double %48, %54
  %56 = fadd double %55, %55
  %57 = fmul double %56, %56
  %58 = call double @llvm.fma.f64(double %57, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %59 = call double @llvm.fma.f64(double %58, double %57, double noundef 0x3EF3B2669F02676F) #140
  %60 = call double @llvm.fma.f64(double %59, double %57, double noundef 0x3F1745CBA9AB0956) #140
  %61 = call double @llvm.fma.f64(double %60, double %57, double noundef 0x3F3C71C72D1B5154) #140
  %62 = call double @llvm.fma.f64(double %61, double %57, double noundef 0x3F624924923BE72D) #140
  %63 = call double @llvm.fma.f64(double %62, double %57, double noundef 0x3F8999999999A3C4) #140
  %64 = call double @llvm.fma.f64(double %63, double %57, double noundef 0x3FB5555555555554) #140
  %65 = fsub double %48, %56
  %66 = fmul double %65, 2.000000e+00
  %67 = fneg double %56
  %68 = call double @llvm.fma.f64(double %67, double %48, double %66) #140
  %69 = fmul double %54, %68
  %70 = fmul double %57, %64
  %71 = call double @llvm.fma.f64(double %70, double %56, double %69) #140
  %72 = xor i32 %e.i.1, -2147483648
  %73 = call double @llvm.nvvm.lohi.i2d(i32 %72, i32 noundef 1127219200) #142
  %74 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %75 = fsub double %73, %74
  %76 = call double @llvm.fma.f64(double %75, double noundef 0x3FE62E42FEFA39EF, double %56) #140
  %77 = fneg double %75
  %78 = call double @llvm.fma.f64(double %77, double noundef 0x3FE62E42FEFA39EF, double %76) #140
  %79 = fsub double %78, %56
  %80 = fsub double %71, %79
  %81 = call double @llvm.fma.f64(double %75, double noundef 0x3C7ABC9E3B39803F, double %80) #140
  %82 = fadd double %76, %81
  br label %__nv_log.exit

83:                                               ; preds = %31
  %84 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %85 = call i32 @llvm.nvvm.d2i.hi(double %.0) #142
  %86 = bitcast i32 %85 to float
  %87 = fcmp oeq float %86, 0.000000e+00
  %q.i.0 = select i1 %87, double 0xFFF0000000000000, double %84
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %47, %83, %5
  %t.0 = phi double [ %21, %5 ], [ %82, %47 ], [ %q.i.0, %83 ]
  ret double %t.0
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_exp(double %a) local_unnamed_addr #80 {
__internal_fast_icmp_abs_lt.exit3:
  %0 = call double @llvm.fma.f64(double %a, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %1 = call i32 @llvm.nvvm.d2i.lo(double %0) #142
  %2 = fadd double %0, 0xC338000000000000
  %3 = call double @llvm.fma.f64(double %2, double noundef 0xBFE62E42FEFA39EF, double %a) #140
  %4 = call double @llvm.fma.f64(double %2, double noundef 0xBC7ABC9E3B39803F, double %3) #140
  %5 = call double @llvm.fma.f64(double %4, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %6 = call double @llvm.fma.f64(double %5, double %4, double noundef 0x3EC71DEE62401315) #140
  %7 = call double @llvm.fma.f64(double %6, double %4, double noundef 0x3EFA01997C89EB71) #140
  %8 = call double @llvm.fma.f64(double %7, double %4, double noundef 0x3F2A01A014761F65) #140
  %9 = call double @llvm.fma.f64(double %8, double %4, double noundef 0x3F56C16C1852B7AF) #140
  %10 = call double @llvm.fma.f64(double %9, double %4, double noundef 0x3F81111111122322) #140
  %11 = call double @llvm.fma.f64(double %10, double %4, double noundef 0x3FA55555555502A1) #140
  %12 = call double @llvm.fma.f64(double %11, double %4, double noundef 0x3FC5555555555511) #140
  %13 = call double @llvm.fma.f64(double %12, double %4, double noundef 0x3FE000000000000B) #140
  %14 = call double @llvm.fma.f64(double %13, double %4, double noundef 1.000000e+00) #140
  %15 = call double @llvm.fma.f64(double %14, double %4, double noundef 1.000000e+00) #140
  %16 = call i32 @llvm.nvvm.d2i.lo(double %15) #142
  %17 = call i32 @llvm.nvvm.d2i.hi(double %15) #142
  %18 = shl i32 %1, 20
  %19 = add i32 %17, %18
  %20 = call double @llvm.nvvm.lohi.i2d(i32 %16, i32 %19) #142
  %21 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %22 = bitcast i32 %21 to float
  %23 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %23, 0
  %24 = call float @llvm.nvvm.fabs.ftz.f(float %22) #139
  %25 = call float @llvm.fabs.f32(float %22)
  %.01 = select i1 %.not, float %25, float %24
  %26 = fcmp olt float %.01, 0x4010C46560000000
  br i1 %26, label %41, label %__internal_fast_icmp_abs_lt.exit

__internal_fast_icmp_abs_lt.exit:                 ; preds = %__internal_fast_icmp_abs_lt.exit3
  %27 = fcmp olt double %a, 0.000000e+00
  %28 = fadd double %a, 0x7FF0000000000000
  %z.0 = select i1 %27, double 0.000000e+00, double %28
  %29 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %29, 0
  %.0 = select i1 %.not1, float %25, float %24
  %30 = fcmp olt float %.0, 0x4010E90000000000
  br i1 %30, label %31, label %41

31:                                               ; preds = %__internal_fast_icmp_abs_lt.exit
  %32 = sdiv i32 %1, 2
  %33 = shl i32 %32, 20
  %34 = add i32 %17, %33
  %35 = call double @llvm.nvvm.lohi.i2d(i32 %16, i32 %34) #139
  %36 = sub nsw i32 %1, %32
  %37 = shl i32 %36, 20
  %38 = add nsw i32 %37, 1072693248
  %39 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %38) #139
  %40 = fmul double %39, %35
  br label %41

41:                                               ; preds = %__internal_fast_icmp_abs_lt.exit, %31, %__internal_fast_icmp_abs_lt.exit3
  %z.2 = phi double [ %20, %__internal_fast_icmp_abs_lt.exit3 ], [ %40, %31 ], [ %z.0, %__internal_fast_icmp_abs_lt.exit ]
  ret double %z.2
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_exp2(double %a) local_unnamed_addr #81 {
__internal_fast_icmp_abs_lt.exit4:
  %0 = fadd double %a, 0x4338000000000000
  %1 = fadd double %0, 0xC338000000000000
  %2 = call i32 @llvm.nvvm.d2i.lo(double %0) #142
  %3 = fsub double %a, %1
  %4 = fmul double %3, 0x3C7ABC9E3B39803F
  %5 = call double @llvm.fma.f64(double %3, double noundef 0x3FE62E42FEFA39EF, double %4) #140
  %6 = call double @llvm.fma.f64(double %5, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %7 = call double @llvm.fma.f64(double %6, double %5, double noundef 0x3EC71DEE62401315) #140
  %8 = call double @llvm.fma.f64(double %7, double %5, double noundef 0x3EFA01997C89EB71) #140
  %9 = call double @llvm.fma.f64(double %8, double %5, double noundef 0x3F2A01A014761F65) #140
  %10 = call double @llvm.fma.f64(double %9, double %5, double noundef 0x3F56C16C1852B7AF) #140
  %11 = call double @llvm.fma.f64(double %10, double %5, double noundef 0x3F81111111122322) #140
  %12 = call double @llvm.fma.f64(double %11, double %5, double noundef 0x3FA55555555502A1) #140
  %13 = call double @llvm.fma.f64(double %12, double %5, double noundef 0x3FC5555555555511) #140
  %14 = call double @llvm.fma.f64(double %13, double %5, double noundef 0x3FE000000000000B) #140
  %15 = call double @llvm.fma.f64(double %14, double %5, double noundef 1.000000e+00) #140
  %16 = call double @llvm.fma.f64(double %15, double %5, double noundef 1.000000e+00) #140
  %17 = call i32 @llvm.nvvm.d2i.lo(double %16) #142
  %18 = call i32 @llvm.nvvm.d2i.hi(double %16) #142
  %19 = shl i32 %2, 20
  %20 = add i32 %18, %19
  %21 = call double @llvm.nvvm.lohi.i2d(i32 %17, i32 %20) #142
  %22 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %23 = bitcast i32 %22 to float
  %24 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %24, 0
  %25 = call float @llvm.nvvm.fabs.ftz.f(float %23) #139
  %26 = call float @llvm.fabs.f32(float %23)
  %.01 = select i1 %.not, float %26, float %25
  %27 = fcmp olt float %.01, 0x4011FE0000000000
  br i1 %27, label %45, label %__internal_fast_icmp_abs_lt.exit

__internal_fast_icmp_abs_lt.exit:                 ; preds = %__internal_fast_icmp_abs_lt.exit4
  %28 = icmp slt i32 %22, 0
  %29 = select i1 %28, double 0.000000e+00, double 0x7FF0000000000000
  %30 = call double @llvm.fabs.f64(double %a)
  %31 = fcmp ugt double %30, 0x7FF0000000000000
  %32 = fadd double %a, %a
  %z.0 = select i1 %31, double %32, double %29
  %33 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %33, 0
  %.0 = select i1 %.not1, float %26, float %25
  %34 = fcmp olt float %.0, 0x4012198000000000
  br i1 %34, label %35, label %45

35:                                               ; preds = %__internal_fast_icmp_abs_lt.exit
  %36 = sdiv i32 %2, 2
  %37 = shl i32 %36, 20
  %38 = add i32 %18, %37
  %39 = call double @llvm.nvvm.lohi.i2d(i32 %17, i32 %38) #139
  %40 = sub nsw i32 %2, %36
  %41 = shl i32 %40, 20
  %42 = add nsw i32 %41, 1072693248
  %43 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %42) #139
  %44 = fmul double %43, %39
  br label %45

45:                                               ; preds = %__internal_fast_icmp_abs_lt.exit, %35, %__internal_fast_icmp_abs_lt.exit4
  %z.2 = phi double [ %21, %__internal_fast_icmp_abs_lt.exit4 ], [ %44, %35 ], [ %z.0, %__internal_fast_icmp_abs_lt.exit ]
  ret double %z.2
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_exp10(double %a) local_unnamed_addr #82 {
__internal_fast_icmp_abs_lt.exit4:
  %0 = call double @llvm.fma.f64(double %a, double noundef 0x400A934F0979A371, double noundef 0x4338000000000000) #140
  %1 = call i32 @llvm.nvvm.d2i.lo(double %0) #142
  %2 = fadd double %0, 0xC338000000000000
  %3 = call double @llvm.fma.f64(double %2, double noundef 0xBFD34413509F79FF, double %a) #140
  %4 = call double @llvm.fma.f64(double %2, double noundef 0x3C49DC1DA994FD21, double %3) #140
  %5 = fmul double %4, 0xBCAF48AD494EA3E9
  %6 = call double @llvm.fma.f64(double %4, double noundef 0x40026BB1BBB55516, double %5) #140
  %7 = call double @llvm.fma.f64(double %6, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %8 = call double @llvm.fma.f64(double %7, double %6, double noundef 0x3EC71DEE62401315) #140
  %9 = call double @llvm.fma.f64(double %8, double %6, double noundef 0x3EFA01997C89EB71) #140
  %10 = call double @llvm.fma.f64(double %9, double %6, double noundef 0x3F2A01A014761F65) #140
  %11 = call double @llvm.fma.f64(double %10, double %6, double noundef 0x3F56C16C1852B7AF) #140
  %12 = call double @llvm.fma.f64(double %11, double %6, double noundef 0x3F81111111122322) #140
  %13 = call double @llvm.fma.f64(double %12, double %6, double noundef 0x3FA55555555502A1) #140
  %14 = call double @llvm.fma.f64(double %13, double %6, double noundef 0x3FC5555555555511) #140
  %15 = call double @llvm.fma.f64(double %14, double %6, double noundef 0x3FE000000000000B) #140
  %16 = call double @llvm.fma.f64(double %15, double %6, double noundef 1.000000e+00) #140
  %17 = call double @llvm.fma.f64(double %16, double %6, double noundef 1.000000e+00) #140
  %18 = call i32 @llvm.nvvm.d2i.lo(double %17) #142
  %19 = call i32 @llvm.nvvm.d2i.hi(double %17) #142
  %20 = shl i32 %1, 20
  %21 = add i32 %19, %20
  %22 = call double @llvm.nvvm.lohi.i2d(i32 %18, i32 %21) #142
  %23 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %24 = bitcast i32 %23 to float
  %25 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %25, 0
  %26 = call float @llvm.nvvm.fabs.ftz.f(float %24) #139
  %27 = call float @llvm.fabs.f32(float %24)
  %.01 = select i1 %.not, float %27, float %26
  %28 = fcmp olt float %.01, 0x400E674E20000000
  br i1 %28, label %46, label %__internal_fast_icmp_abs_lt.exit

__internal_fast_icmp_abs_lt.exit:                 ; preds = %__internal_fast_icmp_abs_lt.exit4
  %29 = icmp slt i32 %23, 0
  %30 = select i1 %29, double 0.000000e+00, double 0x7FF0000000000000
  %31 = call double @llvm.fabs.f64(double %a)
  %32 = fcmp ugt double %31, 0x7FF0000000000000
  %33 = fadd double %a, %a
  %z.0 = select i1 %32, double %33, double %30
  %34 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %34, 0
  %.0 = select i1 %.not1, float %27, float %26
  %35 = fcmp olt float %.0, 0x400E873700000000
  br i1 %35, label %36, label %46

36:                                               ; preds = %__internal_fast_icmp_abs_lt.exit
  %37 = sdiv i32 %1, 2
  %38 = shl i32 %37, 20
  %39 = add i32 %19, %38
  %40 = call double @llvm.nvvm.lohi.i2d(i32 %18, i32 %39) #139
  %41 = sub nsw i32 %1, %37
  %42 = shl i32 %41, 20
  %43 = add nsw i32 %42, 1072693248
  %44 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %43) #139
  %45 = fmul double %44, %40
  br label %46

46:                                               ; preds = %__internal_fast_icmp_abs_lt.exit, %36, %__internal_fast_icmp_abs_lt.exit4
  %z.2 = phi double [ %22, %__internal_fast_icmp_abs_lt.exit4 ], [ %45, %36 ], [ %z.0, %__internal_fast_icmp_abs_lt.exit ]
  ret double %z.2
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_expm1(double %a) local_unnamed_addr #83 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = bitcast i32 %1 to float
  %3 = fcmp olt float %2, 0x4010C5C860000000
  %4 = fcmp ogt float %2, 0xC009500000000000
  %or.cond = and i1 %3, %4
  br i1 %or.cond, label %__internal_expm1_scaled.exit, label %32

__internal_expm1_scaled.exit:                     ; preds = %0
  %5 = call double @llvm.fma.f64(double %a, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %6 = call i32 @llvm.nvvm.d2i.lo(double %5) #142
  %7 = fadd double %5, 0xC338000000000000
  %8 = call double @llvm.fma.f64(double %7, double noundef 0xBFE62E42FEFA39EF, double %a) #140
  %9 = call double @llvm.fma.f64(double %7, double noundef 0xBC7ABC9E3B39803F, double %8) #140
  %10 = shl i32 %1, 1
  %11 = icmp ult i32 %10, 2142496327
  %spec.select = select i1 %11, double %a, double %9
  %spec.select2 = select i1 %11, i32 0, i32 %6
  %12 = call double @llvm.fma.f64(double %spec.select, double noundef 0x3E21F4076ACD15B6, double noundef 0x3E5AF86D8EBD13CD) #140
  %13 = call double @llvm.fma.f64(double %12, double %spec.select, double noundef 0x3E927E5092BA033D) #140
  %14 = call double @llvm.fma.f64(double %13, double %spec.select, double noundef 0x3EC71DDE6C5F9DA1) #140
  %15 = call double @llvm.fma.f64(double %14, double %spec.select, double noundef 0x3EFA01A018D034E6) #140
  %16 = call double @llvm.fma.f64(double %15, double %spec.select, double noundef 0x3F2A01A01B3B6940) #140
  %17 = call double @llvm.fma.f64(double %16, double %spec.select, double noundef 0x3F56C16C16C1B5DD) #140
  %18 = call double @llvm.fma.f64(double %17, double %spec.select, double noundef 0x3F8111111110F74D) #140
  %19 = call double @llvm.fma.f64(double %18, double %spec.select, double noundef 0x3FA555555555554D) #140
  %20 = call double @llvm.fma.f64(double %19, double %spec.select, double noundef 0x3FC5555555555557) #140
  %21 = call double @llvm.fma.f64(double %20, double %spec.select, double noundef 5.000000e-01) #140
  %22 = fmul double %spec.select, %21
  %23 = call double @llvm.fma.f64(double %22, double %spec.select, double %spec.select) #140
  %24 = icmp eq i32 %spec.select2, 1024
  %spec.select2.op = shl i32 %spec.select2, 20
  %spec.select2.op.op = add i32 %spec.select2.op, 1072693248
  %25 = select i1 %24, i32 2145386496, i32 %spec.select2.op.op
  %26 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %25) #142
  %27 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 noundef 1072693248) #142
  %28 = fsub double %26, %27
  %29 = call double @llvm.fma.f64(double %23, double %26, double %28) #140
  %30 = select i1 %24, double %29, double -0.000000e+00
  %t.i.0 = fadd double %29, %30
  %31 = icmp eq i32 %10, 0
  %t.i.1 = select i1 %31, double %spec.select, double %t.i.0
  br label %38

32:                                               ; preds = %0
  %33 = icmp slt i32 %1, 0
  %34 = select i1 %33, double -1.000000e+00, double 0x7FF0000000000000
  %35 = call double @llvm.fabs.f64(double %a) #140
  %36 = fcmp ugt double %35, 0x7FF0000000000000
  %37 = fadd double %a, %a
  %t.0 = select i1 %36, double %37, double %34
  br label %38

38:                                               ; preds = %32, %__internal_expm1_scaled.exit
  %t.1 = phi double [ %t.i.1, %__internal_expm1_scaled.exit ], [ %t.0, %32 ]
  ret double %t.1
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_cosh(double %a) local_unnamed_addr #84 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = and i32 %1, 2147483647
  %3 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %4 = call double @llvm.nvvm.lohi.i2d(i32 %3, i32 %2) #142
  %5 = call i32 @llvm.nvvm.d2i.hi(double %4) #142
  %6 = icmp ult i32 %5, 1082536911
  br i1 %6, label %7, label %36

7:                                                ; preds = %0
  %8 = call double @llvm.fma.f64(double %4, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %9 = call i32 @llvm.nvvm.d2i.lo(double %8) #142
  %10 = fadd double %8, 0xC338000000000000
  %11 = call double @llvm.fma.f64(double %10, double noundef 0xBFE62E42FEFA39EF, double %4) #140
  %12 = call double @llvm.fma.f64(double %10, double noundef 0xBC7ABC9E3B39803F, double %11) #140
  %13 = call double @llvm.fma.f64(double %12, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %14 = call double @llvm.fma.f64(double %13, double %12, double noundef 0x3EC71DEE62401315) #140
  %15 = call double @llvm.fma.f64(double %14, double %12, double noundef 0x3EFA01997C89EB71) #140
  %16 = call double @llvm.fma.f64(double %15, double %12, double noundef 0x3F2A01A014761F65) #140
  %17 = call double @llvm.fma.f64(double %16, double %12, double noundef 0x3F56C16C1852B7AF) #140
  %18 = call double @llvm.fma.f64(double %17, double %12, double noundef 0x3F81111111122322) #140
  %19 = call double @llvm.fma.f64(double %18, double %12, double noundef 0x3FA55555555502A1) #140
  %20 = call double @llvm.fma.f64(double %19, double %12, double noundef 0x3FC5555555555511) #140
  %21 = call double @llvm.fma.f64(double %20, double %12, double noundef 0x3FE000000000000B) #140
  %22 = call double @llvm.fma.f64(double %21, double %12, double noundef 1.000000e+00) #140
  %23 = call double @llvm.fma.f64(double %22, double %12, double noundef 1.000000e+00) #140
  %24 = call i32 @llvm.nvvm.d2i.lo(double %23) #142
  %25 = call i32 @llvm.nvvm.d2i.hi(double %23) #142
  %26 = shl i32 %9, 20
  %27 = add i32 %26, -2097152
  %28 = add i32 %27, %25
  %29 = call double @llvm.nvvm.lohi.i2d(i32 %24, i32 %28) #142
  %30 = call double @llvm.nvvm.rcp.approx.ftz.d(double %29) #142
  %31 = fneg double %29
  %32 = call double @llvm.fma.f64(double %31, double %30, double noundef 1.000000e+00) #140
  %33 = call double @llvm.fma.f64(double %32, double %32, double %32) #140
  %34 = call double @llvm.fma.f64(double %33, double %30, double %30) #140
  %35 = call double @llvm.fma.f64(double %34, double noundef 6.250000e-02, double %29) #140
  br label %38

36:                                               ; preds = %0
  %37 = fcmp ole double %a, 0x7FF0000000000000
  %.0 = select i1 %37, double 0x7FF0000000000000, double %a
  br label %38

38:                                               ; preds = %36, %7
  %.1 = phi double [ %35, %7 ], [ %.0, %36 ]
  %39 = fadd double %.1, %.1
  ret double %39
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_sinh(double %a) local_unnamed_addr #85 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = and i32 %1, 2147483647
  %3 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %4 = call double @llvm.nvvm.lohi.i2d(i32 %3, i32 %2) #142
  %5 = icmp ult i32 %2, 1072693248
  br i1 %5, label %6, label %__internal_expm1_scaled.exit

6:                                                ; preds = %0
  %7 = fmul double %4, %4
  %8 = call double @llvm.fma.f64(double %7, double noundef 0x3D6B4C75AB274C53, double noundef 0x3DE611A561D87DEF) #140
  %9 = call double @llvm.fma.f64(double %8, double %7, double noundef 0x3E5AE64671B18F5C) #140
  %10 = call double @llvm.fma.f64(double %9, double %7, double noundef 0x3EC71DE3A465B1E4) #140
  %11 = call double @llvm.fma.f64(double %10, double %7, double noundef 0x3F2A01A01A02899D) #140
  %12 = call double @llvm.fma.f64(double %11, double %7, double noundef 0x3F811111111110A6) #140
  %13 = call double @llvm.fma.f64(double %12, double %7, double noundef 0x3FC5555555555556) #140
  %14 = fmul double %7, %13
  %15 = call double @llvm.fma.f64(double %14, double %4, double %4) #140
  br label %49

__internal_expm1_scaled.exit:                     ; preds = %0
  %16 = call i32 @llvm.nvvm.d2i.hi(double %4) #142
  %17 = call double @llvm.fma.f64(double %4, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %18 = call i32 @llvm.nvvm.d2i.lo(double %17) #142
  %19 = add nsw i32 %18, -1
  %20 = fadd double %17, 0xC338000000000000
  %21 = call double @llvm.fma.f64(double %20, double noundef 0xBFE62E42FEFA39EF, double %4) #140
  %22 = call double @llvm.fma.f64(double %20, double noundef 0xBC7ABC9E3B39803F, double %21) #140
  %23 = shl i32 %16, 1
  %24 = icmp ult i32 %23, 2142496327
  %spec.select = select i1 %24, double %4, double %22
  %spec.select1 = select i1 %24, i32 0, i32 %19
  %25 = call double @llvm.fma.f64(double %spec.select, double noundef 0x3E21F4076ACD15B6, double noundef 0x3E5AF86D8EBD13CD) #140
  %26 = call double @llvm.fma.f64(double %25, double %spec.select, double noundef 0x3E927E5092BA033D) #140
  %27 = call double @llvm.fma.f64(double %26, double %spec.select, double noundef 0x3EC71DDE6C5F9DA1) #140
  %28 = call double @llvm.fma.f64(double %27, double %spec.select, double noundef 0x3EFA01A018D034E6) #140
  %29 = call double @llvm.fma.f64(double %28, double %spec.select, double noundef 0x3F2A01A01B3B6940) #140
  %30 = call double @llvm.fma.f64(double %29, double %spec.select, double noundef 0x3F56C16C16C1B5DD) #140
  %31 = call double @llvm.fma.f64(double %30, double %spec.select, double noundef 0x3F8111111110F74D) #140
  %32 = call double @llvm.fma.f64(double %31, double %spec.select, double noundef 0x3FA555555555554D) #140
  %33 = call double @llvm.fma.f64(double %32, double %spec.select, double noundef 0x3FC5555555555557) #140
  %34 = call double @llvm.fma.f64(double %33, double %spec.select, double noundef 5.000000e-01) #140
  %35 = fmul double %spec.select, %34
  %36 = call double @llvm.fma.f64(double %35, double %spec.select, double %spec.select) #140
  %37 = icmp eq i32 %spec.select1, 1024
  %spec.select1.op = shl i32 %spec.select1, 20
  %spec.select1.op.op = add i32 %spec.select1.op, 1072693248
  %38 = select i1 %37, i32 2145386496, i32 %spec.select1.op.op
  %39 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %38) #142
  %40 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 noundef 1071644672) #142
  %41 = fsub double %39, %40
  %42 = call double @llvm.fma.f64(double %36, double %39, double %41) #140
  %43 = select i1 %37, double %42, double -0.000000e+00
  %t.i.0 = fadd double %42, %43
  %44 = icmp eq i32 %23, 0
  %t.i.1 = select i1 %44, double %spec.select, double %t.i.0
  %45 = call double @llvm.fma.f64(double %t.i.1, double noundef 2.000000e+00, double noundef 1.000000e+00) #140
  %46 = fdiv double %t.i.1, %45
  %47 = fadd double %t.i.1, %46
  %48 = fcmp oge double %4, 0x408633CE8FB9F87E
  %z.0 = select i1 %48, double 0x7FF0000000000000, double %47
  br label %49

49:                                               ; preds = %__internal_expm1_scaled.exit, %6
  %z.1 = phi double [ %15, %6 ], [ %z.0, %__internal_expm1_scaled.exit ]
  %50 = call i32 @llvm.nvvm.d2i.lo(double %z.1) #142
  %51 = call i32 @llvm.nvvm.d2i.hi(double %z.1) #142
  %52 = and i32 %1, -2147483648
  %53 = or i32 %51, %52
  %54 = call double @llvm.nvvm.lohi.i2d(i32 %50, i32 %53) #142
  ret double %54
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_tanh(double %a) local_unnamed_addr #86 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = and i32 %1, 2147483647
  %3 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %4 = call double @llvm.nvvm.lohi.i2d(i32 %3, i32 %2) #142
  %5 = fcmp ult double %4, 0x3FE4F92224DD2F1A
  br i1 %5, label %43, label %__internal_mexpm1.exit

__internal_mexpm1.exit:                           ; preds = %0
  %6 = fmul double %4, 2.000000e+00
  %7 = fptrunc double %6 to float
  %8 = fmul float %7, 0x3FF7154760000000
  %9 = call float @llvm.round.f32(float %8) #140
  %10 = fpext float %9 to double
  %11 = fneg double %10
  %12 = call double @llvm.fma.f64(double %11, double noundef 0x3FE62E42FEFA39EF, double %6) #140
  %13 = call double @llvm.fma.f64(double %12, double noundef 0x3E5AE904A4741B81, double noundef 0x3E928A27F89B6999) #140
  %14 = call double @llvm.fma.f64(double %13, double %12, double noundef 0x3EC71DE715FF7E07) #140
  %15 = call double @llvm.fma.f64(double %14, double %12, double noundef 0x3EFA019A6B0AC45A) #140
  %16 = call double @llvm.fma.f64(double %15, double %12, double noundef 0x3F2A01A017EED94F) #140
  %17 = call double @llvm.fma.f64(double %16, double %12, double noundef 0x3F56C16C17F2A71B) #140
  %18 = call double @llvm.fma.f64(double %17, double %12, double noundef 0x3F811111111173C4) #140
  %19 = call double @llvm.fma.f64(double %18, double %12, double noundef 0x3FA555555555211A) #140
  %20 = call double @llvm.fma.f64(double %19, double %12, double noundef 0x3FC5555555555540) #140
  %21 = call double @llvm.fma.f64(double %20, double %12, double noundef 0x3FE0000000000005) #140
  %22 = fmul double %12, %21
  %23 = call double @llvm.fma.f64(double %22, double %12, double %12) #140
  %24 = call float @llvm.nvvm.ex2.approx.ftz.f(float %9) #142
  %25 = fpext float %24 to double
  %26 = fsub double 1.000000e+00, %25
  %27 = fneg double %23
  %28 = call double @llvm.fma.f64(double %27, double %25, double %26) #140
  %29 = fsub double 2.000000e+00, %28
  %30 = call double @llvm.nvvm.rcp.approx.ftz.d(double %29) #142
  %31 = fneg double %29
  %32 = call double @llvm.fma.f64(double %31, double %30, double noundef 1.000000e+00) #140
  %33 = call double @llvm.fma.f64(double %32, double %32, double %32) #140
  %34 = call double @llvm.fma.f64(double %33, double %30, double %30) #140
  %35 = fneg double %34
  %36 = call double @llvm.fma.f64(double %35, double noundef 2.000000e+00, double noundef 1.000000e+00) #140
  %37 = icmp ugt i32 %2, 1077088193
  %r.0 = select i1 %37, double 1.000000e+00, double %36
  %38 = call i32 @llvm.nvvm.d2i.lo(double %r.0) #142
  %39 = call i32 @llvm.nvvm.d2i.hi(double %r.0) #142
  %40 = and i32 %1, -2147483648
  %41 = or i32 %39, %40
  %42 = call double @llvm.nvvm.lohi.i2d(i32 %38, i32 %41) #142
  br label %57

43:                                               ; preds = %0
  %44 = fmul double %a, %a
  %45 = call double @llvm.fma.f64(double %44, double noundef 0xBEF0BC46E2F5E964, double noundef 0x3F14359F420AFC3D) #140
  %46 = call double @llvm.fma.f64(double %45, double %44, double noundef 0xBF2DF9F0728C5D84) #140
  %47 = call double @llvm.fma.f64(double %46, double %44, double noundef 0x3F4337D1CEC4F033) #140
  %48 = call double @llvm.fma.f64(double %47, double %44, double noundef 0xBF57D6E9674335B3) #140
  %49 = call double @llvm.fma.f64(double %48, double %44, double noundef 0x3F6D6D000D7AAD3D) #140
  %50 = call double @llvm.fma.f64(double %49, double %44, double noundef 0xBF8226E1F3CF1EF5) #140
  %51 = call double @llvm.fma.f64(double %50, double %44, double noundef 0x3F9664F47EC0C8CF) #140
  %52 = call double @llvm.fma.f64(double %51, double %44, double noundef 0xBFABA1BA1B80AB40) #140
  %53 = call double @llvm.fma.f64(double %52, double %44, double noundef 0x3FC111111110FA4A) #140
  %54 = call double @llvm.fma.f64(double %53, double %44, double noundef 0xBFD5555555555550) #140
  %55 = call double @llvm.fma.f64(double %54, double %44, double noundef 0.000000e+00) #140
  %56 = call double @llvm.fma.f64(double %55, double %a, double %a) #140
  br label %57

57:                                               ; preds = %43, %__internal_mexpm1.exit
  %r.1 = phi double [ %42, %__internal_mexpm1.exit ], [ %56, %43 ]
  ret double %r.1
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_atan2(double %a, double %b) local_unnamed_addr #87 {
  %1 = call double @llvm.fabs.f64(double %b) #140
  %2 = call double @llvm.fabs.f64(double %a) #140
  %3 = fcmp oeq double %b, 0.000000e+00
  %4 = fcmp oeq double %a, 0.000000e+00
  %or.cond = select i1 %3, i1 %4, i1 false
  br i1 %or.cond, label %5, label %15

5:                                                ; preds = %0
  %6 = call i32 @llvm.nvvm.d2i.hi(double %b) #142
  %7 = icmp slt i32 %6, 0
  %8 = select i1 %7, double 0x400921FB54442D18, double 0.000000e+00
  %9 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %10 = call i32 @llvm.nvvm.d2i.lo(double noundef %8) #142
  %11 = call i32 @llvm.nvvm.d2i.hi(double noundef %8) #142
  %12 = and i32 %9, -2147483648
  %13 = or i32 %11, %12
  %14 = call double @llvm.nvvm.lohi.i2d(i32 %10, i32 %13) #142
  br label %66

15:                                               ; preds = %0
  %16 = fcmp oeq double %1, 0x7FF0000000000000
  %17 = fcmp oeq double %2, 0x7FF0000000000000
  %or.cond2 = select i1 %16, i1 %17, i1 false
  br i1 %or.cond2, label %18, label %28

18:                                               ; preds = %15
  %19 = call i32 @llvm.nvvm.d2i.hi(double %b) #142
  %20 = icmp slt i32 %19, 0
  %21 = select i1 %20, double 0x4002D97C7F3321D2, double 0x3FE921FB54442D18
  %22 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %23 = call i32 @llvm.nvvm.d2i.lo(double noundef %21) #142
  %24 = call i32 @llvm.nvvm.d2i.hi(double noundef %21) #142
  %25 = and i32 %22, -2147483648
  %26 = or i32 %24, %25
  %27 = call double @llvm.nvvm.lohi.i2d(i32 %23, i32 %26) #142
  br label %66

28:                                               ; preds = %15
  %29 = call double @llvm.maxnum.f64(double %2, double %1) #140
  %30 = call double @llvm.minnum.f64(double %2, double %1) #140
  %31 = fdiv double %30, %29
  %32 = fmul double %31, %31
  %33 = call double @llvm.fma.f64(double %32, double noundef 0xBEF53E1D2A25FF7E, double noundef 0x3F2D3B63DBB65B49) #140
  %34 = call double @llvm.fma.f64(double %33, double %32, double noundef 0xBF5312788DDE082E) #140
  %35 = call double @llvm.fma.f64(double %34, double %32, double noundef 0x3F6F9690C8249315) #140
  %36 = call double @llvm.fma.f64(double %35, double %32, double noundef 0xBF82CF5AABC7CF0D) #140
  %37 = call double @llvm.fma.f64(double %36, double %32, double noundef 0x3F9162B0B2A3BFDE) #140
  %38 = call double @llvm.fma.f64(double %37, double %32, double noundef 0xBF9A7256FEB6FC6B) #140
  %39 = call double @llvm.fma.f64(double %38, double %32, double noundef 0x3FA171560CE4A489) #140
  %40 = call double @llvm.fma.f64(double %39, double %32, double noundef 0xBFA4F44D841450E4) #140
  %41 = call double @llvm.fma.f64(double %40, double %32, double noundef 0x3FA7EE3D3F36BB95) #140
  %42 = call double @llvm.fma.f64(double %41, double %32, double noundef 0xBFAAD32AE04A9FD1) #140
  %43 = call double @llvm.fma.f64(double %42, double %32, double noundef 0x3FAE17813D66954F) #140
  %44 = call double @llvm.fma.f64(double %43, double %32, double noundef 0xBFB11089CA9A5BCD) #140
  %45 = call double @llvm.fma.f64(double %44, double %32, double noundef 0x3FB3B12B2DB51738) #140
  %46 = call double @llvm.fma.f64(double %45, double %32, double noundef 0xBFB745D022F8DC5C) #140
  %47 = call double @llvm.fma.f64(double %46, double %32, double noundef 0x3FBC71C709DFE927) #140
  %48 = call double @llvm.fma.f64(double %47, double %32, double noundef 0xBFC2492491FA1744) #140
  %49 = call double @llvm.fma.f64(double %48, double %32, double noundef 0x3FC99999999840D2) #140
  %50 = call double @llvm.fma.f64(double %49, double %32, double noundef 0xBFD555555555544C) #140
  %51 = fmul double %32, %50
  %52 = call double @llvm.fma.f64(double %51, double %31, double %31) #140
  %53 = fcmp ogt double %2, %1
  %54 = fsub double 0x3FF921FB54442D18, %52
  %t0.0 = select i1 %53, double %54, double %52
  %55 = call i32 @llvm.nvvm.d2i.hi(double %b) #142
  %56 = icmp slt i32 %55, 0
  %57 = fsub double 0x400921FB54442D18, %t0.0
  %t0.1 = select i1 %56, double %57, double %t0.0
  %58 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %59 = call i32 @llvm.nvvm.d2i.lo(double %t0.1) #142
  %60 = call i32 @llvm.nvvm.d2i.hi(double %t0.1) #142
  %61 = and i32 %58, -2147483648
  %62 = or i32 %60, %61
  %63 = call double @llvm.nvvm.lohi.i2d(i32 %59, i32 %62) #142
  %64 = fadd double %2, %1
  %65 = fcmp ole double %64, 0x7FF0000000000000
  %t0.2 = select i1 %65, double %63, double %64
  br label %66

66:                                               ; preds = %18, %28, %5
  %t0.4 = phi double [ %14, %5 ], [ %27, %18 ], [ %t0.2, %28 ]
  ret double %t0.4
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_atan(double %a) local_unnamed_addr #88 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ogt double %1, 1.000000e+00
  br i1 %2, label %3, label %10

3:                                                ; preds = %0
  %4 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #142
  %5 = fneg double %1
  %6 = call double @llvm.fma.f64(double %5, double %4, double noundef 1.000000e+00) #140
  %7 = call double @llvm.fma.f64(double %6, double %6, double %6) #140
  %8 = call double @llvm.fma.f64(double %7, double %4, double %4) #140
  %9 = fcmp oeq double %1, 0x7FF0000000000000
  %t1.0 = select i1 %9, double 0.000000e+00, double %8
  br label %10

10:                                               ; preds = %3, %0
  %t1.1 = phi double [ %t1.0, %3 ], [ %1, %0 ]
  %11 = fmul double %t1.1, %t1.1
  %12 = call double @llvm.fma.f64(double %11, double noundef 0xBEF53E1D2A25FF7E, double noundef 0x3F2D3B63DBB65B49) #140
  %13 = call double @llvm.fma.f64(double %12, double %11, double noundef 0xBF5312788DDE082E) #140
  %14 = call double @llvm.fma.f64(double %13, double %11, double noundef 0x3F6F9690C8249315) #140
  %15 = call double @llvm.fma.f64(double %14, double %11, double noundef 0xBF82CF5AABC7CF0D) #140
  %16 = call double @llvm.fma.f64(double %15, double %11, double noundef 0x3F9162B0B2A3BFDE) #140
  %17 = call double @llvm.fma.f64(double %16, double %11, double noundef 0xBF9A7256FEB6FC6B) #140
  %18 = call double @llvm.fma.f64(double %17, double %11, double noundef 0x3FA171560CE4A489) #140
  %19 = call double @llvm.fma.f64(double %18, double %11, double noundef 0xBFA4F44D841450E4) #140
  %20 = call double @llvm.fma.f64(double %19, double %11, double noundef 0x3FA7EE3D3F36BB95) #140
  %21 = call double @llvm.fma.f64(double %20, double %11, double noundef 0xBFAAD32AE04A9FD1) #140
  %22 = call double @llvm.fma.f64(double %21, double %11, double noundef 0x3FAE17813D66954F) #140
  %23 = call double @llvm.fma.f64(double %22, double %11, double noundef 0xBFB11089CA9A5BCD) #140
  %24 = call double @llvm.fma.f64(double %23, double %11, double noundef 0x3FB3B12B2DB51738) #140
  %25 = call double @llvm.fma.f64(double %24, double %11, double noundef 0xBFB745D022F8DC5C) #140
  %26 = call double @llvm.fma.f64(double %25, double %11, double noundef 0x3FBC71C709DFE927) #140
  %27 = call double @llvm.fma.f64(double %26, double %11, double noundef 0xBFC2492491FA1744) #140
  %28 = call double @llvm.fma.f64(double %27, double %11, double noundef 0x3FC99999999840D2) #140
  %29 = call double @llvm.fma.f64(double %28, double %11, double noundef 0xBFD555555555544C) #140
  %30 = fmul double %11, %29
  %31 = call double @llvm.fma.f64(double %30, double %t1.1, double %t1.1) #140
  %32 = fsub double 0x3FF921FB54442D18, %31
  %t1.2 = select i1 %2, double %32, double %31
  %33 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %34 = call i32 @llvm.nvvm.d2i.lo(double %t1.2) #142
  %35 = call i32 @llvm.nvvm.d2i.hi(double %t1.2) #142
  %36 = and i32 %33, -2147483648
  %37 = or i32 %35, %36
  %38 = call double @llvm.nvvm.lohi.i2d(i32 %34, i32 %37) #142
  ret double %38
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_asin(double %a) local_unnamed_addr #89 {
__nv_fabsf.exit:
  %0 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %1 = bitcast i32 %0 to float
  %2 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %2, 0
  %3 = call float @llvm.nvvm.fabs.ftz.f(float %1) #139
  %4 = call float @llvm.fabs.f32(float %1)
  %.0 = select i1 %.not, float %4, float %3
  %5 = fcmp olt float %.0, 0x3FFC4CCCC0000000
  br i1 %5, label %6, label %22

6:                                                ; preds = %__nv_fabsf.exit
  %7 = fmul double %a, %a
  %8 = call double @llvm.fma.f64(double %7, double noundef 0x3FB0066BDC1895E9, double noundef 0xBFB3823B180754AF)
  %9 = call double @llvm.fma.f64(double %8, double %7, double noundef 0x3FB11E52CC2F79AE)
  %10 = call double @llvm.fma.f64(double %9, double %7, double noundef 0xBF924EAF3526861B)
  %11 = call double @llvm.fma.f64(double %10, double %7, double noundef 0x3F91DF02A31E6CB7)
  %12 = call double @llvm.fma.f64(double %11, double %7, double noundef 0x3F847D18B0EEC6CC)
  %13 = call double @llvm.fma.f64(double %12, double %7, double noundef 0x3F8D0AF961BA53B0)
  %14 = call double @llvm.fma.f64(double %13, double %7, double noundef 0x3F91BF7734CF1C48)
  %15 = call double @llvm.fma.f64(double %14, double %7, double noundef 0x3F96E91483144EF7)
  %16 = call double @llvm.fma.f64(double %15, double %7, double noundef 0x3F9F1C6E0A4F9F81)
  %17 = call double @llvm.fma.f64(double %16, double %7, double noundef 0x3FA6DB6DC27FA92B)
  %18 = call double @llvm.fma.f64(double %17, double %7, double noundef 0x3FB333333320F91B)
  %19 = call double @llvm.fma.f64(double %18, double %7, double noundef 0x3FC5555555555F4D)
  %20 = fmul double %7, %19
  %21 = call double @llvm.fma.f64(double %20, double %a, double %a)
  br label %65

22:                                               ; preds = %__nv_fabsf.exit
  %23 = call double @llvm.fabs.f64(double %a)
  %24 = call double @llvm.fma.f64(double %23, double noundef -5.000000e-01, double noundef 5.000000e-01)
  %25 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %24) #141, !srcloc !83
  %26 = call i32 @llvm.nvvm.d2i.lo(double %25) #139
  %27 = call i32 @llvm.nvvm.d2i.hi(double %25) #139
  %28 = add i32 %27, -1048576
  %29 = call double @llvm.nvvm.lohi.i2d(i32 %26, i32 %28) #139
  %30 = fmul double %24, %25
  %31 = fneg double %30
  %32 = call double @llvm.fma.f64(double %30, double %31, double %24)
  %33 = call double @llvm.fma.f64(double %32, double %29, double %30)
  %34 = fneg double %33
  %35 = call double @llvm.fma.f64(double %25, double %34, double noundef 1.000000e+00)
  %36 = call double @llvm.fma.f64(double %35, double %29, double %29)
  %37 = call double @llvm.fma.f64(double %33, double %34, double %24)
  %38 = call double @llvm.fma.f64(double %37, double %36, double %33)
  %39 = call i32 @llvm.nvvm.d2i.hi(double %24) #139
  %40 = icmp slt i32 %39, 0
  %spec.select = select i1 %40, double 0xFFF8000000000000, double %38
  %41 = fcmp one double %24, 0.000000e+00
  %t0.1 = select i1 %41, double %spec.select, double %24
  %42 = call double @llvm.fma.f64(double %24, double noundef 0x3FB0066BDC1895E9, double noundef 0xBFB3823B180754AF)
  %43 = call double @llvm.fma.f64(double %42, double %24, double noundef 0x3FB11E52CC2F79AE)
  %44 = call double @llvm.fma.f64(double %43, double %24, double noundef 0xBF924EAF3526861B)
  %45 = call double @llvm.fma.f64(double %44, double %24, double noundef 0x3F91DF02A31E6CB7)
  %46 = call double @llvm.fma.f64(double %45, double %24, double noundef 0x3F847D18B0EEC6CC)
  %47 = call double @llvm.fma.f64(double %46, double %24, double noundef 0x3F8D0AF961BA53B0)
  %48 = call double @llvm.fma.f64(double %47, double %24, double noundef 0x3F91BF7734CF1C48)
  %49 = call double @llvm.fma.f64(double %48, double %24, double noundef 0x3F96E91483144EF7)
  %50 = call double @llvm.fma.f64(double %49, double %24, double noundef 0x3F9F1C6E0A4F9F81)
  %51 = call double @llvm.fma.f64(double %50, double %24, double noundef 0x3FA6DB6DC27FA92B)
  %52 = call double @llvm.fma.f64(double %51, double %24, double noundef 0x3FB333333320F91B)
  %53 = call double @llvm.fma.f64(double %52, double %24, double noundef 0x3FC5555555555F4D)
  %54 = fmul double %24, %53
  %55 = fmul double %t0.1, -2.000000e+00
  %56 = call double @llvm.fma.f64(double %55, double %54, double noundef 0x3C91A62633145C07)
  %57 = fadd double %55, 0x3FE921FB54442D18
  %58 = fadd double %57, %56
  %59 = fadd double %58, 0x3FE921FB54442D18
  %60 = call i32 @llvm.nvvm.d2i.lo(double %59) #139
  %61 = call i32 @llvm.nvvm.d2i.hi(double %59) #139
  %62 = and i32 %0, -2147483648
  %63 = or i32 %61, %62
  %64 = call double @llvm.nvvm.lohi.i2d(i32 %60, i32 %63) #139
  br label %65

65:                                               ; preds = %22, %6
  %t1.0 = phi double [ %21, %6 ], [ %64, %22 ]
  ret double %t1.0
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_acos(double %a) local_unnamed_addr #90 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call double @llvm.fabs.f64(double %a) #140
  %3 = call i32 @llvm.nvvm.d2i.hi(double %2) #142
  %4 = icmp slt i32 %3, 1071801958
  br i1 %4, label %5, label %28

5:                                                ; preds = %0
  %6 = fmul double %a, %a
  %7 = call double @llvm.fma.f64(double %6, double noundef 0x3FB0066BDC1895E9, double noundef 0xBFB3823B180754AF) #140
  %8 = call double @llvm.fma.f64(double %7, double %6, double noundef 0x3FB11E52CC2F79AE) #140
  %9 = call double @llvm.fma.f64(double %8, double %6, double noundef 0xBF924EAF3526861B) #140
  %10 = call double @llvm.fma.f64(double %9, double %6, double noundef 0x3F91DF02A31E6CB7) #140
  %11 = call double @llvm.fma.f64(double %10, double %6, double noundef 0x3F847D18B0EEC6CC) #140
  %12 = call double @llvm.fma.f64(double %11, double %6, double noundef 0x3F8D0AF961BA53B0) #140
  %13 = call double @llvm.fma.f64(double %12, double %6, double noundef 0x3F91BF7734CF1C48) #140
  %14 = call double @llvm.fma.f64(double %13, double %6, double noundef 0x3F96E91483144EF7) #140
  %15 = call double @llvm.fma.f64(double %14, double %6, double noundef 0x3F9F1C6E0A4F9F81) #140
  %16 = call double @llvm.fma.f64(double %15, double %6, double noundef 0x3FA6DB6DC27FA92B) #140
  %17 = call double @llvm.fma.f64(double %16, double %6, double noundef 0x3FB333333320F91B) #140
  %18 = call double @llvm.fma.f64(double %17, double %6, double noundef 0x3FC5555555555F4D) #140
  %19 = fmul double %6, %18
  %20 = call double @llvm.fma.f64(double %19, double %2, double %2) #140
  %21 = icmp slt i32 %1, 0
  br i1 %21, label %22, label %25

22:                                               ; preds = %5
  %23 = fadd double %20, 0x3C91A62633145C07
  %24 = fadd double %23, 0x3FF921FB54442D18
  br label %73

25:                                               ; preds = %5
  %26 = fadd double %20, 0xBC91A62633145C07
  %27 = fsub double 0x3FF921FB54442D18, %26
  br label %73

28:                                               ; preds = %0
  %29 = fsub double 1.000000e+00, %2
  %30 = call i32 @llvm.nvvm.d2i.lo(double %29) #142
  %31 = call i32 @llvm.nvvm.d2i.hi(double %29) #142
  %32 = add i32 %31, -1048576
  %33 = call double @llvm.nvvm.lohi.i2d(i32 %30, i32 %32) #142
  %34 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %33) #141, !srcloc !83
  %35 = call i32 @llvm.nvvm.d2i.lo(double %34) #139
  %36 = call i32 @llvm.nvvm.d2i.hi(double %34) #139
  %37 = add i32 %36, -1048576
  %38 = call double @llvm.nvvm.lohi.i2d(i32 %35, i32 %37) #139
  %39 = fmul double %33, %34
  %40 = fneg double %39
  %41 = call double @llvm.fma.f64(double %39, double %40, double %33)
  %42 = call double @llvm.fma.f64(double %41, double %38, double %39)
  %43 = fneg double %42
  %44 = call double @llvm.fma.f64(double %34, double %43, double noundef 1.000000e+00)
  %45 = call double @llvm.fma.f64(double %44, double %38, double %38)
  %46 = call double @llvm.fma.f64(double %42, double %43, double %33)
  %47 = call double @llvm.fma.f64(double %46, double %45, double %42)
  %48 = call i32 @llvm.nvvm.d2i.lo(double %47) #139
  %49 = call i32 @llvm.nvvm.d2i.hi(double %47) #139
  %50 = add i32 %49, 1048576
  %51 = call double @llvm.nvvm.lohi.i2d(i32 %48, i32 %50) #139
  %52 = call double @llvm.fma.f64(double %29, double noundef 0x3EC715B371155F70, double noundef 0xBEBAC2FE66FAAC4B)
  %53 = call double @llvm.fma.f64(double %52, double %29, double noundef 0x3ED9A9B88EFCD9B8)
  %54 = call double @llvm.fma.f64(double %53, double %29, double noundef 0x3EDD0F40A8A0C4C3)
  %55 = call double @llvm.fma.f64(double %54, double %29, double noundef 0x3EF46D4CFA9E0E1F)
  %56 = call double @llvm.fma.f64(double %55, double %29, double noundef 0x3F079C168D1E2422)
  %57 = call double @llvm.fma.f64(double %56, double %29, double noundef 0x3F1C9A88C3BCA540)
  %58 = call double @llvm.fma.f64(double %57, double %29, double noundef 0x3F31C4E64BD476DF)
  %59 = call double @llvm.fma.f64(double %58, double %29, double noundef 0x3F46E8BA60009C8F)
  %60 = call double @llvm.fma.f64(double %59, double %29, double noundef 0x3F5F1C71C62B05A2)
  %61 = call double @llvm.fma.f64(double %60, double %29, double noundef 0x3F76DB6DB6DC9F2C)
  %62 = call double @llvm.fma.f64(double %61, double %29, double noundef 0x3F9333333333329C)
  %63 = call double @llvm.fma.f64(double %62, double %29, double noundef 0x3FB5555555555555)
  %64 = icmp slt i32 %31, 1
  %65 = fmul double %2, 0.000000e+00
  %66 = fmul double %29, %63
  %67 = call double @llvm.fma.f64(double %66, double %51, double %51)
  %t0.1 = select i1 %64, double %65, double %67
  %68 = icmp slt i32 %31, 0
  %69 = fmul double %t0.1, 0x7FF0000000000000
  %t0.2 = select i1 %68, double %69, double %t0.1
  %70 = icmp slt i32 %1, 0
  %71 = fadd double %t0.2, 0xBCA1A62633145C07
  %72 = fsub double 0x400921FB54442D18, %71
  %t0.3 = select i1 %70, double %72, double %t0.2
  br label %73

73:                                               ; preds = %22, %25, %28
  %t0.4 = phi double [ %t0.3, %28 ], [ %24, %22 ], [ %27, %25 ]
  ret double %t0.4
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_acosh(double %a) local_unnamed_addr #91 {
  %1 = fadd double %a, -1.000000e+00
  %2 = call i32 @llvm.nvvm.d2i.hi(double %1) #142
  %3 = icmp ugt i32 %2, 1127219199
  br i1 %3, label %4, label %69

4:                                                ; preds = %0
  %5 = call i32 @llvm.nvvm.d2i.lo(double %1) #142
  %6 = icmp slt i32 %2, 1048576
  br i1 %6, label %7, label %11

7:                                                ; preds = %4
  %8 = fmul double %1, 0x4350000000000000
  %9 = call i32 @llvm.nvvm.d2i.hi(double %8) #142
  %10 = call i32 @llvm.nvvm.d2i.lo(double %8) #142
  br label %11

11:                                               ; preds = %7, %4
  %.01 = phi double [ %8, %7 ], [ %1, %4 ]
  %ihi.i.0 = phi i32 [ %9, %7 ], [ %2, %4 ]
  %ilo.i.0 = phi i32 [ %10, %7 ], [ %5, %4 ]
  %e.i.0 = phi i32 [ -1077, %7 ], [ -1023, %4 ]
  %12 = add i32 %ihi.i.0, -1
  %13 = icmp ult i32 %12, 2146435071
  br i1 %13, label %14, label %63

14:                                               ; preds = %11
  %15 = lshr i32 %ihi.i.0, 20
  %16 = add nsw i32 %e.i.0, %15
  %17 = and i32 %ihi.i.0, -2146435073
  %18 = or i32 %17, 1072693248
  %19 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %18) #142
  %20 = icmp sgt i32 %18, 1073127582
  br i1 %20, label %21, label %27

21:                                               ; preds = %14
  %22 = call i32 @llvm.nvvm.d2i.lo(double %19) #142
  %23 = call i32 @llvm.nvvm.d2i.hi(double %19) #142
  %24 = add i32 %23, -1048576
  %25 = call double @llvm.nvvm.lohi.i2d(i32 %22, i32 %24) #142
  %26 = add nsw i32 %16, 1
  br label %27

27:                                               ; preds = %21, %14
  %m.i.0 = phi double [ %25, %21 ], [ %19, %14 ]
  %e.i.1 = phi i32 [ %26, %21 ], [ %16, %14 ]
  %28 = fadd double %m.i.0, -1.000000e+00
  %29 = fadd double %m.i.0, 1.000000e+00
  %30 = call double @llvm.nvvm.rcp.approx.ftz.d(double %29) #142
  %31 = fneg double %29
  %32 = call double @llvm.fma.f64(double %31, double %30, double noundef 1.000000e+00) #140
  %33 = call double @llvm.fma.f64(double %32, double %32, double %32) #140
  %34 = call double @llvm.fma.f64(double %33, double %30, double %30) #140
  %35 = fmul double %28, %34
  %36 = fadd double %35, %35
  %37 = fmul double %36, %36
  %38 = call double @llvm.fma.f64(double %37, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %39 = call double @llvm.fma.f64(double %38, double %37, double noundef 0x3EF3B2669F02676F) #140
  %40 = call double @llvm.fma.f64(double %39, double %37, double noundef 0x3F1745CBA9AB0956) #140
  %41 = call double @llvm.fma.f64(double %40, double %37, double noundef 0x3F3C71C72D1B5154) #140
  %42 = call double @llvm.fma.f64(double %41, double %37, double noundef 0x3F624924923BE72D) #140
  %43 = call double @llvm.fma.f64(double %42, double %37, double noundef 0x3F8999999999A3C4) #140
  %44 = call double @llvm.fma.f64(double %43, double %37, double noundef 0x3FB5555555555554) #140
  %45 = fsub double %28, %36
  %46 = fmul double %45, 2.000000e+00
  %47 = fneg double %36
  %48 = call double @llvm.fma.f64(double %47, double %28, double %46) #140
  %49 = fmul double %34, %48
  %50 = fmul double %37, %44
  %51 = call double @llvm.fma.f64(double %50, double %36, double %49) #140
  %52 = xor i32 %e.i.1, -2147483648
  %53 = call double @llvm.nvvm.lohi.i2d(i32 %52, i32 noundef 1127219200) #142
  %54 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %55 = fsub double %53, %54
  %56 = call double @llvm.fma.f64(double %55, double noundef 0x3FE62E42FEFA39EF, double %36) #140
  %57 = fneg double %55
  %58 = call double @llvm.fma.f64(double %57, double noundef 0x3FE62E42FEFA39EF, double %56) #140
  %59 = fsub double %58, %36
  %60 = fsub double %51, %59
  %61 = call double @llvm.fma.f64(double %55, double noundef 0x3C7ABC9E3B39803F, double %60) #140
  %62 = fadd double %56, %61
  br label %__nv_log.exit

63:                                               ; preds = %11
  %64 = call double @llvm.fma.f64(double %.01, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %65 = call i32 @llvm.nvvm.d2i.hi(double %.01) #142
  %66 = bitcast i32 %65 to float
  %67 = fcmp oeq float %66, 0.000000e+00
  %q.i.0 = select i1 %67, double 0xFFF0000000000000, double %64
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %63, %27
  %q.i.1 = phi double [ %62, %27 ], [ %q.i.0, %63 ]
  %68 = fadd double %q.i.1, 0x3FE62E42FEFA39EF
  br label %174

69:                                               ; preds = %0
  %70 = call double @llvm.fma.f64(double %a, double %1, double %1) #140
  %71 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %70) #141, !srcloc !83
  %72 = call i32 @llvm.nvvm.d2i.lo(double %71) #139
  %73 = call i32 @llvm.nvvm.d2i.hi(double %71) #139
  %74 = add i32 %73, -1048576
  %75 = call double @llvm.nvvm.lohi.i2d(i32 %72, i32 %74) #139
  %76 = fmul double %70, %71
  %77 = fneg double %76
  %78 = call double @llvm.fma.f64(double %76, double %77, double %70)
  %79 = call double @llvm.fma.f64(double %78, double %75, double %76)
  %80 = fneg double %79
  %81 = call double @llvm.fma.f64(double %71, double %80, double noundef 1.000000e+00)
  %82 = call double @llvm.fma.f64(double %81, double %75, double %75)
  %83 = call double @llvm.fma.f64(double %79, double %80, double %70)
  %84 = call double @llvm.fma.f64(double %83, double %82, double %79)
  %85 = fadd double %1, %84
  %86 = call i32 @llvm.nvvm.d2i.hi(double %85) #139
  %87 = bitcast i32 %86 to float
  %88 = fcmp olt float %87, 0x3FFCAAAAA0000000
  %89 = fcmp ogt float %87, 0xBFFB333320000000
  %or.cond2 = and i1 %88, %89
  br i1 %or.cond2, label %90, label %107

90:                                               ; preds = %69
  %91 = fadd double %85, 2.000000e+00
  %92 = fdiv double %85, %91
  %93 = fneg double %85
  %94 = fmul double %92, %93
  %95 = fadd double %85, %94
  %96 = fmul double %95, %95
  %97 = call double @llvm.fma.f64(double %96, double noundef 0x3EB372FB2FBE14B5, double noundef 0x3ED087FFCEB2DC44)
  %98 = call double @llvm.fma.f64(double %97, double %96, double noundef 0x3EF3B9FF890F468C)
  %99 = call double @llvm.fma.f64(double %98, double %96, double noundef 0x3F17457EFD51BAF8)
  %100 = call double @llvm.fma.f64(double %99, double %96, double noundef 0x3F3C71C8DE3CE825)
  %101 = call double @llvm.fma.f64(double %100, double %96, double noundef 0x3F6249248FA4661F)
  %102 = call double @llvm.fma.f64(double %101, double %96, double noundef 0x3F899999999D70C4)
  %103 = call double @llvm.fma.f64(double %102, double %96, double noundef 0x3FB5555555555462)
  %104 = fmul double %96, %103
  %105 = call double @llvm.fma.f64(double %104, double %95, double %94)
  %106 = fadd double %85, %105
  br label %__nv_log1p.exit

107:                                              ; preds = %69
  %108 = fadd double %85, 1.000000e+00
  %109 = call i32 @llvm.nvvm.d2i.hi(double %108) #139
  %110 = call i32 @llvm.nvvm.d2i.lo(double %108) #139
  %111 = icmp slt i32 %109, 1048576
  br i1 %111, label %112, label %116

112:                                              ; preds = %107
  %113 = fmul double %108, 0x4350000000000000
  %114 = call i32 @llvm.nvvm.d2i.hi(double %113) #139
  %115 = call i32 @llvm.nvvm.d2i.lo(double %113) #139
  br label %116

116:                                              ; preds = %112, %107
  %.0 = phi double [ %113, %112 ], [ %108, %107 ]
  %ihi.i.i.0 = phi i32 [ %114, %112 ], [ %109, %107 ]
  %ilo.i.i.0 = phi i32 [ %115, %112 ], [ %110, %107 ]
  %e.i.i.0 = phi i32 [ -1077, %112 ], [ -1023, %107 ]
  %117 = add i32 %ihi.i.i.0, -1
  %118 = icmp ult i32 %117, 2146435071
  br i1 %118, label %119, label %168

119:                                              ; preds = %116
  %120 = lshr i32 %ihi.i.i.0, 20
  %121 = add nsw i32 %e.i.i.0, %120
  %122 = and i32 %ihi.i.i.0, -2146435073
  %123 = or i32 %122, 1072693248
  %124 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i.0, i32 %123) #139
  %125 = icmp sgt i32 %123, 1073127582
  br i1 %125, label %126, label %132

126:                                              ; preds = %119
  %127 = call i32 @llvm.nvvm.d2i.lo(double %124) #139
  %128 = call i32 @llvm.nvvm.d2i.hi(double %124) #139
  %129 = add i32 %128, -1048576
  %130 = call double @llvm.nvvm.lohi.i2d(i32 %127, i32 %129) #139
  %131 = add nsw i32 %121, 1
  br label %132

132:                                              ; preds = %126, %119
  %m.i.i.0 = phi double [ %130, %126 ], [ %124, %119 ]
  %e.i.i.1 = phi i32 [ %131, %126 ], [ %121, %119 ]
  %133 = fadd double %m.i.i.0, -1.000000e+00
  %134 = fadd double %m.i.i.0, 1.000000e+00
  %135 = call double @llvm.nvvm.rcp.approx.ftz.d(double %134) #139
  %136 = fneg double %134
  %137 = call double @llvm.fma.f64(double %136, double %135, double noundef 1.000000e+00)
  %138 = call double @llvm.fma.f64(double %137, double %137, double %137)
  %139 = call double @llvm.fma.f64(double %138, double %135, double %135)
  %140 = fmul double %133, %139
  %141 = fadd double %140, %140
  %142 = fmul double %141, %141
  %143 = call double @llvm.fma.f64(double %142, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %144 = call double @llvm.fma.f64(double %143, double %142, double noundef 0x3EF3B2669F02676F)
  %145 = call double @llvm.fma.f64(double %144, double %142, double noundef 0x3F1745CBA9AB0956)
  %146 = call double @llvm.fma.f64(double %145, double %142, double noundef 0x3F3C71C72D1B5154)
  %147 = call double @llvm.fma.f64(double %146, double %142, double noundef 0x3F624924923BE72D)
  %148 = call double @llvm.fma.f64(double %147, double %142, double noundef 0x3F8999999999A3C4)
  %149 = call double @llvm.fma.f64(double %148, double %142, double noundef 0x3FB5555555555554)
  %150 = fsub double %133, %141
  %151 = fmul double %150, 2.000000e+00
  %152 = fneg double %141
  %153 = call double @llvm.fma.f64(double %152, double %133, double %151)
  %154 = fmul double %139, %153
  %155 = fmul double %142, %149
  %156 = call double @llvm.fma.f64(double %155, double %141, double %154)
  %157 = xor i32 %e.i.i.1, -2147483648
  %158 = call double @llvm.nvvm.lohi.i2d(i32 %157, i32 noundef 1127219200) #139
  %159 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %160 = fsub double %158, %159
  %161 = call double @llvm.fma.f64(double %160, double noundef 0x3FE62E42FEFA39EF, double %141)
  %162 = fneg double %160
  %163 = call double @llvm.fma.f64(double %162, double noundef 0x3FE62E42FEFA39EF, double %161)
  %164 = fsub double %163, %141
  %165 = fsub double %156, %164
  %166 = call double @llvm.fma.f64(double %160, double noundef 0x3C7ABC9E3B39803F, double %165)
  %167 = fadd double %161, %166
  br label %__nv_log1p.exit

168:                                              ; preds = %116
  %169 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %170 = call i32 @llvm.nvvm.d2i.hi(double %.0) #139
  %171 = bitcast i32 %170 to float
  %172 = fcmp oeq float %171, 0.000000e+00
  %q.i.i.0 = select i1 %172, double 0xFFF0000000000000, double %169
  br label %__nv_log1p.exit

__nv_log1p.exit:                                  ; preds = %132, %168, %90
  %t.i.0 = phi double [ %106, %90 ], [ %167, %132 ], [ %q.i.i.0, %168 ]
  %173 = icmp eq i32 %2, 0
  %spec.select = select i1 %173, double %1, double %t.i.0
  br label %174

174:                                              ; preds = %__nv_log1p.exit, %__nv_log.exit
  %r.1 = phi double [ %68, %__nv_log.exit ], [ %spec.select, %__nv_log1p.exit ]
  ret double %r.1
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_asinh(double %a) local_unnamed_addr #92 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %3 = and i32 %1, 2147483647
  %4 = call double @llvm.nvvm.lohi.i2d(i32 %2, i32 %3) #142
  %5 = icmp ugt i32 %3, 1138753535
  br i1 %5, label %6, label %72

6:                                                ; preds = %0
  %7 = call i32 @llvm.nvvm.d2i.hi(double %4) #142
  %8 = call i32 @llvm.nvvm.d2i.lo(double %4) #142
  %9 = icmp slt i32 %7, 1048576
  br i1 %9, label %10, label %14

10:                                               ; preds = %6
  %11 = fmul double %4, 0x4350000000000000
  %12 = call i32 @llvm.nvvm.d2i.hi(double %11) #142
  %13 = call i32 @llvm.nvvm.d2i.lo(double %11) #142
  br label %14

14:                                               ; preds = %10, %6
  %.01 = phi double [ %11, %10 ], [ %4, %6 ]
  %ihi.i.0 = phi i32 [ %12, %10 ], [ %7, %6 ]
  %ilo.i.0 = phi i32 [ %13, %10 ], [ %8, %6 ]
  %e.i.0 = phi i32 [ -1077, %10 ], [ -1023, %6 ]
  %15 = add i32 %ihi.i.0, -1
  %16 = icmp ult i32 %15, 2146435071
  br i1 %16, label %17, label %66

17:                                               ; preds = %14
  %18 = lshr i32 %ihi.i.0, 20
  %19 = add nsw i32 %e.i.0, %18
  %20 = and i32 %ihi.i.0, -2146435073
  %21 = or i32 %20, 1072693248
  %22 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %21) #142
  %23 = icmp sgt i32 %21, 1073127582
  br i1 %23, label %24, label %30

24:                                               ; preds = %17
  %25 = call i32 @llvm.nvvm.d2i.lo(double %22) #142
  %26 = call i32 @llvm.nvvm.d2i.hi(double %22) #142
  %27 = add i32 %26, -1048576
  %28 = call double @llvm.nvvm.lohi.i2d(i32 %25, i32 %27) #142
  %29 = add nsw i32 %19, 1
  br label %30

30:                                               ; preds = %24, %17
  %m.i.0 = phi double [ %28, %24 ], [ %22, %17 ]
  %e.i.1 = phi i32 [ %29, %24 ], [ %19, %17 ]
  %31 = fadd double %m.i.0, -1.000000e+00
  %32 = fadd double %m.i.0, 1.000000e+00
  %33 = call double @llvm.nvvm.rcp.approx.ftz.d(double %32) #142
  %34 = fneg double %32
  %35 = call double @llvm.fma.f64(double %34, double %33, double noundef 1.000000e+00) #140
  %36 = call double @llvm.fma.f64(double %35, double %35, double %35) #140
  %37 = call double @llvm.fma.f64(double %36, double %33, double %33) #140
  %38 = fmul double %31, %37
  %39 = fadd double %38, %38
  %40 = fmul double %39, %39
  %41 = call double @llvm.fma.f64(double %40, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %42 = call double @llvm.fma.f64(double %41, double %40, double noundef 0x3EF3B2669F02676F) #140
  %43 = call double @llvm.fma.f64(double %42, double %40, double noundef 0x3F1745CBA9AB0956) #140
  %44 = call double @llvm.fma.f64(double %43, double %40, double noundef 0x3F3C71C72D1B5154) #140
  %45 = call double @llvm.fma.f64(double %44, double %40, double noundef 0x3F624924923BE72D) #140
  %46 = call double @llvm.fma.f64(double %45, double %40, double noundef 0x3F8999999999A3C4) #140
  %47 = call double @llvm.fma.f64(double %46, double %40, double noundef 0x3FB5555555555554) #140
  %48 = fsub double %31, %39
  %49 = fmul double %48, 2.000000e+00
  %50 = fneg double %39
  %51 = call double @llvm.fma.f64(double %50, double %31, double %49) #140
  %52 = fmul double %37, %51
  %53 = fmul double %40, %47
  %54 = call double @llvm.fma.f64(double %53, double %39, double %52) #140
  %55 = xor i32 %e.i.1, -2147483648
  %56 = call double @llvm.nvvm.lohi.i2d(i32 %55, i32 noundef 1127219200) #142
  %57 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %58 = fsub double %56, %57
  %59 = call double @llvm.fma.f64(double %58, double noundef 0x3FE62E42FEFA39EF, double %39) #140
  %60 = fneg double %58
  %61 = call double @llvm.fma.f64(double %60, double noundef 0x3FE62E42FEFA39EF, double %59) #140
  %62 = fsub double %61, %39
  %63 = fsub double %54, %62
  %64 = call double @llvm.fma.f64(double %58, double noundef 0x3C7ABC9E3B39803F, double %63) #140
  %65 = fadd double %59, %64
  br label %__nv_log.exit

66:                                               ; preds = %14
  %67 = call double @llvm.fma.f64(double %.01, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %68 = call i32 @llvm.nvvm.d2i.hi(double %.01) #142
  %69 = bitcast i32 %68 to float
  %70 = fcmp oeq float %69, 0.000000e+00
  %q.i.0 = select i1 %70, double 0xFFF0000000000000, double %67
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %66, %30
  %q.i.1 = phi double [ %65, %30 ], [ %q.i.0, %66 ]
  %71 = fadd double %q.i.1, 0x3FE62E42FEFA39EF
  br label %__nv_log1p.exit

72:                                               ; preds = %0
  %73 = fmul double %a, %a
  %74 = call double @llvm.fma.f64(double %a, double %a, double noundef 1.000000e+00) #140
  %75 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %74) #141, !srcloc !83
  %76 = call i32 @llvm.nvvm.d2i.lo(double %75) #139
  %77 = call i32 @llvm.nvvm.d2i.hi(double %75) #139
  %78 = add i32 %77, -1048576
  %79 = call double @llvm.nvvm.lohi.i2d(i32 %76, i32 %78) #139
  %80 = fmul double %74, %75
  %81 = fneg double %80
  %82 = call double @llvm.fma.f64(double %80, double %81, double %74)
  %83 = call double @llvm.fma.f64(double %82, double %79, double %80)
  %84 = fneg double %83
  %85 = call double @llvm.fma.f64(double %75, double %84, double noundef 1.000000e+00)
  %86 = call double @llvm.fma.f64(double %85, double %79, double %79)
  %87 = call double @llvm.fma.f64(double %83, double %84, double %74)
  %88 = call double @llvm.fma.f64(double %87, double %86, double %83)
  %89 = fadd double %88, 1.000000e+00
  %90 = fdiv double %73, %89
  %91 = fadd double %4, %90
  %92 = call i32 @llvm.nvvm.d2i.hi(double %91) #139
  %93 = bitcast i32 %92 to float
  %94 = fcmp olt float %93, 0x3FFCAAAAA0000000
  %95 = fcmp ogt float %93, 0xBFFB333320000000
  %or.cond2 = and i1 %94, %95
  br i1 %or.cond2, label %96, label %113

96:                                               ; preds = %72
  %97 = fadd double %91, 2.000000e+00
  %98 = fdiv double %91, %97
  %99 = fneg double %91
  %100 = fmul double %98, %99
  %101 = fadd double %91, %100
  %102 = fmul double %101, %101
  %103 = call double @llvm.fma.f64(double %102, double noundef 0x3EB372FB2FBE14B5, double noundef 0x3ED087FFCEB2DC44)
  %104 = call double @llvm.fma.f64(double %103, double %102, double noundef 0x3EF3B9FF890F468C)
  %105 = call double @llvm.fma.f64(double %104, double %102, double noundef 0x3F17457EFD51BAF8)
  %106 = call double @llvm.fma.f64(double %105, double %102, double noundef 0x3F3C71C8DE3CE825)
  %107 = call double @llvm.fma.f64(double %106, double %102, double noundef 0x3F6249248FA4661F)
  %108 = call double @llvm.fma.f64(double %107, double %102, double noundef 0x3F899999999D70C4)
  %109 = call double @llvm.fma.f64(double %108, double %102, double noundef 0x3FB5555555555462)
  %110 = fmul double %102, %109
  %111 = call double @llvm.fma.f64(double %110, double %101, double %100)
  %112 = fadd double %91, %111
  br label %__nv_log1p.exit

113:                                              ; preds = %72
  %114 = fadd double %91, 1.000000e+00
  %115 = call i32 @llvm.nvvm.d2i.hi(double %114) #139
  %116 = call i32 @llvm.nvvm.d2i.lo(double %114) #139
  %117 = icmp slt i32 %115, 1048576
  br i1 %117, label %118, label %122

118:                                              ; preds = %113
  %119 = fmul double %114, 0x4350000000000000
  %120 = call i32 @llvm.nvvm.d2i.hi(double %119) #139
  %121 = call i32 @llvm.nvvm.d2i.lo(double %119) #139
  br label %122

122:                                              ; preds = %118, %113
  %.0 = phi double [ %119, %118 ], [ %114, %113 ]
  %ihi.i.i.0 = phi i32 [ %120, %118 ], [ %115, %113 ]
  %ilo.i.i.0 = phi i32 [ %121, %118 ], [ %116, %113 ]
  %e.i.i.0 = phi i32 [ -1077, %118 ], [ -1023, %113 ]
  %123 = add i32 %ihi.i.i.0, -1
  %124 = icmp ult i32 %123, 2146435071
  br i1 %124, label %125, label %174

125:                                              ; preds = %122
  %126 = lshr i32 %ihi.i.i.0, 20
  %127 = add nsw i32 %e.i.i.0, %126
  %128 = and i32 %ihi.i.i.0, -2146435073
  %129 = or i32 %128, 1072693248
  %130 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i.0, i32 %129) #139
  %131 = icmp sgt i32 %129, 1073127582
  br i1 %131, label %132, label %138

132:                                              ; preds = %125
  %133 = call i32 @llvm.nvvm.d2i.lo(double %130) #139
  %134 = call i32 @llvm.nvvm.d2i.hi(double %130) #139
  %135 = add i32 %134, -1048576
  %136 = call double @llvm.nvvm.lohi.i2d(i32 %133, i32 %135) #139
  %137 = add nsw i32 %127, 1
  br label %138

138:                                              ; preds = %132, %125
  %m.i.i.0 = phi double [ %136, %132 ], [ %130, %125 ]
  %e.i.i.1 = phi i32 [ %137, %132 ], [ %127, %125 ]
  %139 = fadd double %m.i.i.0, -1.000000e+00
  %140 = fadd double %m.i.i.0, 1.000000e+00
  %141 = call double @llvm.nvvm.rcp.approx.ftz.d(double %140) #139
  %142 = fneg double %140
  %143 = call double @llvm.fma.f64(double %142, double %141, double noundef 1.000000e+00)
  %144 = call double @llvm.fma.f64(double %143, double %143, double %143)
  %145 = call double @llvm.fma.f64(double %144, double %141, double %141)
  %146 = fmul double %139, %145
  %147 = fadd double %146, %146
  %148 = fmul double %147, %147
  %149 = call double @llvm.fma.f64(double %148, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %150 = call double @llvm.fma.f64(double %149, double %148, double noundef 0x3EF3B2669F02676F)
  %151 = call double @llvm.fma.f64(double %150, double %148, double noundef 0x3F1745CBA9AB0956)
  %152 = call double @llvm.fma.f64(double %151, double %148, double noundef 0x3F3C71C72D1B5154)
  %153 = call double @llvm.fma.f64(double %152, double %148, double noundef 0x3F624924923BE72D)
  %154 = call double @llvm.fma.f64(double %153, double %148, double noundef 0x3F8999999999A3C4)
  %155 = call double @llvm.fma.f64(double %154, double %148, double noundef 0x3FB5555555555554)
  %156 = fsub double %139, %147
  %157 = fmul double %156, 2.000000e+00
  %158 = fneg double %147
  %159 = call double @llvm.fma.f64(double %158, double %139, double %157)
  %160 = fmul double %145, %159
  %161 = fmul double %148, %155
  %162 = call double @llvm.fma.f64(double %161, double %147, double %160)
  %163 = xor i32 %e.i.i.1, -2147483648
  %164 = call double @llvm.nvvm.lohi.i2d(i32 %163, i32 noundef 1127219200) #139
  %165 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %166 = fsub double %164, %165
  %167 = call double @llvm.fma.f64(double %166, double noundef 0x3FE62E42FEFA39EF, double %147)
  %168 = fneg double %166
  %169 = call double @llvm.fma.f64(double %168, double noundef 0x3FE62E42FEFA39EF, double %167)
  %170 = fsub double %169, %147
  %171 = fsub double %162, %170
  %172 = call double @llvm.fma.f64(double %166, double noundef 0x3C7ABC9E3B39803F, double %171)
  %173 = fadd double %167, %172
  br label %__nv_log1p.exit

174:                                              ; preds = %122
  %175 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %176 = call i32 @llvm.nvvm.d2i.hi(double %.0) #139
  %177 = bitcast i32 %176 to float
  %178 = fcmp oeq float %177, 0.000000e+00
  %q.i.i.0 = select i1 %178, double 0xFFF0000000000000, double %175
  br label %__nv_log1p.exit

__nv_log1p.exit:                                  ; preds = %96, %174, %138, %__nv_log.exit
  %t.0 = phi double [ %71, %__nv_log.exit ], [ %112, %96 ], [ %173, %138 ], [ %q.i.i.0, %174 ]
  %179 = call i32 @llvm.nvvm.d2i.lo(double %t.0) #139
  %180 = call i32 @llvm.nvvm.d2i.hi(double %t.0) #139
  %181 = and i32 %1, -2147483648
  %182 = or i32 %180, %181
  %183 = call double @llvm.nvvm.lohi.i2d(i32 %179, i32 %182) #139
  ret double %183
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_atanh(double %a) local_unnamed_addr #93 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fmul double %1, 2.000000e+00
  %3 = fsub double 1.000000e+00, %1
  %4 = fdiv double %2, %3
  %5 = call i32 @llvm.nvvm.d2i.hi(double %4) #142
  %6 = bitcast i32 %5 to float
  %7 = fcmp olt float %6, 0x3FFCAAAAA0000000
  %8 = fcmp ogt float %6, 0xBFFB333320000000
  %or.cond = and i1 %7, %8
  br i1 %or.cond, label %9, label %26

9:                                                ; preds = %0
  %10 = fadd double %4, 2.000000e+00
  %11 = fdiv double %4, %10
  %12 = fneg double %4
  %13 = fmul double %11, %12
  %14 = fadd double %4, %13
  %15 = fmul double %14, %14
  %16 = call double @llvm.fma.f64(double %15, double noundef 0x3EB372FB2FBE14B5, double noundef 0x3ED087FFCEB2DC44) #140
  %17 = call double @llvm.fma.f64(double %16, double %15, double noundef 0x3EF3B9FF890F468C) #140
  %18 = call double @llvm.fma.f64(double %17, double %15, double noundef 0x3F17457EFD51BAF8) #140
  %19 = call double @llvm.fma.f64(double %18, double %15, double noundef 0x3F3C71C8DE3CE825) #140
  %20 = call double @llvm.fma.f64(double %19, double %15, double noundef 0x3F6249248FA4661F) #140
  %21 = call double @llvm.fma.f64(double %20, double %15, double noundef 0x3F899999999D70C4) #140
  %22 = call double @llvm.fma.f64(double %21, double %15, double noundef 0x3FB5555555555462) #140
  %23 = fmul double %15, %22
  %24 = call double @llvm.fma.f64(double %23, double %14, double %13) #140
  %25 = fadd double %4, %24
  br label %__nv_log1p.exit

26:                                               ; preds = %0
  %27 = fadd double %4, 1.000000e+00
  %28 = call i32 @llvm.nvvm.d2i.hi(double %27) #142
  %29 = call i32 @llvm.nvvm.d2i.lo(double %27) #142
  %30 = icmp slt i32 %28, 1048576
  br i1 %30, label %31, label %35

31:                                               ; preds = %26
  %32 = fmul double %27, 0x4350000000000000
  %33 = call i32 @llvm.nvvm.d2i.hi(double %32) #142
  %34 = call i32 @llvm.nvvm.d2i.lo(double %32) #142
  br label %35

35:                                               ; preds = %31, %26
  %.0 = phi double [ %32, %31 ], [ %27, %26 ]
  %ihi.i.i.0 = phi i32 [ %33, %31 ], [ %28, %26 ]
  %ilo.i.i.0 = phi i32 [ %34, %31 ], [ %29, %26 ]
  %e.i.i.0 = phi i32 [ -1077, %31 ], [ -1023, %26 ]
  %36 = add i32 %ihi.i.i.0, -1
  %37 = icmp ult i32 %36, 2146435071
  br i1 %37, label %38, label %87

38:                                               ; preds = %35
  %39 = lshr i32 %ihi.i.i.0, 20
  %40 = add nsw i32 %e.i.i.0, %39
  %41 = and i32 %ihi.i.i.0, -2146435073
  %42 = or i32 %41, 1072693248
  %43 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i.0, i32 %42) #142
  %44 = icmp sgt i32 %42, 1073127582
  br i1 %44, label %45, label %51

45:                                               ; preds = %38
  %46 = call i32 @llvm.nvvm.d2i.lo(double %43) #142
  %47 = call i32 @llvm.nvvm.d2i.hi(double %43) #142
  %48 = add i32 %47, -1048576
  %49 = call double @llvm.nvvm.lohi.i2d(i32 %46, i32 %48) #142
  %50 = add nsw i32 %40, 1
  br label %51

51:                                               ; preds = %45, %38
  %m.i.i.0 = phi double [ %49, %45 ], [ %43, %38 ]
  %e.i.i.1 = phi i32 [ %50, %45 ], [ %40, %38 ]
  %52 = fadd double %m.i.i.0, -1.000000e+00
  %53 = fadd double %m.i.i.0, 1.000000e+00
  %54 = call double @llvm.nvvm.rcp.approx.ftz.d(double %53) #142
  %55 = fneg double %53
  %56 = call double @llvm.fma.f64(double %55, double %54, double noundef 1.000000e+00) #140
  %57 = call double @llvm.fma.f64(double %56, double %56, double %56) #140
  %58 = call double @llvm.fma.f64(double %57, double %54, double %54) #140
  %59 = fmul double %52, %58
  %60 = fadd double %59, %59
  %61 = fmul double %60, %60
  %62 = call double @llvm.fma.f64(double %61, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %63 = call double @llvm.fma.f64(double %62, double %61, double noundef 0x3EF3B2669F02676F) #140
  %64 = call double @llvm.fma.f64(double %63, double %61, double noundef 0x3F1745CBA9AB0956) #140
  %65 = call double @llvm.fma.f64(double %64, double %61, double noundef 0x3F3C71C72D1B5154) #140
  %66 = call double @llvm.fma.f64(double %65, double %61, double noundef 0x3F624924923BE72D) #140
  %67 = call double @llvm.fma.f64(double %66, double %61, double noundef 0x3F8999999999A3C4) #140
  %68 = call double @llvm.fma.f64(double %67, double %61, double noundef 0x3FB5555555555554) #140
  %69 = fsub double %52, %60
  %70 = fmul double %69, 2.000000e+00
  %71 = fneg double %60
  %72 = call double @llvm.fma.f64(double %71, double %52, double %70) #140
  %73 = fmul double %58, %72
  %74 = fmul double %61, %68
  %75 = call double @llvm.fma.f64(double %74, double %60, double %73) #140
  %76 = xor i32 %e.i.i.1, -2147483648
  %77 = call double @llvm.nvvm.lohi.i2d(i32 %76, i32 noundef 1127219200) #142
  %78 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %79 = fsub double %77, %78
  %80 = call double @llvm.fma.f64(double %79, double noundef 0x3FE62E42FEFA39EF, double %60) #140
  %81 = fneg double %79
  %82 = call double @llvm.fma.f64(double %81, double noundef 0x3FE62E42FEFA39EF, double %80) #140
  %83 = fsub double %82, %60
  %84 = fsub double %75, %83
  %85 = call double @llvm.fma.f64(double %79, double noundef 0x3C7ABC9E3B39803F, double %84) #140
  %86 = fadd double %80, %85
  br label %__nv_log1p.exit

87:                                               ; preds = %35
  %88 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %89 = call i32 @llvm.nvvm.d2i.hi(double %.0) #142
  %90 = bitcast i32 %89 to float
  %91 = fcmp oeq float %90, 0.000000e+00
  %q.i.i.0 = select i1 %91, double 0xFFF0000000000000, double %88
  br label %__nv_log1p.exit

__nv_log1p.exit:                                  ; preds = %51, %87, %9
  %t.i.0 = phi double [ %25, %9 ], [ %86, %51 ], [ %q.i.i.0, %87 ]
  %92 = fmul double %t.i.0, 5.000000e-01
  %93 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %94 = call i32 @llvm.nvvm.d2i.lo(double %92) #142
  %95 = call i32 @llvm.nvvm.d2i.hi(double %92) #142
  %96 = and i32 %93, -2147483648
  %97 = or i32 %95, %96
  %98 = call double @llvm.nvvm.lohi.i2d(i32 %94, i32 %97) #142
  ret double %98
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_hypot(double %a, double %b) local_unnamed_addr #94 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = call double @llvm.fabs.f64(double %b) #140
  %3 = bitcast double %2 to i64
  %4 = bitcast double %1 to i64
  %min.cond.not = icmp ugt i64 %3, %4
  %min = select i1 %min.cond.not, i64 %4, i64 %3
  %5 = bitcast i64 %min to double
  %max = select i1 %min.cond.not, i64 %3, i64 %4
  %6 = bitcast i64 %max to double
  %7 = call i32 @llvm.nvvm.d2i.hi(double %6) #142
  %8 = and i32 %7, -4194304
  %9 = sub i32 2144337920, %8
  %10 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %9) #142
  %11 = fmul double %10, %5
  %12 = fmul double %10, %6
  %13 = fmul double %11, %11
  %14 = call double @llvm.fma.f64(double %12, double %12, double %13) #140
  %15 = call double @llvm.minnum.f64(double %14, double noundef 0x7FEFFFFFFFFFFFFF) #140
  %16 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %15) #141, !srcloc !84
  %17 = fneg double %16
  %18 = fmul double %16, %17
  %19 = call double @llvm.fma.f64(double %15, double %18, double noundef 1.000000e+00)
  %20 = call double @llvm.fma.f64(double %19, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %21 = fmul double %16, %19
  %22 = call double @llvm.fma.f64(double %20, double %21, double %16)
  %23 = fmul double %14, %22
  %24 = or i32 %8, 1048576
  %25 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %24) #139
  %26 = fmul double %25, %23
  %27 = fcmp oeq double %5, 0.000000e+00
  %res.0 = select i1 %27, double %6, double %26
  %28 = call i32 @llvm.nvvm.d2i.hi(double %5) #139
  %29 = call i32 @llvm.nvvm.d2i.hi(double noundef 0x7FF0000000000000) #139
  %.not = icmp ult i32 %28, %29
  %res.1 = select i1 %.not, double %res.0, double %5
  ret double %res.1
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_rhypot(double %a, double %b) local_unnamed_addr #95 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = call double @llvm.fabs.f64(double %b) #140
  %3 = bitcast double %2 to i64
  %4 = bitcast double %1 to i64
  %min.cond.not = icmp ugt i64 %3, %4
  %min = select i1 %min.cond.not, i64 %4, i64 %3
  %5 = bitcast i64 %min to double
  %max = select i1 %min.cond.not, i64 %3, i64 %4
  %6 = bitcast i64 %max to double
  %7 = call i32 @llvm.nvvm.d2i.hi(double %6) #142
  %8 = and i32 %7, -4194304
  %9 = sub i32 2144337920, %8
  %10 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %9) #142
  %11 = fmul double %10, %5
  %12 = fmul double %10, %6
  %13 = fmul double %11, %11
  %14 = call double @llvm.fma.f64(double %12, double %12, double %13) #140
  %15 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %14) #141, !srcloc !84
  %16 = fneg double %15
  %17 = fmul double %15, %16
  %18 = call double @llvm.fma.f64(double %14, double %17, double noundef 1.000000e+00)
  %19 = call double @llvm.fma.f64(double %18, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %20 = fmul double %15, %18
  %21 = call double @llvm.fma.f64(double %19, double %20, double %15)
  %22 = fmul double %10, %21
  %23 = call i32 @llvm.nvvm.d2i.hi(double %22) #139
  %24 = call i32 @llvm.nvvm.d2i.hi(double noundef 0x7FF0000000000000) #139
  %.not = icmp ult i32 %23, %24
  br i1 %.not, label %33, label %25

25:                                               ; preds = %0
  %26 = call double @llvm.fabs.f64(double %6)
  %27 = fcmp ugt double %26, 0x7FF0000000000000
  %28 = fadd double %6, %5
  %29 = fcmp une double %6, 0x7FF0000000000000
  %30 = select i1 %29, i32 %24, i32 0
  %31 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %30) #139
  %res.0 = select i1 %27, double %28, double %31
  %32 = fcmp oeq double %5, 0x7FF0000000000000
  %res.1 = select i1 %32, double 0.000000e+00, double %res.0
  br label %33

33:                                               ; preds = %25, %0
  %res.2 = phi double [ %res.1, %25 ], [ %22, %0 ]
  ret double %res.2
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_norm3d(double %a, double %b, double %c) local_unnamed_addr #96 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = call double @llvm.fabs.f64(double %b) #140
  %3 = call double @llvm.fabs.f64(double %c) #140
  %4 = fadd double %1, %2
  %5 = fadd double %4, %3
  %6 = call double @llvm.minnum.f64(double %1, double %2) #140
  %7 = call double @llvm.maxnum.f64(double %1, double %2) #140
  %8 = call double @llvm.minnum.f64(double %7, double %3) #140
  %9 = call double @llvm.maxnum.f64(double %7, double %3) #140
  %10 = call i32 @llvm.nvvm.d2i.hi(double %9) #142
  %11 = and i32 %10, -4194304
  %12 = sub i32 2144337920, %11
  %13 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %12) #142
  %14 = fmul double %8, %13
  %15 = fmul double %6, %13
  %16 = fmul double %9, %13
  %17 = fmul double %14, %14
  %18 = call double @llvm.fma.f64(double %15, double %15, double %17) #140
  %19 = call double @llvm.fma.f64(double %16, double %16, double %18) #140
  %20 = call double @llvm.minnum.f64(double %19, double noundef 0x7FEFFFFFFFFFFFFF) #140
  %21 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %20) #141, !srcloc !84
  %22 = fneg double %21
  %23 = fmul double %21, %22
  %24 = call double @llvm.fma.f64(double %20, double %23, double noundef 1.000000e+00)
  %25 = call double @llvm.fma.f64(double %24, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %26 = fmul double %21, %24
  %27 = call double @llvm.fma.f64(double %25, double %26, double %21)
  %28 = fmul double %19, %27
  %29 = or i32 %11, 1048576
  %30 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %29) #139
  %31 = fmul double %30, %28
  %32 = fcmp ogt double %5, %9
  %res.0 = select i1 %32, double %31, double %5
  %33 = call i32 @llvm.nvvm.d2i.hi(double noundef 0x7FF0000000000000) #139
  %.not = icmp ult i32 %10, %33
  %res.1 = select i1 %.not, double %res.0, double %9
  ret double %res.1
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_rnorm3d(double %a, double %b, double %c) local_unnamed_addr #97 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = call double @llvm.fabs.f64(double %b) #140
  %3 = call double @llvm.fabs.f64(double %c) #140
  %4 = fadd double %1, %2
  %5 = fadd double %4, %3
  %6 = call double @llvm.minnum.f64(double %1, double %2) #140
  %7 = call double @llvm.maxnum.f64(double %1, double %2) #140
  %8 = call double @llvm.minnum.f64(double %7, double %3) #140
  %9 = call double @llvm.maxnum.f64(double %7, double %3) #140
  %10 = call i32 @llvm.nvvm.d2i.hi(double %9) #142
  %11 = and i32 %10, -4194304
  %12 = sub i32 2144337920, %11
  %13 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %12) #142
  %14 = fmul double %8, %13
  %15 = fmul double %6, %13
  %16 = fmul double %9, %13
  %17 = fmul double %14, %14
  %18 = call double @llvm.fma.f64(double %15, double %15, double %17) #140
  %19 = call double @llvm.fma.f64(double %16, double %16, double %18) #140
  %20 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %19) #141, !srcloc !84
  %21 = fneg double %20
  %22 = fmul double %20, %21
  %23 = call double @llvm.fma.f64(double %19, double %22, double noundef 1.000000e+00)
  %24 = call double @llvm.fma.f64(double %23, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %25 = fmul double %20, %23
  %26 = call double @llvm.fma.f64(double %24, double %25, double %20)
  %27 = fmul double %13, %26
  %28 = fcmp ogt double %5, 0.000000e+00
  %29 = fadd double %5, 0x7FF0000000000000
  %res.0 = select i1 %28, double %27, double %29
  %30 = fcmp oeq double %9, 0x7FF0000000000000
  %res.1 = select i1 %30, double 0.000000e+00, double %res.0
  ret double %res.1
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_norm4d(double %a, double %b, double %c, double %d) local_unnamed_addr #98 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = call double @llvm.fabs.f64(double %b) #140
  %3 = call double @llvm.fabs.f64(double %c) #140
  %4 = call double @llvm.fabs.f64(double %d) #140
  %5 = fadd double %1, %2
  %6 = fadd double %5, %3
  %7 = fadd double %6, %4
  %8 = call double @llvm.minnum.f64(double %1, double %2) #140
  %9 = call double @llvm.maxnum.f64(double %1, double %2) #140
  %10 = call double @llvm.minnum.f64(double %9, double %3) #140
  %11 = call double @llvm.maxnum.f64(double %9, double %3) #140
  %12 = call double @llvm.minnum.f64(double %11, double %4) #140
  %13 = call double @llvm.maxnum.f64(double %11, double %4) #140
  %14 = call i32 @llvm.nvvm.d2i.hi(double %13) #142
  %15 = and i32 %14, -4194304
  %16 = sub i32 2144337920, %15
  %17 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %16) #142
  %18 = fmul double %12, %17
  %19 = fmul double %10, %17
  %20 = fmul double %8, %17
  %21 = fmul double %13, %17
  %22 = fmul double %18, %18
  %23 = call double @llvm.fma.f64(double %19, double %19, double %22) #140
  %24 = call double @llvm.fma.f64(double %20, double %20, double %23) #140
  %25 = call double @llvm.fma.f64(double %21, double %21, double %24) #140
  %26 = call double @llvm.minnum.f64(double %25, double noundef 0x7FEFFFFFFFFFFFFF) #140
  %27 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %26) #141, !srcloc !84
  %28 = fneg double %27
  %29 = fmul double %27, %28
  %30 = call double @llvm.fma.f64(double %26, double %29, double noundef 1.000000e+00)
  %31 = call double @llvm.fma.f64(double %30, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %32 = fmul double %27, %30
  %33 = call double @llvm.fma.f64(double %31, double %32, double %27)
  %34 = fmul double %25, %33
  %35 = or i32 %15, 1048576
  %36 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %35) #139
  %37 = fmul double %36, %34
  %38 = fcmp ogt double %7, %13
  %res.0 = select i1 %38, double %37, double %7
  %39 = call i32 @llvm.nvvm.d2i.hi(double noundef 0x7FF0000000000000) #139
  %.not = icmp ult i32 %14, %39
  %res.1 = select i1 %.not, double %res.0, double %13
  ret double %res.1
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_rnorm4d(double %a, double %b, double %c, double %d) local_unnamed_addr #99 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = call double @llvm.fabs.f64(double %b) #140
  %3 = call double @llvm.fabs.f64(double %c) #140
  %4 = call double @llvm.fabs.f64(double %d) #140
  %5 = fadd double %1, %2
  %6 = fadd double %5, %3
  %7 = fadd double %6, %4
  %8 = call double @llvm.minnum.f64(double %1, double %2) #140
  %9 = call double @llvm.maxnum.f64(double %1, double %2) #140
  %10 = call double @llvm.minnum.f64(double %9, double %3) #140
  %11 = call double @llvm.maxnum.f64(double %9, double %3) #140
  %12 = call double @llvm.minnum.f64(double %11, double %4) #140
  %13 = call double @llvm.maxnum.f64(double %11, double %4) #140
  %14 = call i32 @llvm.nvvm.d2i.hi(double %13) #142
  %15 = and i32 %14, -4194304
  %16 = sub i32 2144337920, %15
  %17 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %16) #142
  %18 = fmul double %12, %17
  %19 = fmul double %10, %17
  %20 = fmul double %8, %17
  %21 = fmul double %13, %17
  %22 = fmul double %18, %18
  %23 = call double @llvm.fma.f64(double %19, double %19, double %22) #140
  %24 = call double @llvm.fma.f64(double %20, double %20, double %23) #140
  %25 = call double @llvm.fma.f64(double %21, double %21, double %24) #140
  %26 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %25) #141, !srcloc !84
  %27 = fneg double %26
  %28 = fmul double %26, %27
  %29 = call double @llvm.fma.f64(double %25, double %28, double noundef 1.000000e+00)
  %30 = call double @llvm.fma.f64(double %29, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %31 = fmul double %26, %29
  %32 = call double @llvm.fma.f64(double %30, double %31, double %26)
  %33 = fmul double %17, %32
  %34 = fcmp ogt double %7, 0.000000e+00
  %35 = fadd double %7, 0x7FF0000000000000
  %res.0 = select i1 %34, double %33, double %35
  %36 = fcmp oeq double %13, 0x7FF0000000000000
  %res.1 = select i1 %36, double 0.000000e+00, double %res.0
  ret double %res.1
}

; Function Attrs: argmemonly noinline nosync nounwind readonly
define dso_local double @__nv_norm(i32 %dim, double* nocapture nofree noundef nonnull readonly align 8 dereferenceable(8) %t) local_unnamed_addr #100 {
  %1 = load double, double* %t, align 8
  %2 = call double @llvm.fabs.f64(double %1)
  %3 = icmp sgt i32 %dim, 1
  br i1 %3, label %.lr.ph7, label %._crit_edge8

.lr.ph7:                                          ; preds = %0
  %4 = add nsw i32 %dim, -1
  %5 = add nsw i32 %dim, -2
  %xtraiter12 = and i32 %4, 3
  %6 = icmp ult i32 %5, 3
  br i1 %6, label %._crit_edge8.unr-lcssa, label %.lr.ph7.new

.lr.ph7.new:                                      ; preds = %.lr.ph7
  %unroll_iter15 = and i32 %4, -4
  br label %7

7:                                                ; preds = %7, %.lr.ph7.new
  %i.05 = phi i32 [ 1, %.lr.ph7.new ], [ %31, %7 ]
  %largest.04 = phi double [ %2, %.lr.ph7.new ], [ %30, %7 ]
  %niter16 = phi i32 [ 0, %.lr.ph7.new ], [ %niter16.next.3, %7 ]
  %8 = zext i32 %i.05 to i64
  %9 = getelementptr inbounds double, double* %t, i64 %8
  %10 = load double, double* %9, align 8
  %11 = call double @llvm.fabs.f64(double %10)
  %12 = call double @llvm.maxnum.f64(double %11, double %largest.04)
  %13 = add nuw nsw i32 %i.05, 1
  %14 = zext i32 %13 to i64
  %15 = getelementptr inbounds double, double* %t, i64 %14
  %16 = load double, double* %15, align 8
  %17 = call double @llvm.fabs.f64(double %16)
  %18 = call double @llvm.maxnum.f64(double %17, double %12)
  %19 = add nuw nsw i32 %i.05, 2
  %20 = zext i32 %19 to i64
  %21 = getelementptr inbounds double, double* %t, i64 %20
  %22 = load double, double* %21, align 8
  %23 = call double @llvm.fabs.f64(double %22)
  %24 = call double @llvm.maxnum.f64(double %23, double %18)
  %25 = add nuw nsw i32 %i.05, 3
  %26 = zext i32 %25 to i64
  %27 = getelementptr inbounds double, double* %t, i64 %26
  %28 = load double, double* %27, align 8
  %29 = call double @llvm.fabs.f64(double %28)
  %30 = call double @llvm.maxnum.f64(double %29, double %24)
  %31 = add nuw nsw i32 %i.05, 4
  %niter16.next.3 = add i32 %niter16, 4
  %niter16.ncmp.3.not = icmp eq i32 %niter16.next.3, %unroll_iter15
  br i1 %niter16.ncmp.3.not, label %._crit_edge8.unr-lcssa, label %7

._crit_edge8.unr-lcssa:                           ; preds = %7, %.lr.ph7
  %.lcssa.ph = phi double [ undef, %.lr.ph7 ], [ %30, %7 ]
  %i.05.unr = phi i32 [ 1, %.lr.ph7 ], [ %31, %7 ]
  %largest.04.unr = phi double [ %2, %.lr.ph7 ], [ %30, %7 ]
  %lcmp.mod13.not = icmp eq i32 %xtraiter12, 0
  br i1 %lcmp.mod13.not, label %._crit_edge8, label %.epil.preheader11

.epil.preheader11:                                ; preds = %._crit_edge8.unr-lcssa, %.epil.preheader11
  %i.05.epil = phi i32 [ %37, %.epil.preheader11 ], [ %i.05.unr, %._crit_edge8.unr-lcssa ]
  %largest.04.epil = phi double [ %36, %.epil.preheader11 ], [ %largest.04.unr, %._crit_edge8.unr-lcssa ]
  %epil.iter = phi i32 [ %epil.iter.next, %.epil.preheader11 ], [ 0, %._crit_edge8.unr-lcssa ]
  %32 = zext i32 %i.05.epil to i64
  %33 = getelementptr inbounds double, double* %t, i64 %32
  %34 = load double, double* %33, align 8
  %35 = call double @llvm.fabs.f64(double %34)
  %36 = call double @llvm.maxnum.f64(double %35, double %largest.04.epil)
  %37 = add nuw nsw i32 %i.05.epil, 1
  %epil.iter.next = add i32 %epil.iter, 1
  %epil.iter.cmp.not = icmp eq i32 %epil.iter.next, %xtraiter12
  br i1 %epil.iter.cmp.not, label %._crit_edge8, label %.epil.preheader11, !llvm.loop !85

._crit_edge8:                                     ; preds = %.epil.preheader11, %._crit_edge8.unr-lcssa, %0
  %largest.0.lcssa = phi double [ %2, %0 ], [ %.lcssa.ph, %._crit_edge8.unr-lcssa ], [ %36, %.epil.preheader11 ]
  %38 = call i32 @llvm.nvvm.d2i.hi(double %largest.0.lcssa) #139
  %39 = and i32 %38, -4194304
  %40 = sub i32 2144337920, %39
  %41 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %40) #139
  %42 = fmul double %largest.0.lcssa, %41
  %43 = icmp sgt i32 %dim, 0
  br i1 %43, label %.lr.ph, label %._crit_edge

.lr.ph:                                           ; preds = %._crit_edge8
  %xtraiter = and i32 %dim, 1
  %44 = icmp eq i32 %dim, 1
  br i1 %44, label %._crit_edge.unr-lcssa, label %.lr.ph.new

.lr.ph.new:                                       ; preds = %.lr.ph
  %unroll_iter = and i32 %dim, -2
  br label %45

45:                                               ; preds = %._crit_edge17, %.lr.ph.new
  %46 = phi double [ %1, %.lr.ph.new ], [ %.pre, %._crit_edge17 ]
  %j.03 = phi i32 [ 0, %.lr.ph.new ], [ %j.1.1, %._crit_edge17 ]
  %i.12 = phi i32 [ 0, %.lr.ph.new ], [ %63, %._crit_edge17 ]
  %res0.01 = phi double [ 0.000000e+00, %.lr.ph.new ], [ %res0.1.1, %._crit_edge17 ]
  %niter = phi i32 [ 0, %.lr.ph.new ], [ %niter.next.1, %._crit_edge17 ]
  %47 = call double @llvm.fabs.f64(double %46)
  %48 = fmul double %41, %47
  %49 = fcmp une double %48, %42
  %50 = icmp ne i32 %i.12, %j.03
  %or.cond = select i1 %49, i1 true, i1 %50
  %51 = call double @llvm.fma.f64(double %48, double %48, double %res0.01)
  %res0.1 = select i1 %or.cond, double %51, double %res0.01
  %52 = zext i1 %or.cond to i32
  %j.1 = add nuw nsw i32 %j.03, %52
  %53 = or i32 %i.12, 1
  %54 = zext i32 %53 to i64
  %55 = getelementptr inbounds double, double* %t, i64 %54
  %56 = load double, double* %55, align 8
  %57 = call double @llvm.fabs.f64(double %56)
  %58 = fmul double %41, %57
  %59 = fcmp une double %58, %42
  %60 = icmp ne i32 %53, %j.1
  %or.cond.1 = select i1 %59, i1 true, i1 %60
  %61 = call double @llvm.fma.f64(double %58, double %58, double %res0.1)
  %res0.1.1 = select i1 %or.cond.1, double %61, double %res0.1
  %62 = zext i1 %or.cond.1 to i32
  %j.1.1 = add nuw nsw i32 %j.1, %62
  %63 = add nuw nsw i32 %i.12, 2
  %niter.next.1 = add i32 %niter, 2
  %niter.ncmp.1.not = icmp eq i32 %niter.next.1, %unroll_iter
  br i1 %niter.ncmp.1.not, label %._crit_edge.unr-lcssa, label %._crit_edge17

._crit_edge17:                                    ; preds = %45
  %.phi.trans.insert = zext i32 %63 to i64
  %.phi.trans.insert18 = getelementptr inbounds double, double* %t, i64 %.phi.trans.insert
  %.pre = load double, double* %.phi.trans.insert18, align 8
  br label %45

._crit_edge.unr-lcssa:                            ; preds = %45, %.lr.ph
  %res0.1.lcssa.ph = phi double [ undef, %.lr.ph ], [ %res0.1.1, %45 ]
  %j.03.unr = phi i32 [ 0, %.lr.ph ], [ %j.1.1, %45 ]
  %i.12.unr = phi i32 [ 0, %.lr.ph ], [ %63, %45 ]
  %res0.01.unr = phi double [ 0.000000e+00, %.lr.ph ], [ %res0.1.1, %45 ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %._crit_edge, label %.epil.preheader

.epil.preheader:                                  ; preds = %._crit_edge.unr-lcssa
  %64 = zext i32 %i.12.unr to i64
  %65 = getelementptr inbounds double, double* %t, i64 %64
  %66 = load double, double* %65, align 8
  %67 = call double @llvm.fabs.f64(double %66)
  %68 = fmul double %41, %67
  %69 = fcmp une double %68, %42
  %70 = icmp ne i32 %i.12.unr, %j.03.unr
  %or.cond.epil = select i1 %69, i1 true, i1 %70
  %71 = call double @llvm.fma.f64(double %68, double %68, double %res0.01.unr)
  %res0.1.epil = select i1 %or.cond.epil, double %71, double %res0.01.unr
  br label %._crit_edge

._crit_edge:                                      ; preds = %.epil.preheader, %._crit_edge.unr-lcssa, %._crit_edge8
  %res0.0.lcssa = phi double [ 0.000000e+00, %._crit_edge8 ], [ %res0.1.lcssa.ph, %._crit_edge.unr-lcssa ], [ %res0.1.epil, %.epil.preheader ]
  %72 = call double @llvm.fma.f64(double %42, double %42, double %res0.0.lcssa)
  %73 = call double @llvm.minnum.f64(double %72, double noundef 0x7FEFFFFFFFFFFFFF)
  %74 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %73) #141, !srcloc !84
  %75 = fneg double %74
  %76 = fmul double %74, %75
  %77 = call double @llvm.fma.f64(double %73, double %76, double noundef 1.000000e+00)
  %78 = fcmp oeq double %res0.0.lcssa, 0.000000e+00
  br i1 %78, label %88, label %79

79:                                               ; preds = %._crit_edge
  %80 = call double @llvm.fma.f64(double %77, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %81 = fmul double %74, %77
  %82 = call double @llvm.fma.f64(double %80, double %81, double %74)
  %83 = fmul double %72, %82
  %84 = or i32 %39, 1048576
  %85 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %84) #139
  %86 = fmul double %85, %83
  %87 = call i32 @llvm.nvvm.d2i.hi(double noundef 0x7FF0000000000000) #139
  %.not = icmp ult i32 %38, %87
  br i1 %.not, label %89, label %88

88:                                               ; preds = %79, %._crit_edge
  br label %89

89:                                               ; preds = %88, %79
  %res.0 = phi double [ %largest.0.lcssa, %88 ], [ %86, %79 ]
  ret double %res.0
}

; Function Attrs: argmemonly nofree noinline nosync nounwind readonly
define dso_local double @__nv_rnorm(i32 %dim, double* nocapture nofree noundef nonnull readonly align 8 dereferenceable(8) %t) local_unnamed_addr #101 {
  %1 = load double, double* %t, align 8
  %2 = call double @llvm.fabs.f64(double %1)
  %3 = icmp sgt i32 %dim, 1
  br i1 %3, label %.lr.ph7, label %._crit_edge8

.lr.ph7:                                          ; preds = %0
  %4 = add nsw i32 %dim, -1
  %5 = add nsw i32 %dim, -2
  %xtraiter12 = and i32 %4, 3
  %6 = icmp ult i32 %5, 3
  br i1 %6, label %._crit_edge8.unr-lcssa, label %.lr.ph7.new

.lr.ph7.new:                                      ; preds = %.lr.ph7
  %unroll_iter15 = and i32 %4, -4
  br label %7

7:                                                ; preds = %7, %.lr.ph7.new
  %i.05 = phi i32 [ 1, %.lr.ph7.new ], [ %31, %7 ]
  %largest.04 = phi double [ %2, %.lr.ph7.new ], [ %30, %7 ]
  %niter16 = phi i32 [ 0, %.lr.ph7.new ], [ %niter16.next.3, %7 ]
  %8 = zext i32 %i.05 to i64
  %9 = getelementptr inbounds double, double* %t, i64 %8
  %10 = load double, double* %9, align 8
  %11 = call double @llvm.fabs.f64(double %10)
  %12 = call double @llvm.maxnum.f64(double %11, double %largest.04)
  %13 = add nuw nsw i32 %i.05, 1
  %14 = zext i32 %13 to i64
  %15 = getelementptr inbounds double, double* %t, i64 %14
  %16 = load double, double* %15, align 8
  %17 = call double @llvm.fabs.f64(double %16)
  %18 = call double @llvm.maxnum.f64(double %17, double %12)
  %19 = add nuw nsw i32 %i.05, 2
  %20 = zext i32 %19 to i64
  %21 = getelementptr inbounds double, double* %t, i64 %20
  %22 = load double, double* %21, align 8
  %23 = call double @llvm.fabs.f64(double %22)
  %24 = call double @llvm.maxnum.f64(double %23, double %18)
  %25 = add nuw nsw i32 %i.05, 3
  %26 = zext i32 %25 to i64
  %27 = getelementptr inbounds double, double* %t, i64 %26
  %28 = load double, double* %27, align 8
  %29 = call double @llvm.fabs.f64(double %28)
  %30 = call double @llvm.maxnum.f64(double %29, double %24)
  %31 = add nuw nsw i32 %i.05, 4
  %niter16.next.3 = add i32 %niter16, 4
  %niter16.ncmp.3.not = icmp eq i32 %niter16.next.3, %unroll_iter15
  br i1 %niter16.ncmp.3.not, label %._crit_edge8.unr-lcssa, label %7

._crit_edge8.unr-lcssa:                           ; preds = %7, %.lr.ph7
  %.lcssa.ph = phi double [ undef, %.lr.ph7 ], [ %30, %7 ]
  %i.05.unr = phi i32 [ 1, %.lr.ph7 ], [ %31, %7 ]
  %largest.04.unr = phi double [ %2, %.lr.ph7 ], [ %30, %7 ]
  %lcmp.mod13.not = icmp eq i32 %xtraiter12, 0
  br i1 %lcmp.mod13.not, label %._crit_edge8, label %.epil.preheader11

.epil.preheader11:                                ; preds = %._crit_edge8.unr-lcssa, %.epil.preheader11
  %i.05.epil = phi i32 [ %37, %.epil.preheader11 ], [ %i.05.unr, %._crit_edge8.unr-lcssa ]
  %largest.04.epil = phi double [ %36, %.epil.preheader11 ], [ %largest.04.unr, %._crit_edge8.unr-lcssa ]
  %epil.iter = phi i32 [ %epil.iter.next, %.epil.preheader11 ], [ 0, %._crit_edge8.unr-lcssa ]
  %32 = zext i32 %i.05.epil to i64
  %33 = getelementptr inbounds double, double* %t, i64 %32
  %34 = load double, double* %33, align 8
  %35 = call double @llvm.fabs.f64(double %34)
  %36 = call double @llvm.maxnum.f64(double %35, double %largest.04.epil)
  %37 = add nuw nsw i32 %i.05.epil, 1
  %epil.iter.next = add i32 %epil.iter, 1
  %epil.iter.cmp.not = icmp eq i32 %epil.iter.next, %xtraiter12
  br i1 %epil.iter.cmp.not, label %._crit_edge8, label %.epil.preheader11, !llvm.loop !87

._crit_edge8:                                     ; preds = %.epil.preheader11, %._crit_edge8.unr-lcssa, %0
  %largest.0.lcssa = phi double [ %2, %0 ], [ %.lcssa.ph, %._crit_edge8.unr-lcssa ], [ %36, %.epil.preheader11 ]
  %38 = call i32 @llvm.nvvm.d2i.hi(double %largest.0.lcssa) #139
  %39 = and i32 %38, -4194304
  %40 = sub i32 2144337920, %39
  %41 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %40) #139
  %42 = fmul double %largest.0.lcssa, %41
  %43 = icmp sgt i32 %dim, 0
  br i1 %43, label %.lr.ph, label %._crit_edge

.lr.ph:                                           ; preds = %._crit_edge8
  %xtraiter = and i32 %dim, 1
  %44 = icmp eq i32 %dim, 1
  br i1 %44, label %._crit_edge.unr-lcssa, label %.lr.ph.new

.lr.ph.new:                                       ; preds = %.lr.ph
  %unroll_iter = and i32 %dim, -2
  br label %45

45:                                               ; preds = %._crit_edge17, %.lr.ph.new
  %46 = phi double [ %1, %.lr.ph.new ], [ %.pre, %._crit_edge17 ]
  %j.03 = phi i32 [ 0, %.lr.ph.new ], [ %j.1.1, %._crit_edge17 ]
  %i.12 = phi i32 [ 0, %.lr.ph.new ], [ %63, %._crit_edge17 ]
  %res0.01 = phi double [ 0.000000e+00, %.lr.ph.new ], [ %res0.1.1, %._crit_edge17 ]
  %niter = phi i32 [ 0, %.lr.ph.new ], [ %niter.next.1, %._crit_edge17 ]
  %47 = call double @llvm.fabs.f64(double %46)
  %48 = fmul double %41, %47
  %49 = fcmp une double %48, %42
  %50 = icmp ne i32 %i.12, %j.03
  %or.cond = select i1 %49, i1 true, i1 %50
  %51 = call double @llvm.fma.f64(double %48, double %48, double %res0.01)
  %res0.1 = select i1 %or.cond, double %51, double %res0.01
  %52 = zext i1 %or.cond to i32
  %j.1 = add nuw nsw i32 %j.03, %52
  %53 = or i32 %i.12, 1
  %54 = zext i32 %53 to i64
  %55 = getelementptr inbounds double, double* %t, i64 %54
  %56 = load double, double* %55, align 8
  %57 = call double @llvm.fabs.f64(double %56)
  %58 = fmul double %41, %57
  %59 = fcmp une double %58, %42
  %60 = icmp ne i32 %53, %j.1
  %or.cond.1 = select i1 %59, i1 true, i1 %60
  %61 = call double @llvm.fma.f64(double %58, double %58, double %res0.1)
  %res0.1.1 = select i1 %or.cond.1, double %61, double %res0.1
  %62 = zext i1 %or.cond.1 to i32
  %j.1.1 = add nuw nsw i32 %j.1, %62
  %63 = add nuw nsw i32 %i.12, 2
  %niter.next.1 = add i32 %niter, 2
  %niter.ncmp.1.not = icmp eq i32 %niter.next.1, %unroll_iter
  br i1 %niter.ncmp.1.not, label %._crit_edge.unr-lcssa, label %._crit_edge17

._crit_edge17:                                    ; preds = %45
  %.phi.trans.insert = zext i32 %63 to i64
  %.phi.trans.insert18 = getelementptr inbounds double, double* %t, i64 %.phi.trans.insert
  %.pre = load double, double* %.phi.trans.insert18, align 8
  br label %45

._crit_edge.unr-lcssa:                            ; preds = %45, %.lr.ph
  %res0.1.lcssa.ph = phi double [ undef, %.lr.ph ], [ %res0.1.1, %45 ]
  %j.03.unr = phi i32 [ 0, %.lr.ph ], [ %j.1.1, %45 ]
  %i.12.unr = phi i32 [ 0, %.lr.ph ], [ %63, %45 ]
  %res0.01.unr = phi double [ 0.000000e+00, %.lr.ph ], [ %res0.1.1, %45 ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %._crit_edge, label %.epil.preheader

.epil.preheader:                                  ; preds = %._crit_edge.unr-lcssa
  %64 = zext i32 %i.12.unr to i64
  %65 = getelementptr inbounds double, double* %t, i64 %64
  %66 = load double, double* %65, align 8
  %67 = call double @llvm.fabs.f64(double %66)
  %68 = fmul double %41, %67
  %69 = fcmp une double %68, %42
  %70 = icmp ne i32 %i.12.unr, %j.03.unr
  %or.cond.epil = select i1 %69, i1 true, i1 %70
  %71 = call double @llvm.fma.f64(double %68, double %68, double %res0.01.unr)
  %res0.1.epil = select i1 %or.cond.epil, double %71, double %res0.01.unr
  br label %._crit_edge

._crit_edge:                                      ; preds = %.epil.preheader, %._crit_edge.unr-lcssa, %._crit_edge8
  %res0.0.lcssa = phi double [ 0.000000e+00, %._crit_edge8 ], [ %res0.1.lcssa.ph, %._crit_edge.unr-lcssa ], [ %res0.1.epil, %.epil.preheader ]
  %72 = call double @llvm.fma.f64(double %42, double %42, double %res0.0.lcssa)
  %73 = call double @llvm.nvvm.rsqrt.approx.d(double %72) #139
  %74 = fmul double %41, %73
  %75 = fcmp oeq double %largest.0.lcssa, 0x7FF0000000000000
  %res.0 = select i1 %75, double 0.000000e+00, double %74
  ret double %res.0
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_cbrt(double %a) local_unnamed_addr #102 {
  %1 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %3 = and i32 %2, 2147483647
  %4 = fcmp une double %a, 0.000000e+00
  %5 = icmp ult i32 %3, 2146435072
  %or.cond = select i1 %4, i1 %5, i1 false
  br i1 %or.cond, label %__nv_float2int_rn.exit, label %50

__nv_float2int_rn.exit:                           ; preds = %0
  %6 = icmp ult i32 %3, 1048576
  %7 = call double @llvm.nvvm.lohi.i2d(i32 %1, i32 %3) #142
  %8 = fmul double %7, 0x4350000000000000
  %9 = call i32 @llvm.nvvm.d2i.lo(double %8) #142
  %10 = call i32 @llvm.nvvm.d2i.hi(double %8) #142
  %ilo.0 = select i1 %6, i32 %9, i32 %1
  %ihi.0 = select i1 %6, i32 %10, i32 %3
  %denorm.0.neg = select i1 %6, i32 4078, i32 1048576
  %expo.0 = lshr i32 %ihi.0, 20
  %11 = add nsw i32 %expo.0, -1022
  %12 = sitofp i32 %11 to float
  %13 = fmul float %12, 0x3FD5555560000000
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %14, 0
  %15 = call i32 @llvm.nvvm.f2i.rn.ftz(float %13) #139
  %16 = call i32 @llvm.nvvm.f2i.rn(float %13) #139
  %.0 = select i1 %.not, i32 %16, i32 %15
  %.neg = mul i32 %.0, -3145728
  %17 = add i32 %.neg, %ihi.0
  %18 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.0, i32 %17) #139
  %19 = fptrunc double %18 to float
  %20 = call float @llvm.nvvm.lg2.approx.ftz.f(float %19) #139
  %21 = fmul float %20, 0x3FD5555560000000
  %22 = call float @llvm.nvvm.ex2.approx.ftz.f(float %21) #139
  %23 = fpext float %22 to double
  %24 = fmul double %23, %23
  %25 = call i32 @llvm.nvvm.d2i.lo(double %24) #139
  %26 = call i32 @llvm.nvvm.d2i.hi(double %24) #139
  %27 = add i32 %26, 1048576
  %28 = call double @llvm.nvvm.lohi.i2d(i32 %25, i32 %27) #139
  %29 = call double @llvm.fma.f64(double %28, double %23, double %18)
  %30 = call double @llvm.nvvm.rcp.approx.ftz.d(double %29) #139
  %31 = fneg double %29
  %32 = call double @llvm.fma.f64(double %31, double %30, double noundef 1.000000e+00)
  %33 = call double @llvm.fma.f64(double %32, double %32, double %32)
  %34 = call double @llvm.fma.f64(double %33, double %30, double %30)
  %35 = fneg double %23
  %36 = call double @llvm.fma.f64(double %24, double %35, double %18)
  %37 = fmul double %36, %34
  %38 = call double @llvm.fma.f64(double %23, double %37, double %23)
  %39 = call i32 @llvm.nvvm.d2i.lo(double %38) #139
  %40 = call i32 @llvm.nvvm.d2i.hi(double %38) #139
  %41 = add i32 %.0, %denorm.0.neg
  %42 = shl i32 %41, 20
  %43 = add nsw i32 %40, %42
  %44 = call double @llvm.nvvm.lohi.i2d(i32 %39, i32 %43) #139
  %45 = call i32 @llvm.nvvm.d2i.lo(double %44) #139
  %46 = call i32 @llvm.nvvm.d2i.hi(double %44) #139
  %47 = and i32 %2, -2147483648
  %48 = or i32 %46, %47
  %49 = call double @llvm.nvvm.lohi.i2d(i32 %45, i32 %48) #139
  br label %52

50:                                               ; preds = %0
  %51 = fadd double %a, %a
  br label %52

52:                                               ; preds = %50, %__nv_float2int_rn.exit
  %t.0 = phi double [ %49, %__nv_float2int_rn.exit ], [ %51, %50 ]
  ret double %t.0
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_rcbrt(double %a) local_unnamed_addr #103 {
  %1 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %3 = and i32 %2, 2147483647
  %4 = fcmp une double %a, 0.000000e+00
  %5 = icmp ult i32 %3, 2146435072
  %or.cond = select i1 %4, i1 %5, i1 false
  br i1 %or.cond, label %__nv_float2int_rn.exit, label %43

__nv_float2int_rn.exit:                           ; preds = %0
  %6 = icmp ult i32 %3, 1048576
  %7 = call double @llvm.nvvm.lohi.i2d(i32 %1, i32 %3) #142
  %8 = fmul double %7, 0x4350000000000000
  %9 = call i32 @llvm.nvvm.d2i.lo(double %8) #142
  %10 = call i32 @llvm.nvvm.d2i.hi(double %8) #142
  %ilo.0 = select i1 %6, i32 %9, i32 %1
  %ihi.0 = select i1 %6, i32 %10, i32 %3
  %denorm.0.neg.neg = select i1 %6, i32 18, i32 1048576
  %expo.0 = lshr i32 %ihi.0, 20
  %11 = add nsw i32 %expo.0, -1022
  %12 = sitofp i32 %11 to float
  %13 = fmul float %12, 0x3FD5555560000000
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %14, 0
  %15 = call i32 @llvm.nvvm.f2i.rn.ftz(float %13) #139
  %16 = call i32 @llvm.nvvm.f2i.rn(float %13) #139
  %.0 = select i1 %.not, i32 %16, i32 %15
  %.neg = mul i32 %.0, -3145728
  %17 = add i32 %.neg, %ihi.0
  %18 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.0, i32 %17) #139
  %19 = fptrunc double %18 to float
  %20 = call float @llvm.nvvm.lg2.approx.ftz.f(float %19) #139
  %21 = fmul float %20, 0xBFD5555560000000
  %22 = call float @llvm.nvvm.ex2.approx.ftz.f(float %21) #139
  %23 = fpext float %22 to double
  %24 = fmul double %23, %23
  %25 = fmul double %18, %23
  %26 = fneg double %25
  %27 = call double @llvm.fma.f64(double %24, double %26, double noundef 1.000000e+00)
  %28 = call double @llvm.fma.f64(double %18, double %23, double %26)
  %29 = fneg double %28
  %30 = call double @llvm.fma.f64(double %24, double %29, double %27)
  %31 = fneg double %24
  %32 = call double @llvm.fma.f64(double %23, double %23, double %31)
  %33 = fneg double %32
  %34 = call double @llvm.fma.f64(double %25, double %33, double %30)
  %35 = call double @llvm.fma.f64(double %34, double noundef 0x3FCC71C71C71C71C, double noundef 0x3FD5555555555555)
  %36 = fmul double %34, %23
  %37 = call double @llvm.fma.f64(double %35, double %36, double %23)
  %38 = call i32 @llvm.nvvm.d2i.lo(double %37) #139
  %39 = call i32 @llvm.nvvm.d2i.hi(double %37) #139
  %.neg1 = sub i32 %denorm.0.neg.neg, %.0
  %40 = shl i32 %.neg1, 20
  %41 = add nsw i32 %39, %40
  %42 = call double @llvm.nvvm.lohi.i2d(i32 %38, i32 %41) #139
  br label %49

43:                                               ; preds = %0
  %44 = xor i32 %3, 2146435072
  %45 = call double @llvm.nvvm.lohi.i2d(i32 %1, i32 %44) #139
  %46 = call double @llvm.fabs.f64(double %a)
  %47 = fcmp ugt double %46, 0x7FF0000000000000
  %48 = fadd double %a, %a
  %spec.select = select i1 %47, double %48, double %45
  br label %49

49:                                               ; preds = %43, %__nv_float2int_rn.exit
  %t.1 = phi double [ %42, %__nv_float2int_rn.exit ], [ %spec.select, %43 ]
  %50 = call i32 @llvm.nvvm.d2i.lo(double %t.1) #139
  %51 = call i32 @llvm.nvvm.d2i.hi(double %t.1) #139
  %52 = and i32 %2, -2147483648
  %53 = or i32 %51, %52
  %54 = call double @llvm.nvvm.lohi.i2d(i32 %50, i32 %53) #139
  ret double %54
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_pow(double %a, double %b) local_unnamed_addr #104 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.hi(double %b) #142
  %3 = lshr i32 %2, 20
  %4 = and i32 %3, 2047
  %5 = add nsw i32 %4, -1012
  %6 = bitcast double %b to i64
  %7 = zext i32 %5 to i64
  %8 = shl i64 %6, %7
  %9 = icmp eq i64 %8, -9223372036854775808
  %10 = zext i1 %9 to i32
  %11 = call double @llvm.fabs.f64(double %a) #140
  %12 = call fastcc double @__internal_accurate_pow(double %11, double %b) #139
  %13 = fcmp oeq double %a, 0.000000e+00
  br i1 %13, label %14, label %20

14:                                               ; preds = %0
  %15 = call double @llvm.fabs.f64(double %b)
  %16 = fcmp oeq double %15, 5.000000e-01
  %bIsOddInteger.0 = select i1 %16, i32 0, i32 %10
  %.not = icmp eq i32 %bIsOddInteger.0, 0
  %thi.0 = select i1 %.not, i32 0, i32 %1
  %17 = icmp slt i32 %2, 0
  %18 = or i32 %thi.0, 2146435072
  %thi.1 = select i1 %17, i32 %18, i32 %thi.0
  %19 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %thi.1) #139
  br label %28

20:                                               ; preds = %0
  %21 = icmp slt i32 %1, 0
  %or.cond = select i1 %21, i1 %9, i1 false
  %22 = call i32 @llvm.nvvm.d2i.lo(double %12) #139
  %23 = call i32 @llvm.nvvm.d2i.hi(double %12) #139
  %24 = xor i32 %23, -2147483648
  %25 = call double @llvm.nvvm.lohi.i2d(i32 %22, i32 %24) #139
  %t.0 = select i1 %or.cond, double %25, double %12
  %26 = call double @llvm.trunc.f64(double %b)
  %27 = fcmp une double %26, %b
  %or.cond2 = select i1 %21, i1 %27, i1 false
  %t.1 = select i1 %or.cond2, double 0xFFF8000000000000, double %t.0
  br label %28

28:                                               ; preds = %20, %14
  %t.2 = phi double [ %19, %14 ], [ %t.1, %20 ]
  %bIsOddInteger.1 = phi i32 [ %bIsOddInteger.0, %14 ], [ %10, %20 ]
  %29 = fadd double %a, %b
  %30 = call i32 @llvm.nvvm.d2i.hi(double %29) #139
  %31 = and i32 %30, 2146435072
  %32 = icmp eq i32 %31, 2146435072
  br i1 %32, label %33, label %84

33:                                               ; preds = %28
  %34 = fcmp ugt double %11, 0x7FF0000000000000
  %35 = call double @llvm.fabs.f64(double %b)
  %36 = fcmp ugt double %35, 0x7FF0000000000000
  %or.cond8 = select i1 %34, i1 true, i1 %36
  br i1 %or.cond8, label %84, label %37

37:                                               ; preds = %33
  %38 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %39 = icmp eq i32 %38, 200
  br i1 %39, label %46, label %40

40:                                               ; preds = %37
  %41 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %42 = icmp eq i32 %41, 350
  br i1 %42, label %46, label %43

43:                                               ; preds = %40
  %44 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %45 = icmp eq i32 %44, 370
  br i1 %45, label %46, label %48

46:                                               ; preds = %43, %40, %37
  %47 = fcmp oeq double %35, 0x7FF0000000000000
  br label %__nv_isinfd.exit4

48:                                               ; preds = %43
  %49 = call i32 @llvm.nvvm.d2i.lo(double %b) #139
  %50 = and i32 %2, 2147483647
  %51 = icmp eq i32 %50, 2146435072
  %52 = icmp eq i32 %49, 0
  %53 = select i1 %51, i1 %52, i1 false
  br label %__nv_isinfd.exit4

__nv_isinfd.exit4:                                ; preds = %48, %46
  %.01.in = phi i1 [ %47, %46 ], [ %53, %48 ]
  br i1 %.01.in, label %54, label %60

54:                                               ; preds = %__nv_isinfd.exit4
  %55 = fcmp ogt double %11, 1.000000e+00
  %thi.2 = select i1 %55, i32 2146435072, i32 0
  %56 = icmp slt i32 %2, 0
  %57 = xor i32 %thi.2, 2146435072
  %thi.3 = select i1 %56, i32 %57, i32 %thi.2
  %58 = fcmp oeq double %a, -1.000000e+00
  %thi.4 = select i1 %58, i32 1072693248, i32 %thi.3
  %59 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %thi.4) #139
  br label %84

60:                                               ; preds = %__nv_isinfd.exit4
  %61 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %62 = icmp eq i32 %61, 200
  br i1 %62, label %69, label %63

63:                                               ; preds = %60
  %64 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %65 = icmp eq i32 %64, 350
  br i1 %65, label %69, label %66

66:                                               ; preds = %63
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %68 = icmp eq i32 %67, 370
  br i1 %68, label %69, label %71

69:                                               ; preds = %66, %63, %60
  %70 = fcmp oeq double %11, 0x7FF0000000000000
  br label %__nv_isinfd.exit

71:                                               ; preds = %66
  %72 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %73 = and i32 %1, 2147483647
  %74 = icmp eq i32 %73, 2146435072
  %75 = icmp eq i32 %72, 0
  %76 = select i1 %74, i1 %75, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %71, %69
  %.0.in = phi i1 [ %70, %69 ], [ %76, %71 ]
  br i1 %.0.in, label %77, label %84

77:                                               ; preds = %__nv_isinfd.exit
  %.inv = icmp slt i32 %2, 0
  %spec.select = select i1 %.inv, i32 0, i32 2146435072
  %78 = icmp slt i32 %1, 0
  %79 = icmp ne i32 %bIsOddInteger.1, 0
  %or.cond3 = select i1 %78, i1 %79, i1 false
  %80 = and i32 %2, 2147483647
  %81 = icmp ne i32 %80, 1071644672
  %or.cond5 = select i1 %or.cond3, i1 %81, i1 false
  %82 = or i32 %spec.select, -2147483648
  %thi.6 = select i1 %or.cond5, i32 %82, i32 %spec.select
  %83 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %thi.6) #139
  br label %84

84:                                               ; preds = %33, %__nv_isinfd.exit, %77, %54, %28
  %t.6 = phi double [ %t.2, %28 ], [ %59, %54 ], [ %83, %77 ], [ %t.2, %__nv_isinfd.exit ], [ %29, %33 ]
  %85 = fcmp oeq double %a, 1.000000e+00
  %86 = fcmp oeq double %b, 0.000000e+00
  %or.cond6 = select i1 %85, i1 true, i1 %86
  %t.7 = select i1 %or.cond6, double 1.000000e+00, double %t.6
  ret double %t.7
}

; Function Attrs: noinline nounwind
define internal fastcc double @__internal_accurate_pow(double %a, double %b) unnamed_addr #105 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %3 = lshr i32 %1, 20
  %4 = icmp ult i32 %1, 1048576
  br i1 %4, label %5, label %11

5:                                                ; preds = %0
  %6 = fmul double %a, 0x4350000000000000
  %7 = call i32 @llvm.nvvm.d2i.hi(double %6) #142
  %8 = call i32 @llvm.nvvm.d2i.lo(double %6) #142
  %9 = lshr i32 %7, 20
  %10 = add nsw i32 %9, -54
  br label %11

11:                                               ; preds = %5, %0
  %ilo.i.0 = phi i32 [ %8, %5 ], [ %2, %0 ]
  %ihi.i.0 = phi i32 [ %7, %5 ], [ %1, %0 ]
  %expo.i.0 = phi i32 [ %10, %5 ], [ %3, %0 ]
  %12 = add nsw i32 %expo.i.0, -1023
  %13 = and i32 %ihi.i.0, -2146435073
  %14 = or i32 %13, 1072693248
  %15 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %14) #142
  %16 = icmp ugt i32 %14, 1073127582
  br i1 %16, label %17, label %__internal_log_ext_prec.exit

17:                                               ; preds = %11
  %18 = call i32 @llvm.nvvm.d2i.lo(double %15) #142
  %19 = call i32 @llvm.nvvm.d2i.hi(double %15) #142
  %20 = add i32 %19, -1048576
  %21 = call double @llvm.nvvm.lohi.i2d(i32 %18, i32 %20) #142
  %22 = add nsw i32 %expo.i.0, -1022
  br label %__internal_log_ext_prec.exit

__internal_log_ext_prec.exit:                     ; preds = %17, %11
  %m.i.0 = phi double [ %21, %17 ], [ %15, %11 ]
  %expo.i.1 = phi i32 [ %22, %17 ], [ %12, %11 ]
  %23 = fadd double %m.i.0, -1.000000e+00
  %24 = fadd double %m.i.0, 1.000000e+00
  %25 = call double @llvm.nvvm.rcp.approx.ftz.d(double %24) #142
  %26 = fneg double %24
  %27 = call double @llvm.fma.f64(double %26, double %25, double noundef 1.000000e+00) #140
  %28 = call double @llvm.fma.f64(double %27, double %27, double %27) #140
  %29 = call double @llvm.fma.f64(double %28, double %25, double %25) #140
  %30 = fmul double %23, %29
  %31 = fadd double %30, %30
  %32 = fmul double %31, %31
  %33 = call double @llvm.fma.f64(double %32, double noundef 0x3EB0F5FF7D2CAFE2, double noundef 0x3ED0F5D241AD3B5A) #140
  %34 = call double @llvm.fma.f64(double %33, double %32, double noundef 0x3EF3B20A75488A3F) #140
  %35 = call double @llvm.fma.f64(double %34, double %32, double noundef 0x3F1745CDE4FAECD5) #140
  %36 = call double @llvm.fma.f64(double %35, double %32, double noundef 0x3F3C71C7258A578B) #140
  %37 = call double @llvm.fma.f64(double %36, double %32, double noundef 0x3F6249249242B910) #140
  %38 = call double @llvm.fma.f64(double %37, double %32, double noundef 0x3F89999999999DFB) #140
  %39 = fmul double %32, %38
  %40 = fsub double %23, %31
  %41 = fmul double %40, 2.000000e+00
  %42 = fneg double %31
  %43 = call double @llvm.fma.f64(double %42, double %23, double %41) #140
  %44 = fmul double %29, %43
  %45 = fadd double %39, 0x3FB5555555555555
  %46 = fsub double 0x3FB5555555555555, %45
  %47 = fadd double %39, %46
  %48 = fadd double %47, 0.000000e+00
  %49 = fadd double %48, 0xBC46A4CB00B9E7B0
  %50 = fadd double %45, %49
  %51 = fsub double %45, %50
  %52 = fadd double %49, %51
  %53 = fneg double %32
  %54 = call double @llvm.fma.f64(double %31, double %31, double %53) #140
  %55 = call i32 @llvm.nvvm.d2i.lo(double %44) #142
  %56 = call i32 @llvm.nvvm.d2i.hi(double %44) #142
  %57 = add i32 %56, 1048576
  %58 = call double @llvm.nvvm.lohi.i2d(i32 %55, i32 %57) #142
  %59 = call double @llvm.fma.f64(double %31, double %58, double %54) #140
  %60 = fmul double %31, %32
  %61 = fneg double %60
  %62 = call double @llvm.fma.f64(double %32, double %31, double %61) #140
  %63 = call double @llvm.fma.f64(double %32, double %44, double %62) #140
  %64 = call double @llvm.fma.f64(double %59, double %31, double %63) #140
  %65 = fmul double %60, %50
  %66 = fneg double %65
  %67 = call double @llvm.fma.f64(double %50, double %60, double %66) #140
  %68 = call double @llvm.fma.f64(double %50, double %64, double %67) #140
  %69 = call double @llvm.fma.f64(double %52, double %60, double %68) #140
  %70 = fadd double %65, %69
  %71 = fsub double %65, %70
  %72 = fadd double %69, %71
  %73 = fadd double %31, %70
  %74 = fsub double %31, %73
  %75 = fadd double %70, %74
  %76 = fadd double %72, %75
  %77 = fadd double %44, %76
  %78 = fadd double %73, %77
  %79 = fsub double %73, %78
  %80 = fadd double %77, %79
  %81 = xor i32 %expo.i.1, -2147483648
  %82 = call double @llvm.nvvm.lohi.i2d(i32 %81, i32 noundef 1127219200) #142
  %83 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %84 = fsub double %82, %83
  %85 = call double @llvm.fma.f64(double %84, double noundef 0x3FE62E42FEFA39EF, double %78) #140
  %86 = fneg double %84
  %87 = call double @llvm.fma.f64(double %86, double noundef 0x3FE62E42FEFA39EF, double %85) #140
  %88 = fsub double %87, %78
  %89 = fsub double %80, %88
  %90 = call double @llvm.fma.f64(double %84, double noundef 0x3C7ABC9E3B39803F, double %89) #140
  %91 = fadd double %85, %90
  %92 = fsub double %85, %91
  %93 = fadd double %90, %92
  %94 = call i32 @llvm.nvvm.d2i.hi(double %b) #142
  %95 = call i32 @llvm.nvvm.d2i.lo(double %b) #142
  %96 = shl i32 %94, 1
  %97 = icmp ugt i32 %96, -33554433
  %98 = and i32 %94, -15728641
  %spec.select = select i1 %97, i32 %98, i32 %94
  %99 = call double @llvm.nvvm.lohi.i2d(i32 %95, i32 %spec.select) #142
  %100 = fmul double %99, %91
  %101 = fneg double %100
  %102 = call double @llvm.fma.f64(double %91, double %99, double %101) #140
  %103 = call double @llvm.fma.f64(double %93, double %99, double %102) #140
  %104 = fadd double %100, %103
  %105 = call double @llvm.fma.f64(double %104, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %106 = call i32 @llvm.nvvm.d2i.lo(double %105) #142
  %107 = fadd double %105, 0xC338000000000000
  %108 = call double @llvm.fma.f64(double %107, double noundef 0xBFE62E42FEFA39EF, double %104) #140
  %109 = call double @llvm.fma.f64(double %107, double noundef 0xBC7ABC9E3B39803F, double %108) #140
  %110 = call double @llvm.fma.f64(double %109, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %111 = call double @llvm.fma.f64(double %110, double %109, double noundef 0x3EC71DEE62401315) #140
  %112 = call double @llvm.fma.f64(double %111, double %109, double noundef 0x3EFA01997C89EB71) #140
  %113 = call double @llvm.fma.f64(double %112, double %109, double noundef 0x3F2A01A014761F65) #140
  %114 = call double @llvm.fma.f64(double %113, double %109, double noundef 0x3F56C16C1852B7AF) #140
  %115 = call double @llvm.fma.f64(double %114, double %109, double noundef 0x3F81111111122322) #140
  %116 = call double @llvm.fma.f64(double %115, double %109, double noundef 0x3FA55555555502A1) #140
  %117 = call double @llvm.fma.f64(double %116, double %109, double noundef 0x3FC5555555555511) #140
  %118 = call double @llvm.fma.f64(double %117, double %109, double noundef 0x3FE000000000000B) #140
  %119 = call double @llvm.fma.f64(double %118, double %109, double noundef 1.000000e+00) #140
  %120 = call double @llvm.fma.f64(double %119, double %109, double noundef 1.000000e+00) #140
  %121 = call i32 @llvm.nvvm.d2i.lo(double %120) #142
  %122 = call i32 @llvm.nvvm.d2i.hi(double %120) #142
  %123 = shl i32 %106, 20
  %124 = add i32 %122, %123
  %125 = call double @llvm.nvvm.lohi.i2d(i32 %121, i32 %124) #142
  %126 = call i32 @llvm.nvvm.d2i.hi(double %104) #142
  %127 = bitcast i32 %126 to float
  %128 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %128, 0
  %129 = call float @llvm.nvvm.fabs.ftz.f(float %127) #139
  %130 = call float @llvm.fabs.f32(float %127)
  %.02 = select i1 %.not, float %130, float %129
  %131 = fcmp olt float %.02, 0x4010C46560000000
  br i1 %131, label %__nv_exp.exit, label %__internal_fast_icmp_abs_lt.exit.i

__internal_fast_icmp_abs_lt.exit.i:               ; preds = %__internal_log_ext_prec.exit
  %132 = fcmp olt double %104, 0.000000e+00
  %133 = fadd double %104, 0x7FF0000000000000
  %z.i.0 = select i1 %132, double 0.000000e+00, double %133
  %134 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %134, 0
  %.01 = select i1 %.not2, float %130, float %129
  %135 = fcmp olt float %.01, 0x4010E90000000000
  br i1 %135, label %136, label %__nv_exp.exit

136:                                              ; preds = %__internal_fast_icmp_abs_lt.exit.i
  %137 = sdiv i32 %106, 2
  %138 = shl i32 %137, 20
  %139 = add i32 %122, %138
  %140 = call double @llvm.nvvm.lohi.i2d(i32 %121, i32 %139) #139
  %141 = sub nsw i32 %106, %137
  %142 = shl i32 %141, 20
  %143 = add nsw i32 %142, 1072693248
  %144 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %143) #139
  %145 = fmul double %144, %140
  br label %__nv_exp.exit

__nv_exp.exit:                                    ; preds = %__internal_fast_icmp_abs_lt.exit.i, %136, %__internal_log_ext_prec.exit
  %z.i.2 = phi double [ %125, %__internal_log_ext_prec.exit ], [ %145, %136 ], [ %z.i.0, %__internal_fast_icmp_abs_lt.exit.i ]
  %146 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %147 = icmp eq i32 %146, 200
  br i1 %147, label %154, label %148

148:                                              ; preds = %__nv_exp.exit
  %149 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %150 = icmp eq i32 %149, 350
  br i1 %150, label %154, label %151

151:                                              ; preds = %148
  %152 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %153 = icmp eq i32 %152, 370
  br i1 %153, label %154, label %157

154:                                              ; preds = %151, %148, %__nv_exp.exit
  %155 = call double @llvm.fabs.f64(double %z.i.2)
  %156 = fcmp oeq double %155, 0x7FF0000000000000
  br label %__nv_isinfd.exit

157:                                              ; preds = %151
  %158 = call i32 @llvm.nvvm.d2i.lo(double %z.i.2) #139
  %159 = call i32 @llvm.nvvm.d2i.hi(double %z.i.2) #139
  %160 = and i32 %159, 2147483647
  %161 = icmp eq i32 %160, 2146435072
  %162 = icmp eq i32 %158, 0
  %163 = select i1 %161, i1 %162, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %157, %154
  %.0.in = phi i1 [ %156, %154 ], [ %163, %157 ]
  %164 = fsub double %100, %104
  %165 = fadd double %103, %164
  %166 = call double @llvm.fma.f64(double %z.i.2, double %165, double %z.i.2)
  %spec.select1 = select i1 %.0.in, double %z.i.2, double %166
  ret double %spec.select1
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_j0(double %a) local_unnamed_addr #106 {
  %q.i.i.i = alloca i32, align 4
  %q.i.i.i.i = alloca i32, align 4
  %q.i.i = alloca i32, align 4
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ugt double %1, 0x400FB319F277BBE5
  br i1 %2, label %23, label %3

3:                                                ; preds = %0
  %4 = fadd double %1, 0xC0033D152E971B40
  %5 = fadd double %4, 0x3CA0F539D7DA258E
  %6 = call double @llvm.fma.f64(double %5, double noundef 0xBCC0D18564C48C61, double noundef 0xBCFCF8F9A8C294BC) #140
  %7 = call double @llvm.fma.f64(double %6, double %5, double noundef 0x3D3FAB983CAE498B) #140
  %8 = call double @llvm.fma.f64(double %7, double %5, double noundef 0x3D7CD7C018579B88) #140
  %9 = call double @llvm.fma.f64(double %8, double %5, double noundef 0xBDBBDD2342D64FDD) #140
  %10 = call double @llvm.fma.f64(double %9, double %5, double noundef 0xBDF5C2D9416B1E2B) #140
  %11 = call double @llvm.fma.f64(double %10, double %5, double noundef 0x3E32951D73174DD5) #140
  %12 = call double @llvm.fma.f64(double %11, double %5, double noundef 0x3E67FF99802CAEB5) #140
  %13 = call double @llvm.fma.f64(double %12, double %5, double noundef 0xBEA1CCE305C4C9F7) #140
  %14 = call double @llvm.fma.f64(double %13, double %5, double noundef 0xBED232C77E29E1BB) #140
  %15 = call double @llvm.fma.f64(double %14, double %5, double noundef 0x3F06ED3B9F0EF757) #140
  %16 = call double @llvm.fma.f64(double %15, double %5, double noundef 0x3F315382BA096A62) #140
  %17 = call double @llvm.fma.f64(double %16, double %5, double noundef 0xBF61F992590D1AE4) #140
  %18 = call double @llvm.fma.f64(double %17, double %5, double noundef 0xBF81BB1CBE1A465F) #140
  %19 = call double @llvm.fma.f64(double %18, double %5, double noundef 0x3FACFAE864368D84) #140
  %20 = call double @llvm.fma.f64(double %19, double %5, double noundef 0x3FBBA1DEEA0294A3) #140
  %21 = call double @llvm.fma.f64(double %20, double %5, double noundef 0xBFE09CDB36551280) #140
  %22 = fmul double %5, %21
  br label %256

23:                                               ; preds = %0
  %24 = fcmp ugt double %1, 0x401C58FD1A62F5EC
  br i1 %24, label %45, label %25

25:                                               ; preds = %23
  %26 = fadd double %1, 0xC016148F5B2C2E45
  %27 = fadd double %26, 0xBC975054CD60A517
  %28 = call double @llvm.fma.f64(double %27, double noundef 0x3CBCB0A8F126B343, double noundef 0x3CF83FD1F333EB61) #140
  %29 = call double @llvm.fma.f64(double %28, double %27, double noundef 0xBD4100E33E3FB413) #140
  %30 = call double @llvm.fma.f64(double %29, double %27, double noundef 0xBD7846076D004627) #140
  %31 = call double @llvm.fma.f64(double %30, double %27, double noundef 0x3DBE2F1D4F90720D) #140
  %32 = call double @llvm.fma.f64(double %31, double %27, double noundef 0x3DF1D03B1E4A119B) #140
  %33 = call double @llvm.fma.f64(double %32, double %27, double noundef 0xBE341D72B1B3BCE9) #140
  %34 = call double @llvm.fma.f64(double %33, double %27, double noundef 0xBE62DA37CE2A9EF8) #140
  %35 = call double @llvm.fma.f64(double %34, double %27, double noundef 0x3EA32E6D9974F763) #140
  %36 = call double @llvm.fma.f64(double %35, double %27, double noundef 0x3ECAD77D744A1879) #140
  %37 = call double @llvm.fma.f64(double %36, double %27, double noundef 0xBF0863F481A37337) #140
  %38 = call double @llvm.fma.f64(double %37, double %27, double noundef 0xBF26F641F418F0F4) #140
  %39 = call double @llvm.fma.f64(double %38, double %27, double noundef 0x3F627E31FE9A969E) #140
  %40 = call double @llvm.fma.f64(double %39, double %27, double noundef 0x3F72F7FFE9025628) #140
  %41 = call double @llvm.fma.f64(double %40, double %27, double noundef 0xBFAB2150CB41E8BF) #140
  %42 = call double @llvm.fma.f64(double %41, double %27, double noundef 0xBF9F8F72E7A848DE) #140
  %43 = call double @llvm.fma.f64(double %42, double %27, double noundef 0x3FD5C6E60A097823) #140
  %44 = fmul double %27, %43
  br label %256

45:                                               ; preds = %23
  %46 = fcmp ugt double %1, 0x402471FCB6A7A8C0
  br i1 %46, label %67, label %47

47:                                               ; preds = %45
  %48 = fadd double %1, 0xC0214EB56CCCDECA
  %49 = fadd double %48, 0x3CB51970714C7C25
  %50 = call double @llvm.fma.f64(double %49, double noundef 0xBCBDB7FFCF659E24, double noundef 0xBCF4B3A71AAAC629) #140
  %51 = call double @llvm.fma.f64(double %50, double %49, double noundef 0x3D417EC150ECDCE7) #140
  %52 = call double @llvm.fma.f64(double %51, double %49, double noundef 0x3D7438F5EA1D10B2) #140
  %53 = call double @llvm.fma.f64(double %52, double %49, double noundef 0xBDBEDAE7EC2C9E87) #140
  %54 = call double @llvm.fma.f64(double %53, double %49, double noundef 0xBDECADD2C4B91F58) #140
  %55 = call double @llvm.fma.f64(double %54, double %49, double noundef 0x3E34582C8EE12204) #140
  %56 = call double @llvm.fma.f64(double %55, double %49, double noundef 0x3E5CEDA451DD20F8) #140
  %57 = call double @llvm.fma.f64(double %56, double %49, double noundef 0xBEA30E8CC3165E2F) #140
  %58 = call double @llvm.fma.f64(double %57, double %49, double noundef 0xBEC3324842BB1A2E) #140
  %59 = call double @llvm.fma.f64(double %58, double %49, double noundef 0x3F07800BC54FBDDB) #140
  %60 = call double @llvm.fma.f64(double %59, double %49, double noundef 0x3F1D79605276949A) #140
  %61 = call double @llvm.fma.f64(double %60, double %49, double noundef 0xBF60E0D60385A629) #140
  %62 = call double @llvm.fma.f64(double %61, double %49, double noundef 0xBF648E63600D82F3) #140
  %63 = call double @llvm.fma.f64(double %62, double %49, double noundef 0x3FA68B984EC6493A) #140
  %64 = call double @llvm.fma.f64(double %63, double %49, double noundef 0x3F900F7FCF183E0B) #140
  %65 = call double @llvm.fma.f64(double %64, double %49, double noundef 0xBFD15F7977A772D4) #140
  %66 = fmul double %49, %65
  br label %256

67:                                               ; preds = %45
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %69 = icmp eq i32 %68, 200
  br i1 %69, label %76, label %70

70:                                               ; preds = %67
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %72 = icmp eq i32 %71, 350
  br i1 %72, label %76, label %73

73:                                               ; preds = %70
  %74 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %75 = icmp eq i32 %74, 370
  br i1 %75, label %76, label %78

76:                                               ; preds = %73, %70, %67
  %77 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit

78:                                               ; preds = %73
  %79 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %80 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %81 = and i32 %80, 2147483647
  %82 = icmp eq i32 %81, 2146435072
  %83 = icmp eq i32 %79, 0
  %84 = select i1 %82, i1 %83, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %78, %76
  %.05.in = phi i1 [ %77, %76 ], [ %84, %78 ]
  br i1 %.05.in, label %256, label %85

85:                                               ; preds = %__nv_isinfd.exit
  %86 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %87 = fneg double %1
  %88 = call double @llvm.fma.f64(double %87, double %86, double noundef 1.000000e+00)
  %89 = call double @llvm.fma.f64(double %88, double %88, double %88)
  %90 = call double @llvm.fma.f64(double %89, double %86, double %86)
  %91 = fmul double %90, %90
  %92 = call double @llvm.fma.f64(double %91, double noundef 0xC0D115CB8C11A9DC, double noundef 0x409927467A655012)
  %93 = call double @llvm.fma.f64(double %92, double %91, double noundef 0xC05751787E247BD4)
  %94 = call double @llvm.fma.f64(double %93, double %91, double noundef 0x401704C4E5FC36B2)
  %95 = call double @llvm.fma.f64(double %94, double %91, double noundef 0xBFE15B747A2FD531)
  %96 = call double @llvm.fma.f64(double %95, double %91, double noundef 0x3FBA7FEACF6CB79B)
  %97 = call double @llvm.fma.f64(double %96, double %91, double noundef 0xBFAFFFFFEDDCF548)
  %98 = call double @llvm.fma.f64(double %97, double %91, double noundef 0x3FEFFFFFFFFFC9E5)
  %99 = call double @llvm.fma.f64(double %91, double noundef 0xC14602FE1C34685E, double noundef 0x410ECD4523B12B84)
  %100 = call double @llvm.fma.f64(double %99, double %91, double noundef 0xC0C7A2FC1972F05A)
  %101 = call double @llvm.fma.f64(double %100, double %91, double noundef 0x407EBA131F7E5BEB)
  %102 = call double @llvm.fma.f64(double %101, double %91, double noundef 0xC0373B92E6E7CC7D)
  %103 = call double @llvm.fma.f64(double %102, double %91, double noundef 0x3FFA31BEE63A2F08)
  %104 = call double @llvm.fma.f64(double %103, double %91, double noundef 0xBFCAD320104D5D05)
  %105 = call double @llvm.fma.f64(double %104, double %91, double noundef 0x3FB0AAAA9C76D07E)
  %106 = call double @llvm.fma.f64(double %105, double %91, double noundef 0xBFBFFFFFFFFDACEC)
  %107 = call double @llvm.fma.f64(double %106, double %90, double %1)
  %108 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %109 = fmul double %108, 0x3FE9884533D43651
  %110 = fmul double %109, %98
  %111 = fmul double %107, 0x3FE45F306DC9C883
  %112 = call i32 @llvm.nvvm.d2i.rn(double %111) #139
  store i32 %112, i32* %q.i.i, align 4, !noalias !68
  %113 = sitofp i32 %112 to double
  %114 = fneg double %113
  %115 = call double @llvm.fma.f64(double %114, double noundef 0x3FF921FB54442D18, double %107)
  %116 = call double @llvm.fma.f64(double %114, double noundef 0x3C91A62633145C00, double %115)
  %117 = call double @llvm.fma.f64(double %114, double noundef 0x397B839A252049C0, double %116)
  %118 = call double @llvm.fabs.f64(double %107)
  %119 = fcmp ult double %118, 0x41E0000000000000
  br i1 %119, label %__internal_trig_reduction_kerneld.exit.i, label %120

120:                                              ; preds = %85
  %121 = call fastcc double @__internal_trig_reduction_slowpathd(double %107, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i) #139
  %.pre = load i32, i32* %q.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i

__internal_trig_reduction_kerneld.exit.i:         ; preds = %120, %85
  %122 = phi i32 [ %.pre, %120 ], [ %112, %85 ]
  %t.i.i.0 = phi double [ %121, %120 ], [ %117, %85 ]
  %123 = and i32 %122, 3
  %124 = uitofp i32 %123 to double
  %125 = fadd double %t.i.i.0, 0xBFE921FB54442D18
  %126 = fmul double %124, 0x3FF921FB54442D18
  %127 = fadd double %125, %126
  %128 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %129 = icmp eq i32 %128, 350
  br i1 %129, label %133, label %130

130:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i
  %131 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %132 = icmp eq i32 %131, 370
  br i1 %132, label %133, label %192

133:                                              ; preds = %130, %__internal_trig_reduction_kerneld.exit.i
  %134 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %135 = icmp eq i32 %134, 200
  br i1 %135, label %142, label %136

136:                                              ; preds = %133
  %137 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %138 = icmp eq i32 %137, 350
  br i1 %138, label %142, label %139

139:                                              ; preds = %136
  %140 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %141 = icmp eq i32 %140, 370
  br i1 %141, label %142, label %145

142:                                              ; preds = %139, %136, %133
  %143 = call double @llvm.fabs.f64(double %127)
  %144 = fcmp oeq double %143, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i

145:                                              ; preds = %139
  %146 = call i32 @llvm.nvvm.d2i.lo(double %127) #139
  %147 = call i32 @llvm.nvvm.d2i.hi(double %127) #139
  %148 = and i32 %147, 2147483647
  %149 = icmp eq i32 %148, 2146435072
  %150 = icmp eq i32 %146, 0
  %151 = select i1 %149, i1 %150, i1 false
  br label %__nv_isinfd.exit.i.i.i

__nv_isinfd.exit.i.i.i:                           ; preds = %145, %142
  %.02.in = phi i1 [ %144, %142 ], [ %151, %145 ]
  br i1 %.02.in, label %152, label %154

152:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %153 = fmul double %127, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

154:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %155 = fmul double %127, 0x3FE45F306DC9C883
  %156 = call i32 @llvm.nvvm.d2i.rn(double %155) #139
  store i32 %156, i32* %q.i.i.i.i, align 4, !noalias !68
  %157 = sitofp i32 %156 to double
  %158 = fneg double %157
  %159 = call double @llvm.fma.f64(double %158, double noundef 0x3FF921FB54442D18, double %127)
  %160 = call double @llvm.fma.f64(double %158, double noundef 0x3C91A62633145C00, double %159)
  %161 = call double @llvm.fma.f64(double %158, double noundef 0x397B839A252049C0, double %160)
  %162 = call double @llvm.fabs.f64(double %127)
  %163 = fcmp ult double %162, 0x41E0000000000000
  br i1 %163, label %__internal_trig_reduction_kerneld.exit.i.i.i, label %164

164:                                              ; preds = %154
  %165 = call fastcc double @__internal_trig_reduction_slowpathd(double %127, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i) #139
  %.pre8 = load i32, i32* %q.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i:     ; preds = %154, %164, %152
  %i.i.i.i.0 = phi i32 [ 0, %152 ], [ %.pre8, %164 ], [ %156, %154 ]
  %.01 = phi double [ %153, %152 ], [ %165, %164 ], [ %161, %154 ]
  %166 = fmul double %.01, %.01
  %167 = call double @llvm.fma.f64(double %166, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %168 = call double @llvm.fma.f64(double %167, double %166, double noundef 0xBE927E4F8E06E6D9)
  %169 = call double @llvm.fma.f64(double %168, double %166, double noundef 0x3EFA01A019DDBCE9)
  %170 = call double @llvm.fma.f64(double %169, double %166, double noundef 0xBF56C16C16C15D47)
  %171 = call double @llvm.fma.f64(double %170, double %166, double noundef 0x3FA5555555555551)
  %172 = call double @llvm.fma.f64(double %171, double %166, double noundef -5.000000e-01)
  %173 = call double @llvm.fma.f64(double %172, double %166, double noundef 1.000000e+00)
  %174 = call double @llvm.fma.f64(double %166, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %175 = call double @llvm.fma.f64(double %174, double %166, double noundef 0x3EC71DE369ACE392)
  %176 = call double @llvm.fma.f64(double %175, double %166, double noundef 0xBF2A01A019DB62A1)
  %177 = call double @llvm.fma.f64(double %176, double %166, double noundef 0x3F81111111110818)
  %178 = call double @llvm.fma.f64(double %177, double %166, double noundef 0xBFC5555555555554)
  %179 = call double @llvm.fma.f64(double %178, double %166, double noundef 0.000000e+00)
  %180 = call double @llvm.fma.f64(double %179, double %.01, double %.01)
  %181 = call i32 @llvm.nvvm.d2i.hi(double %180) #139
  %182 = call i32 @llvm.nvvm.d2i.lo(double %180) #139
  %183 = xor i32 %181, -2147483648
  %184 = call double @llvm.nvvm.lohi.i2d(i32 %182, i32 %183) #139
  %185 = and i32 %i.i.i.i.0, 1
  %.not5 = icmp eq i32 %185, 0
  %spec.select1 = select i1 %.not5, double %173, double %184
  %186 = and i32 %i.i.i.i.0, 2
  %.not6 = icmp eq i32 %186, 0
  br i1 %.not6, label %__internal_cos_offset.exit, label %187

187:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i
  %188 = call i32 @llvm.nvvm.d2i.hi(double %spec.select1) #139
  %189 = call i32 @llvm.nvvm.d2i.lo(double %spec.select1) #139
  %190 = xor i32 %188, -2147483648
  %191 = call double @llvm.nvvm.lohi.i2d(i32 %189, i32 %190) #139
  br label %__internal_cos_offset.exit

192:                                              ; preds = %130
  %193 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %194 = icmp eq i32 %193, 200
  br i1 %194, label %201, label %195

195:                                              ; preds = %192
  %196 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %197 = icmp eq i32 %196, 350
  br i1 %197, label %201, label %198

198:                                              ; preds = %195
  %199 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %200 = icmp eq i32 %199, 370
  br i1 %200, label %201, label %204

201:                                              ; preds = %198, %195, %192
  %202 = call double @llvm.fabs.f64(double %127)
  %203 = fcmp oeq double %202, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i

204:                                              ; preds = %198
  %205 = call i32 @llvm.nvvm.d2i.lo(double %127) #139
  %206 = call i32 @llvm.nvvm.d2i.hi(double %127) #139
  %207 = and i32 %206, 2147483647
  %208 = icmp eq i32 %207, 2146435072
  %209 = icmp eq i32 %205, 0
  %210 = select i1 %208, i1 %209, i1 false
  br label %__nv_isinfd.exit.i.i

__nv_isinfd.exit.i.i:                             ; preds = %204, %201
  %.03.in = phi i1 [ %203, %201 ], [ %210, %204 ]
  br i1 %.03.in, label %211, label %213

211:                                              ; preds = %__nv_isinfd.exit.i.i
  %212 = fmul double %127, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i

213:                                              ; preds = %__nv_isinfd.exit.i.i
  %214 = fmul double %127, 0x3FE45F306DC9C883
  %215 = call i32 @llvm.nvvm.d2i.rn(double %214) #139
  store i32 %215, i32* %q.i.i.i, align 4, !noalias !68
  %216 = sitofp i32 %215 to double
  %217 = fneg double %216
  %218 = call double @llvm.fma.f64(double %217, double noundef 0x3FF921FB54442D18, double %127)
  %219 = call double @llvm.fma.f64(double %217, double noundef 0x3C91A62633145C00, double %218)
  %220 = call double @llvm.fma.f64(double %217, double noundef 0x397B839A252049C0, double %219)
  %221 = call double @llvm.fabs.f64(double %127)
  %222 = fcmp ult double %221, 0x41E0000000000000
  br i1 %222, label %__internal_trig_reduction_kerneld.exit.i.i, label %223

223:                                              ; preds = %213
  %224 = call fastcc double @__internal_trig_reduction_slowpathd(double %127, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i) #139
  %.pre7 = load i32, i32* %q.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i

__internal_trig_reduction_kerneld.exit.i.i:       ; preds = %223, %213
  %225 = phi i32 [ %.pre7, %223 ], [ %215, %213 ]
  %t.i1.i.i.0 = phi double [ %224, %223 ], [ %220, %213 ]
  %phi.bo = add i32 %225, 1
  br label %__internal_sin_cos_kerneld.exit.i.i

__internal_sin_cos_kerneld.exit.i.i:              ; preds = %__internal_trig_reduction_kerneld.exit.i.i, %211
  %i.i.i.0 = phi i32 [ 1, %211 ], [ %phi.bo, %__internal_trig_reduction_kerneld.exit.i.i ]
  %z.i.i.0 = phi double [ %212, %211 ], [ %t.i1.i.i.0, %__internal_trig_reduction_kerneld.exit.i.i ]
  %226 = shl i32 %i.i.i.0, 3
  %227 = and i32 %226, 8
  %228 = zext i32 %227 to i64
  %229 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %228
  %230 = fmul double %z.i.i.0, %z.i.i.0
  %231 = and i32 %i.i.i.0, 1
  %.not = icmp eq i32 %231, 0
  %232 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %233 = getelementptr inbounds double, double addrspace(1)* %229, i64 1
  %234 = load double, double addrspace(1)* %233, align 8
  %235 = call double @llvm.fma.f64(double noundef %232, double %230, double %234)
  %236 = getelementptr inbounds double, double addrspace(1)* %229, i64 2
  %237 = load double, double addrspace(1)* %236, align 8
  %238 = call double @llvm.fma.f64(double %235, double %230, double %237)
  %239 = getelementptr inbounds double, double addrspace(1)* %229, i64 3
  %240 = load double, double addrspace(1)* %239, align 8
  %241 = call double @llvm.fma.f64(double %238, double %230, double %240)
  %242 = getelementptr inbounds double, double addrspace(1)* %229, i64 4
  %243 = load double, double addrspace(1)* %242, align 8
  %244 = call double @llvm.fma.f64(double %241, double %230, double %243)
  %245 = getelementptr inbounds double, double addrspace(1)* %229, i64 5
  %246 = load double, double addrspace(1)* %245, align 8
  %247 = call double @llvm.fma.f64(double %244, double %230, double %246)
  %248 = getelementptr inbounds double, double addrspace(1)* %229, i64 6
  %249 = load double, double addrspace(1)* %248, align 8
  %250 = call double @llvm.fma.f64(double %247, double %230, double %249)
  %251 = call double @llvm.fma.f64(double %250, double %z.i.i.0, double %z.i.i.0)
  %252 = call double @llvm.fma.f64(double %250, double %230, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not, double %251, double %252
  %253 = and i32 %i.i.i.0, 2
  %.not4 = icmp eq i32 %253, 0
  %254 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not4, double %spec.select2, double %254
  br label %__internal_cos_offset.exit

__internal_cos_offset.exit:                       ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i, %187, %__internal_sin_cos_kerneld.exit.i.i
  %.0 = phi double [ %.1, %__internal_sin_cos_kerneld.exit.i.i ], [ %191, %187 ], [ %spec.select1, %__internal_trig_reduction_kerneld.exit.i.i.i ]
  %255 = fmul double %110, %.0
  br label %256

256:                                              ; preds = %25, %__internal_cos_offset.exit, %__nv_isinfd.exit, %47, %3
  %r.3 = phi double [ %22, %3 ], [ %44, %25 ], [ %66, %47 ], [ %255, %__internal_cos_offset.exit ], [ 0.000000e+00, %__nv_isinfd.exit ]
  ret double %r.3
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_j1(double %a) local_unnamed_addr #107 {
  %q.i.i.i = alloca i32, align 4
  %q.i.i.i.i = alloca i32, align 4
  %q.i.i = alloca i32, align 4
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ugt double %1, 0x400353AABAD7B784
  br i1 %2, label %20, label %3

3:                                                ; preds = %0
  %4 = call double @llvm.fma.f64(double %1, double noundef 0x3D020E4ADCDE2AD3, double noundef 0xBD4DD167A0DC3F55) #140
  %5 = call double @llvm.fma.f64(double %4, double %1, double noundef 0x3D5503F5A491E487) #140
  %6 = call double @llvm.fma.f64(double %5, double %1, double noundef 0x3DC1F29940C2403A) #140
  %7 = call double @llvm.fma.f64(double %6, double %1, double noundef 0x3D84CF9302EACDEF) #140
  %8 = call double @llvm.fma.f64(double %7, double %1, double noundef 0xBE384A53DBBCA436) #140
  %9 = call double @llvm.fma.f64(double %8, double %1, double noundef 0x3D9779BEE4F63BCC) #140
  %10 = call double @llvm.fma.f64(double %9, double %1, double noundef 0x3EA6C160E414F3F0) #140
  %11 = call double @llvm.fma.f64(double %10, double %1, double noundef 0x3D8F3D2F12430699) #140
  %12 = call double @llvm.fma.f64(double %11, double %1, double noundef 0xBF0C71C72C0CED04) #140
  %13 = call double @llvm.fma.f64(double %12, double %1, double noundef 0x3D659BCA506F1128) #140
  %14 = call double @llvm.fma.f64(double %13, double %1, double noundef 0x3F65555555506982) #140
  %15 = call double @llvm.fma.f64(double %14, double %1, double noundef 0x3D15BA0B425F1BFB) #140
  %16 = call double @llvm.fma.f64(double %15, double %1, double noundef 0xBFB0000000000065) #140
  %17 = call double @llvm.fma.f64(double %16, double %1, double noundef 0x3C8729A7253FB679) #140
  %18 = call double @llvm.fma.f64(double %17, double %1, double noundef 5.000000e-01) #140
  %19 = fmul double %1, %18
  br label %252

20:                                               ; preds = %0
  %21 = fcmp ugt double %1, 0x4015B1D0574614EA
  br i1 %21, label %41, label %22

22:                                               ; preds = %20
  %23 = fadd double %1, 0xC00EA75575AF6F09
  %24 = fadd double %23, 0x3CA60155A9D1B256
  %25 = call double @llvm.fma.f64(double %24, double noundef 0xBCF8D3CDBB60175E, double noundef 0x3D41011A1DF02DAD) #140
  %26 = call double @llvm.fma.f64(double %25, double %24, double noundef 0x3D76013AC1E5E222) #140
  %27 = call double @llvm.fma.f64(double %26, double %24, double noundef 0xBDBEC315D96D5F03) #140
  %28 = call double @llvm.fma.f64(double %27, double %24, double noundef 0xBDF03BE1B4B57207) #140
  %29 = call double @llvm.fma.f64(double %28, double %24, double noundef 0x3E345695F8B660F7) #140
  %30 = call double @llvm.fma.f64(double %29, double %24, double noundef 0x3E617069FCFCFFF4) #140
  %31 = call double @llvm.fma.f64(double %30, double %24, double noundef 0xBEA33825C36745EB) #140
  %32 = call double @llvm.fma.f64(double %31, double %24, double noundef 0xBEC9799D4F90931B) #140
  %33 = call double @llvm.fma.f64(double %32, double %24, double noundef 0x3F083A06E2F7DF13) #140
  %34 = call double @llvm.fma.f64(double %33, double %24, double noundef 0x3F26E4C2D53A7CF6) #140
  %35 = call double @llvm.fma.f64(double %34, double %24, double noundef 0xBF624B3409957B1C) #140
  %36 = call double @llvm.fma.f64(double %35, double %24, double noundef 0xBF7537544C3325DF) #140
  %37 = call double @llvm.fma.f64(double %36, double %24, double noundef 0x3FAB589D1DA138E2) #140
  %38 = call double @llvm.fma.f64(double %37, double %24, double noundef 0x3FAAE8A39F51AD13) #140
  %39 = call double @llvm.fma.f64(double %38, double %24, double noundef 0xBFD9C6CF582CBF7F) #140
  %40 = fmul double %24, %39
  br label %252

41:                                               ; preds = %20
  %42 = fcmp ugt double %1, 0x40213065E54C1AA9
  br i1 %42, label %62, label %43

43:                                               ; preds = %41
  %44 = fadd double %1, 0xC01C0FF5F3B47250
  %45 = fadd double %44, 0x3C9B226D9D243827
  %46 = call double @llvm.fma.f64(double %45, double noundef 0x3CF3EB867515FAD6, double noundef 0xBD40E8363DB649A9) #140
  %47 = call double @llvm.fma.f64(double %46, double %45, double noundef 0xBD73B7DD4A6608FB) #140
  %48 = call double @llvm.fma.f64(double %47, double %45, double noundef 0x3DBEC5E01482C750) #140
  %49 = call double @llvm.fma.f64(double %48, double %45, double noundef 0x3DEC62BB9E882103) #140
  %50 = call double @llvm.fma.f64(double %49, double %45, double noundef 0xBE34462EED732A23) #140
  %51 = call double @llvm.fma.f64(double %50, double %45, double noundef 0xBE5D48DCAD7DC59B) #140
  %52 = call double @llvm.fma.f64(double %51, double %45, double noundef 0x3EA3026DF29167E9) #140
  %53 = call double @llvm.fma.f64(double %52, double %45, double noundef 0x3EC4255B0119666C) #140
  %54 = call double @llvm.fma.f64(double %53, double %45, double noundef 0xBF0796A751B32693) #140
  %55 = call double @llvm.fma.f64(double %54, double %45, double noundef 0xBF207358BBDBA284) #140
  %56 = call double @llvm.fma.f64(double %55, double %45, double noundef 0x3F613FBC7D6927B1) #140
  %57 = call double @llvm.fma.f64(double %56, double %45, double noundef 0x3F69A4B292E3DD75) #140
  %58 = call double @llvm.fma.f64(double %57, double %45, double noundef 0xBFA80C83BDEEE4FB) #140
  %59 = call double @llvm.fma.f64(double %58, double %45, double noundef 0xBF95E70DC60362BF) #140
  %60 = call double @llvm.fma.f64(double %59, double %45, double noundef 0x3FD33518B3874E8A) #140
  %61 = fmul double %45, %60
  br label %252

62:                                               ; preds = %41
  %63 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %64 = icmp eq i32 %63, 200
  br i1 %64, label %71, label %65

65:                                               ; preds = %62
  %66 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %67 = icmp eq i32 %66, 350
  br i1 %67, label %71, label %68

68:                                               ; preds = %65
  %69 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %70 = icmp eq i32 %69, 370
  br i1 %70, label %71, label %73

71:                                               ; preds = %68, %65, %62
  %72 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit

73:                                               ; preds = %68
  %74 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %75 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %76 = and i32 %75, 2147483647
  %77 = icmp eq i32 %76, 2146435072
  %78 = icmp eq i32 %74, 0
  %79 = select i1 %77, i1 %78, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %73, %71
  %.05.in = phi i1 [ %72, %71 ], [ %79, %73 ]
  br i1 %.05.in, label %252, label %80

80:                                               ; preds = %__nv_isinfd.exit
  %81 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %82 = fneg double %1
  %83 = call double @llvm.fma.f64(double %82, double %81, double noundef 1.000000e+00)
  %84 = call double @llvm.fma.f64(double %83, double %83, double %83)
  %85 = call double @llvm.fma.f64(double %84, double %81, double %81)
  %86 = fmul double %85, %85
  %87 = call double @llvm.fma.f64(double %86, double noundef 0x40CD02EA3F2F6751, double noundef 0xC099C06322A3F8BE)
  %88 = call double @llvm.fma.f64(double %87, double %86, double noundef 0x405B89354DA77324)
  %89 = call double @llvm.fma.f64(double %88, double %86, double noundef 0xC01E352294653188)
  %90 = call double @llvm.fma.f64(double %89, double %86, double noundef 0x3FE9BC7DB16BD7A7)
  %91 = call double @llvm.fma.f64(double %90, double %86, double noundef 0xBFC8BFE1C3A4F741)
  %92 = call double @llvm.fma.f64(double %91, double %86, double noundef 0x3FC7FFFFF0D00BE2)
  %93 = call double @llvm.fma.f64(double %92, double %86, double noundef 0x3FF00000000068CC)
  %94 = call double @llvm.fma.f64(double %86, double noundef 0xC18DA26B212FDC9A, double noundef 0x415A30AC6857BEE0)
  %95 = call double @llvm.fma.f64(double %94, double %86, double noundef 0xC11764222AD7C910)
  %96 = call double @llvm.fma.f64(double %95, double %86, double noundef 0x40CEB02E0C306857)
  %97 = call double @llvm.fma.f64(double %96, double %86, double noundef 0xC08351859FA2B23B)
  %98 = call double @llvm.fma.f64(double %97, double %86, double noundef 0x403E65A07AF51F42)
  %99 = call double @llvm.fma.f64(double %98, double %86, double noundef 0xC002F2B817F77A57)
  %100 = call double @llvm.fma.f64(double %99, double %86, double noundef 0x3FD7BCC34DA069FD)
  %101 = call double @llvm.fma.f64(double %100, double %86, double noundef 0xBFC4FFFFF8A44463)
  %102 = call double @llvm.fma.f64(double %101, double %86, double noundef 0x3FD7FFFFFFFF5CD7)
  %103 = call double @llvm.fma.f64(double %102, double %85, double %1)
  %104 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %105 = fmul double %104, 0x3FE9884533D43651
  %106 = fmul double %105, %93
  %107 = fmul double %103, 0x3FE45F306DC9C883
  %108 = call i32 @llvm.nvvm.d2i.rn(double %107) #139
  store i32 %108, i32* %q.i.i, align 4, !noalias !68
  %109 = sitofp i32 %108 to double
  %110 = fneg double %109
  %111 = call double @llvm.fma.f64(double %110, double noundef 0x3FF921FB54442D18, double %103)
  %112 = call double @llvm.fma.f64(double %110, double noundef 0x3C91A62633145C00, double %111)
  %113 = call double @llvm.fma.f64(double %110, double noundef 0x397B839A252049C0, double %112)
  %114 = call double @llvm.fabs.f64(double %103)
  %115 = fcmp ult double %114, 0x41E0000000000000
  br i1 %115, label %__internal_trig_reduction_kerneld.exit.i, label %116

116:                                              ; preds = %80
  %117 = call fastcc double @__internal_trig_reduction_slowpathd(double %103, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i) #139
  %.pre = load i32, i32* %q.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i

__internal_trig_reduction_kerneld.exit.i:         ; preds = %116, %80
  %118 = phi i32 [ %.pre, %116 ], [ %108, %80 ]
  %t.i.i.0 = phi double [ %117, %116 ], [ %113, %80 ]
  %119 = and i32 %118, 3
  %120 = uitofp i32 %119 to double
  %121 = fadd double %t.i.i.0, 0xC002D97C7F3321D2
  %122 = fmul double %120, 0x3FF921FB54442D18
  %123 = fadd double %121, %122
  %124 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %125 = icmp eq i32 %124, 350
  br i1 %125, label %129, label %126

126:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i
  %127 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %128 = icmp eq i32 %127, 370
  br i1 %128, label %129, label %188

129:                                              ; preds = %126, %__internal_trig_reduction_kerneld.exit.i
  %130 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %131 = icmp eq i32 %130, 200
  br i1 %131, label %138, label %132

132:                                              ; preds = %129
  %133 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %134 = icmp eq i32 %133, 350
  br i1 %134, label %138, label %135

135:                                              ; preds = %132
  %136 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %137 = icmp eq i32 %136, 370
  br i1 %137, label %138, label %141

138:                                              ; preds = %135, %132, %129
  %139 = call double @llvm.fabs.f64(double %123)
  %140 = fcmp oeq double %139, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i

141:                                              ; preds = %135
  %142 = call i32 @llvm.nvvm.d2i.lo(double %123) #139
  %143 = call i32 @llvm.nvvm.d2i.hi(double %123) #139
  %144 = and i32 %143, 2147483647
  %145 = icmp eq i32 %144, 2146435072
  %146 = icmp eq i32 %142, 0
  %147 = select i1 %145, i1 %146, i1 false
  br label %__nv_isinfd.exit.i.i.i

__nv_isinfd.exit.i.i.i:                           ; preds = %141, %138
  %.02.in = phi i1 [ %140, %138 ], [ %147, %141 ]
  br i1 %.02.in, label %148, label %150

148:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %149 = fmul double %123, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

150:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %151 = fmul double %123, 0x3FE45F306DC9C883
  %152 = call i32 @llvm.nvvm.d2i.rn(double %151) #139
  store i32 %152, i32* %q.i.i.i.i, align 4, !noalias !68
  %153 = sitofp i32 %152 to double
  %154 = fneg double %153
  %155 = call double @llvm.fma.f64(double %154, double noundef 0x3FF921FB54442D18, double %123)
  %156 = call double @llvm.fma.f64(double %154, double noundef 0x3C91A62633145C00, double %155)
  %157 = call double @llvm.fma.f64(double %154, double noundef 0x397B839A252049C0, double %156)
  %158 = call double @llvm.fabs.f64(double %123)
  %159 = fcmp ult double %158, 0x41E0000000000000
  br i1 %159, label %__internal_trig_reduction_kerneld.exit.i.i.i, label %160

160:                                              ; preds = %150
  %161 = call fastcc double @__internal_trig_reduction_slowpathd(double %123, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i) #139
  %.pre8 = load i32, i32* %q.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i:     ; preds = %150, %160, %148
  %i.i.i.i.0 = phi i32 [ 0, %148 ], [ %.pre8, %160 ], [ %152, %150 ]
  %.01 = phi double [ %149, %148 ], [ %161, %160 ], [ %157, %150 ]
  %162 = fmul double %.01, %.01
  %163 = call double @llvm.fma.f64(double %162, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %164 = call double @llvm.fma.f64(double %163, double %162, double noundef 0xBE927E4F8E06E6D9)
  %165 = call double @llvm.fma.f64(double %164, double %162, double noundef 0x3EFA01A019DDBCE9)
  %166 = call double @llvm.fma.f64(double %165, double %162, double noundef 0xBF56C16C16C15D47)
  %167 = call double @llvm.fma.f64(double %166, double %162, double noundef 0x3FA5555555555551)
  %168 = call double @llvm.fma.f64(double %167, double %162, double noundef -5.000000e-01)
  %169 = call double @llvm.fma.f64(double %168, double %162, double noundef 1.000000e+00)
  %170 = call double @llvm.fma.f64(double %162, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %171 = call double @llvm.fma.f64(double %170, double %162, double noundef 0x3EC71DE369ACE392)
  %172 = call double @llvm.fma.f64(double %171, double %162, double noundef 0xBF2A01A019DB62A1)
  %173 = call double @llvm.fma.f64(double %172, double %162, double noundef 0x3F81111111110818)
  %174 = call double @llvm.fma.f64(double %173, double %162, double noundef 0xBFC5555555555554)
  %175 = call double @llvm.fma.f64(double %174, double %162, double noundef 0.000000e+00)
  %176 = call double @llvm.fma.f64(double %175, double %.01, double %.01)
  %177 = call i32 @llvm.nvvm.d2i.hi(double %176) #139
  %178 = call i32 @llvm.nvvm.d2i.lo(double %176) #139
  %179 = xor i32 %177, -2147483648
  %180 = call double @llvm.nvvm.lohi.i2d(i32 %178, i32 %179) #139
  %181 = and i32 %i.i.i.i.0, 1
  %.not5 = icmp eq i32 %181, 0
  %spec.select1 = select i1 %.not5, double %169, double %180
  %182 = and i32 %i.i.i.i.0, 2
  %.not6 = icmp eq i32 %182, 0
  br i1 %.not6, label %__internal_cos_offset.exit, label %183

183:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i
  %184 = call i32 @llvm.nvvm.d2i.hi(double %spec.select1) #139
  %185 = call i32 @llvm.nvvm.d2i.lo(double %spec.select1) #139
  %186 = xor i32 %184, -2147483648
  %187 = call double @llvm.nvvm.lohi.i2d(i32 %185, i32 %186) #139
  br label %__internal_cos_offset.exit

188:                                              ; preds = %126
  %189 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %190 = icmp eq i32 %189, 200
  br i1 %190, label %197, label %191

191:                                              ; preds = %188
  %192 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %193 = icmp eq i32 %192, 350
  br i1 %193, label %197, label %194

194:                                              ; preds = %191
  %195 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %196 = icmp eq i32 %195, 370
  br i1 %196, label %197, label %200

197:                                              ; preds = %194, %191, %188
  %198 = call double @llvm.fabs.f64(double %123)
  %199 = fcmp oeq double %198, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i

200:                                              ; preds = %194
  %201 = call i32 @llvm.nvvm.d2i.lo(double %123) #139
  %202 = call i32 @llvm.nvvm.d2i.hi(double %123) #139
  %203 = and i32 %202, 2147483647
  %204 = icmp eq i32 %203, 2146435072
  %205 = icmp eq i32 %201, 0
  %206 = select i1 %204, i1 %205, i1 false
  br label %__nv_isinfd.exit.i.i

__nv_isinfd.exit.i.i:                             ; preds = %200, %197
  %.03.in = phi i1 [ %199, %197 ], [ %206, %200 ]
  br i1 %.03.in, label %207, label %209

207:                                              ; preds = %__nv_isinfd.exit.i.i
  %208 = fmul double %123, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i

209:                                              ; preds = %__nv_isinfd.exit.i.i
  %210 = fmul double %123, 0x3FE45F306DC9C883
  %211 = call i32 @llvm.nvvm.d2i.rn(double %210) #139
  store i32 %211, i32* %q.i.i.i, align 4, !noalias !68
  %212 = sitofp i32 %211 to double
  %213 = fneg double %212
  %214 = call double @llvm.fma.f64(double %213, double noundef 0x3FF921FB54442D18, double %123)
  %215 = call double @llvm.fma.f64(double %213, double noundef 0x3C91A62633145C00, double %214)
  %216 = call double @llvm.fma.f64(double %213, double noundef 0x397B839A252049C0, double %215)
  %217 = call double @llvm.fabs.f64(double %123)
  %218 = fcmp ult double %217, 0x41E0000000000000
  br i1 %218, label %__internal_trig_reduction_kerneld.exit.i.i, label %219

219:                                              ; preds = %209
  %220 = call fastcc double @__internal_trig_reduction_slowpathd(double %123, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i) #139
  %.pre7 = load i32, i32* %q.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i

__internal_trig_reduction_kerneld.exit.i.i:       ; preds = %219, %209
  %221 = phi i32 [ %.pre7, %219 ], [ %211, %209 ]
  %t.i1.i.i.0 = phi double [ %220, %219 ], [ %216, %209 ]
  %phi.bo = add i32 %221, 1
  br label %__internal_sin_cos_kerneld.exit.i.i

__internal_sin_cos_kerneld.exit.i.i:              ; preds = %__internal_trig_reduction_kerneld.exit.i.i, %207
  %i.i.i.0 = phi i32 [ 1, %207 ], [ %phi.bo, %__internal_trig_reduction_kerneld.exit.i.i ]
  %z.i.i.0 = phi double [ %208, %207 ], [ %t.i1.i.i.0, %__internal_trig_reduction_kerneld.exit.i.i ]
  %222 = shl i32 %i.i.i.0, 3
  %223 = and i32 %222, 8
  %224 = zext i32 %223 to i64
  %225 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %224
  %226 = fmul double %z.i.i.0, %z.i.i.0
  %227 = and i32 %i.i.i.0, 1
  %.not = icmp eq i32 %227, 0
  %228 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %229 = getelementptr inbounds double, double addrspace(1)* %225, i64 1
  %230 = load double, double addrspace(1)* %229, align 8
  %231 = call double @llvm.fma.f64(double noundef %228, double %226, double %230)
  %232 = getelementptr inbounds double, double addrspace(1)* %225, i64 2
  %233 = load double, double addrspace(1)* %232, align 8
  %234 = call double @llvm.fma.f64(double %231, double %226, double %233)
  %235 = getelementptr inbounds double, double addrspace(1)* %225, i64 3
  %236 = load double, double addrspace(1)* %235, align 8
  %237 = call double @llvm.fma.f64(double %234, double %226, double %236)
  %238 = getelementptr inbounds double, double addrspace(1)* %225, i64 4
  %239 = load double, double addrspace(1)* %238, align 8
  %240 = call double @llvm.fma.f64(double %237, double %226, double %239)
  %241 = getelementptr inbounds double, double addrspace(1)* %225, i64 5
  %242 = load double, double addrspace(1)* %241, align 8
  %243 = call double @llvm.fma.f64(double %240, double %226, double %242)
  %244 = getelementptr inbounds double, double addrspace(1)* %225, i64 6
  %245 = load double, double addrspace(1)* %244, align 8
  %246 = call double @llvm.fma.f64(double %243, double %226, double %245)
  %247 = call double @llvm.fma.f64(double %246, double %z.i.i.0, double %z.i.i.0)
  %248 = call double @llvm.fma.f64(double %246, double %226, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not, double %247, double %248
  %249 = and i32 %i.i.i.0, 2
  %.not4 = icmp eq i32 %249, 0
  %250 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not4, double %spec.select2, double %250
  br label %__internal_cos_offset.exit

__internal_cos_offset.exit:                       ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i, %183, %__internal_sin_cos_kerneld.exit.i.i
  %.0 = phi double [ %.1, %__internal_sin_cos_kerneld.exit.i.i ], [ %187, %183 ], [ %spec.select1, %__internal_trig_reduction_kerneld.exit.i.i.i ]
  %251 = fmul double %106, %.0
  br label %252

252:                                              ; preds = %22, %__internal_cos_offset.exit, %__nv_isinfd.exit, %43, %3
  %r.3 = phi double [ %19, %3 ], [ %40, %22 ], [ %61, %43 ], [ %251, %__internal_cos_offset.exit ], [ 0.000000e+00, %__nv_isinfd.exit ]
  %253 = fcmp olt double %a, 0.000000e+00
  %254 = fneg double %r.3
  %r.4 = select i1 %253, double %254, double %r.3
  %255 = fcmp olt double %1, 1.000000e-30
  %256 = fmul double %a, 5.000000e-01
  %r.5 = select i1 %255, double %256, double %r.4
  ret double %r.5
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_y0(double %a) local_unnamed_addr #108 {
  %q.i.i.i.i9 = alloca i32, align 4
  %q.i.i.i.i.i = alloca i32, align 4
  %q.i.i.i20 = alloca i32, align 4
  %q.i.i.i = alloca i32, align 4
  %q.i.i.i.i = alloca i32, align 4
  %q.i.i = alloca i32, align 4
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ugt double %1, 0x3FE97F4A8F9D3F28
  br i1 %2, label %335, label %3

3:                                                ; preds = %0
  %4 = fmul double %a, %a
  %5 = call double @llvm.fma.f64(double %4, double noundef 0x3C8EFBD0A1B77C65, double noundef 0xBD13098C51C18514) #140
  %6 = call double @llvm.fma.f64(double %5, double %4, double noundef 0x3D923102D2F5F2F5) #140
  %7 = call double @llvm.fma.f64(double %6, double %4, double noundef 0xBE0A5F2DEE7D526E) #140
  %8 = call double @llvm.fma.f64(double %7, double %4, double noundef 0x3E7BB77E758B38AF) #140
  %9 = call double @llvm.fma.f64(double %8, double %4, double noundef 0xBEE3D1A206EC4F36) #140
  %10 = call double @llvm.fma.f64(double %9, double %4, double noundef 0x3F4183DCD3ED6294) #140
  %11 = call double @llvm.fma.f64(double %10, double %4, double noundef 0xBF903921CF04F123) #140
  %12 = call double @llvm.fma.f64(double %11, double %4, double noundef 0x3FC5DB69D7753176) #140
  %13 = fadd double %4, 0xBFDBA96740000000
  %14 = fadd double %13, 0x3E15A30C80000000
  %15 = fmul double %14, %12
  %16 = call i32 @llvm.nvvm.d2i.hi(double %1) #142
  %17 = call i32 @llvm.nvvm.d2i.lo(double %1) #142
  %18 = icmp slt i32 %16, 1048576
  br i1 %18, label %19, label %23

19:                                               ; preds = %3
  %20 = fmul double %1, 0x4350000000000000
  %21 = call i32 @llvm.nvvm.d2i.hi(double %20) #142
  %22 = call i32 @llvm.nvvm.d2i.lo(double %20) #142
  br label %23

23:                                               ; preds = %19, %3
  %.05 = phi double [ %20, %19 ], [ %1, %3 ]
  %ihi.i.0 = phi i32 [ %21, %19 ], [ %16, %3 ]
  %ilo.i.0 = phi i32 [ %22, %19 ], [ %17, %3 ]
  %e.i.0 = phi i32 [ -1077, %19 ], [ -1023, %3 ]
  %24 = add i32 %ihi.i.0, -1
  %25 = icmp ult i32 %24, 2146435071
  br i1 %25, label %26, label %75

26:                                               ; preds = %23
  %27 = lshr i32 %ihi.i.0, 20
  %28 = add nsw i32 %e.i.0, %27
  %29 = and i32 %ihi.i.0, -2146435073
  %30 = or i32 %29, 1072693248
  %31 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %30) #142
  %32 = icmp sgt i32 %30, 1073127582
  br i1 %32, label %33, label %39

33:                                               ; preds = %26
  %34 = call i32 @llvm.nvvm.d2i.lo(double %31) #142
  %35 = call i32 @llvm.nvvm.d2i.hi(double %31) #142
  %36 = add i32 %35, -1048576
  %37 = call double @llvm.nvvm.lohi.i2d(i32 %34, i32 %36) #142
  %38 = add nsw i32 %28, 1
  br label %39

39:                                               ; preds = %33, %26
  %m.i.0 = phi double [ %37, %33 ], [ %31, %26 ]
  %e.i.1 = phi i32 [ %38, %33 ], [ %28, %26 ]
  %40 = fadd double %m.i.0, -1.000000e+00
  %41 = fadd double %m.i.0, 1.000000e+00
  %42 = call double @llvm.nvvm.rcp.approx.ftz.d(double %41) #142
  %43 = fneg double %41
  %44 = call double @llvm.fma.f64(double %43, double %42, double noundef 1.000000e+00) #140
  %45 = call double @llvm.fma.f64(double %44, double %44, double %44) #140
  %46 = call double @llvm.fma.f64(double %45, double %42, double %42) #140
  %47 = fmul double %40, %46
  %48 = fadd double %47, %47
  %49 = fmul double %48, %48
  %50 = call double @llvm.fma.f64(double %49, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %51 = call double @llvm.fma.f64(double %50, double %49, double noundef 0x3EF3B2669F02676F) #140
  %52 = call double @llvm.fma.f64(double %51, double %49, double noundef 0x3F1745CBA9AB0956) #140
  %53 = call double @llvm.fma.f64(double %52, double %49, double noundef 0x3F3C71C72D1B5154) #140
  %54 = call double @llvm.fma.f64(double %53, double %49, double noundef 0x3F624924923BE72D) #140
  %55 = call double @llvm.fma.f64(double %54, double %49, double noundef 0x3F8999999999A3C4) #140
  %56 = call double @llvm.fma.f64(double %55, double %49, double noundef 0x3FB5555555555554) #140
  %57 = fsub double %40, %48
  %58 = fmul double %57, 2.000000e+00
  %59 = fneg double %48
  %60 = call double @llvm.fma.f64(double %59, double %40, double %58) #140
  %61 = fmul double %46, %60
  %62 = fmul double %49, %56
  %63 = call double @llvm.fma.f64(double %62, double %48, double %61) #140
  %64 = xor i32 %e.i.1, -2147483648
  %65 = call double @llvm.nvvm.lohi.i2d(i32 %64, i32 noundef 1127219200) #142
  %66 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %67 = fsub double %65, %66
  %68 = call double @llvm.fma.f64(double %67, double noundef 0x3FE62E42FEFA39EF, double %48) #140
  %69 = fneg double %67
  %70 = call double @llvm.fma.f64(double %69, double noundef 0x3FE62E42FEFA39EF, double %68) #140
  %71 = fsub double %70, %48
  %72 = fsub double %63, %71
  %73 = call double @llvm.fma.f64(double %67, double noundef 0x3C7ABC9E3B39803F, double %72) #140
  %74 = fadd double %68, %73
  br label %__nv_log.exit

75:                                               ; preds = %23
  %76 = call double @llvm.fma.f64(double %.05, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %77 = call i32 @llvm.nvvm.d2i.hi(double %.05) #142
  %78 = bitcast i32 %77 to float
  %79 = fcmp oeq float %78, 0.000000e+00
  %q.i3.0 = select i1 %79, double 0xFFF0000000000000, double %76
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %75, %39
  %q.i3.1 = phi double [ %74, %39 ], [ %q.i3.0, %75 ]
  %80 = fmul double %q.i3.1, 0x3FE45F306DC9C883
  %81 = fcmp ugt double %1, 0x400FB319F277BBE5
  br i1 %81, label %102, label %82

82:                                               ; preds = %__nv_log.exit
  %83 = fadd double %1, 0xC0033D152E971B40
  %84 = fadd double %83, 0x3CA0F539D7DA258E
  %85 = call double @llvm.fma.f64(double %84, double noundef 0xBCC0D18564C48C61, double noundef 0xBCFCF8F9A8C294BC) #140
  %86 = call double @llvm.fma.f64(double %85, double %84, double noundef 0x3D3FAB983CAE498B) #140
  %87 = call double @llvm.fma.f64(double %86, double %84, double noundef 0x3D7CD7C018579B88) #140
  %88 = call double @llvm.fma.f64(double %87, double %84, double noundef 0xBDBBDD2342D64FDD) #140
  %89 = call double @llvm.fma.f64(double %88, double %84, double noundef 0xBDF5C2D9416B1E2B) #140
  %90 = call double @llvm.fma.f64(double %89, double %84, double noundef 0x3E32951D73174DD5) #140
  %91 = call double @llvm.fma.f64(double %90, double %84, double noundef 0x3E67FF99802CAEB5) #140
  %92 = call double @llvm.fma.f64(double %91, double %84, double noundef 0xBEA1CCE305C4C9F7) #140
  %93 = call double @llvm.fma.f64(double %92, double %84, double noundef 0xBED232C77E29E1BB) #140
  %94 = call double @llvm.fma.f64(double %93, double %84, double noundef 0x3F06ED3B9F0EF757) #140
  %95 = call double @llvm.fma.f64(double %94, double %84, double noundef 0x3F315382BA096A62) #140
  %96 = call double @llvm.fma.f64(double %95, double %84, double noundef 0xBF61F992590D1AE4) #140
  %97 = call double @llvm.fma.f64(double %96, double %84, double noundef 0xBF81BB1CBE1A465F) #140
  %98 = call double @llvm.fma.f64(double %97, double %84, double noundef 0x3FACFAE864368D84) #140
  %99 = call double @llvm.fma.f64(double %98, double %84, double noundef 0x3FBBA1DEEA0294A3) #140
  %100 = call double @llvm.fma.f64(double %99, double %84, double noundef 0xBFE09CDB36551280) #140
  %101 = fmul double %84, %100
  br label %__nv_j0.exit

102:                                              ; preds = %__nv_log.exit
  %103 = fcmp ugt double %1, 0x401C58FD1A62F5EC
  br i1 %103, label %124, label %104

104:                                              ; preds = %102
  %105 = fadd double %1, 0xC016148F5B2C2E45
  %106 = fadd double %105, 0xBC975054CD60A517
  %107 = call double @llvm.fma.f64(double %106, double noundef 0x3CBCB0A8F126B343, double noundef 0x3CF83FD1F333EB61) #140
  %108 = call double @llvm.fma.f64(double %107, double %106, double noundef 0xBD4100E33E3FB413) #140
  %109 = call double @llvm.fma.f64(double %108, double %106, double noundef 0xBD7846076D004627) #140
  %110 = call double @llvm.fma.f64(double %109, double %106, double noundef 0x3DBE2F1D4F90720D) #140
  %111 = call double @llvm.fma.f64(double %110, double %106, double noundef 0x3DF1D03B1E4A119B) #140
  %112 = call double @llvm.fma.f64(double %111, double %106, double noundef 0xBE341D72B1B3BCE9) #140
  %113 = call double @llvm.fma.f64(double %112, double %106, double noundef 0xBE62DA37CE2A9EF8) #140
  %114 = call double @llvm.fma.f64(double %113, double %106, double noundef 0x3EA32E6D9974F763) #140
  %115 = call double @llvm.fma.f64(double %114, double %106, double noundef 0x3ECAD77D744A1879) #140
  %116 = call double @llvm.fma.f64(double %115, double %106, double noundef 0xBF0863F481A37337) #140
  %117 = call double @llvm.fma.f64(double %116, double %106, double noundef 0xBF26F641F418F0F4) #140
  %118 = call double @llvm.fma.f64(double %117, double %106, double noundef 0x3F627E31FE9A969E) #140
  %119 = call double @llvm.fma.f64(double %118, double %106, double noundef 0x3F72F7FFE9025628) #140
  %120 = call double @llvm.fma.f64(double %119, double %106, double noundef 0xBFAB2150CB41E8BF) #140
  %121 = call double @llvm.fma.f64(double %120, double %106, double noundef 0xBF9F8F72E7A848DE) #140
  %122 = call double @llvm.fma.f64(double %121, double %106, double noundef 0x3FD5C6E60A097823) #140
  %123 = fmul double %106, %122
  br label %__nv_j0.exit

124:                                              ; preds = %102
  %125 = fcmp ugt double %1, 0x402471FCB6A7A8C0
  br i1 %125, label %146, label %126

126:                                              ; preds = %124
  %127 = fadd double %1, 0xC0214EB56CCCDECA
  %128 = fadd double %127, 0x3CB51970714C7C25
  %129 = call double @llvm.fma.f64(double %128, double noundef 0xBCBDB7FFCF659E24, double noundef 0xBCF4B3A71AAAC629) #140
  %130 = call double @llvm.fma.f64(double %129, double %128, double noundef 0x3D417EC150ECDCE7) #140
  %131 = call double @llvm.fma.f64(double %130, double %128, double noundef 0x3D7438F5EA1D10B2) #140
  %132 = call double @llvm.fma.f64(double %131, double %128, double noundef 0xBDBEDAE7EC2C9E87) #140
  %133 = call double @llvm.fma.f64(double %132, double %128, double noundef 0xBDECADD2C4B91F58) #140
  %134 = call double @llvm.fma.f64(double %133, double %128, double noundef 0x3E34582C8EE12204) #140
  %135 = call double @llvm.fma.f64(double %134, double %128, double noundef 0x3E5CEDA451DD20F8) #140
  %136 = call double @llvm.fma.f64(double %135, double %128, double noundef 0xBEA30E8CC3165E2F) #140
  %137 = call double @llvm.fma.f64(double %136, double %128, double noundef 0xBEC3324842BB1A2E) #140
  %138 = call double @llvm.fma.f64(double %137, double %128, double noundef 0x3F07800BC54FBDDB) #140
  %139 = call double @llvm.fma.f64(double %138, double %128, double noundef 0x3F1D79605276949A) #140
  %140 = call double @llvm.fma.f64(double %139, double %128, double noundef 0xBF60E0D60385A629) #140
  %141 = call double @llvm.fma.f64(double %140, double %128, double noundef 0xBF648E63600D82F3) #140
  %142 = call double @llvm.fma.f64(double %141, double %128, double noundef 0x3FA68B984EC6493A) #140
  %143 = call double @llvm.fma.f64(double %142, double %128, double noundef 0x3F900F7FCF183E0B) #140
  %144 = call double @llvm.fma.f64(double %143, double %128, double noundef 0xBFD15F7977A772D4) #140
  %145 = fmul double %128, %144
  br label %__nv_j0.exit

146:                                              ; preds = %124
  %147 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %148 = icmp eq i32 %147, 200
  br i1 %148, label %155, label %149

149:                                              ; preds = %146
  %150 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %151 = icmp eq i32 %150, 350
  br i1 %151, label %155, label %152

152:                                              ; preds = %149
  %153 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %154 = icmp eq i32 %153, 370
  br i1 %154, label %155, label %157

155:                                              ; preds = %152, %149, %146
  %156 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i

157:                                              ; preds = %152
  %158 = and i32 %16, 2147483647
  %159 = icmp eq i32 %158, 2146435072
  %160 = icmp eq i32 %17, 0
  %161 = select i1 %159, i1 %160, i1 false
  br label %__nv_isinfd.exit.i

__nv_isinfd.exit.i:                               ; preds = %157, %155
  %.012.in = phi i1 [ %156, %155 ], [ %161, %157 ]
  br i1 %.012.in, label %__nv_j0.exit, label %162

162:                                              ; preds = %__nv_isinfd.exit.i
  %163 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %164 = fneg double %1
  %165 = call double @llvm.fma.f64(double %164, double %163, double noundef 1.000000e+00)
  %166 = call double @llvm.fma.f64(double %165, double %165, double %165)
  %167 = call double @llvm.fma.f64(double %166, double %163, double %163)
  %168 = fmul double %167, %167
  %169 = call double @llvm.fma.f64(double %168, double noundef 0xC0D115CB8C11A9DC, double noundef 0x409927467A655012)
  %170 = call double @llvm.fma.f64(double %169, double %168, double noundef 0xC05751787E247BD4)
  %171 = call double @llvm.fma.f64(double %170, double %168, double noundef 0x401704C4E5FC36B2)
  %172 = call double @llvm.fma.f64(double %171, double %168, double noundef 0xBFE15B747A2FD531)
  %173 = call double @llvm.fma.f64(double %172, double %168, double noundef 0x3FBA7FEACF6CB79B)
  %174 = call double @llvm.fma.f64(double %173, double %168, double noundef 0xBFAFFFFFEDDCF548)
  %175 = call double @llvm.fma.f64(double %174, double %168, double noundef 0x3FEFFFFFFFFFC9E5)
  %176 = call double @llvm.fma.f64(double %168, double noundef 0xC14602FE1C34685E, double noundef 0x410ECD4523B12B84)
  %177 = call double @llvm.fma.f64(double %176, double %168, double noundef 0xC0C7A2FC1972F05A)
  %178 = call double @llvm.fma.f64(double %177, double %168, double noundef 0x407EBA131F7E5BEB)
  %179 = call double @llvm.fma.f64(double %178, double %168, double noundef 0xC0373B92E6E7CC7D)
  %180 = call double @llvm.fma.f64(double %179, double %168, double noundef 0x3FFA31BEE63A2F08)
  %181 = call double @llvm.fma.f64(double %180, double %168, double noundef 0xBFCAD320104D5D05)
  %182 = call double @llvm.fma.f64(double %181, double %168, double noundef 0x3FB0AAAA9C76D07E)
  %183 = call double @llvm.fma.f64(double %182, double %168, double noundef 0xBFBFFFFFFFFDACEC)
  %184 = call double @llvm.fma.f64(double %183, double %167, double %1)
  %185 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %186 = fmul double %185, 0x3FE9884533D43651
  %187 = fmul double %186, %175
  %188 = fmul double %184, 0x3FE45F306DC9C883
  %189 = call i32 @llvm.nvvm.d2i.rn(double %188) #139
  store i32 %189, i32* %q.i.i.i20, align 4, !noalias !68
  %190 = sitofp i32 %189 to double
  %191 = fneg double %190
  %192 = call double @llvm.fma.f64(double %191, double noundef 0x3FF921FB54442D18, double %184)
  %193 = call double @llvm.fma.f64(double %191, double noundef 0x3C91A62633145C00, double %192)
  %194 = call double @llvm.fma.f64(double %191, double noundef 0x397B839A252049C0, double %193)
  %195 = call double @llvm.fabs.f64(double %184)
  %196 = fcmp ult double %195, 0x41E0000000000000
  br i1 %196, label %__internal_trig_reduction_kerneld.exit.i.i25, label %197

197:                                              ; preds = %162
  %198 = call fastcc double @__internal_trig_reduction_slowpathd(double %184, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i20) #139
  %.pre = load i32, i32* %q.i.i.i20, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i25

__internal_trig_reduction_kerneld.exit.i.i25:     ; preds = %197, %162
  %199 = phi i32 [ %.pre, %197 ], [ %189, %162 ]
  %t.i.i.i19.0 = phi double [ %198, %197 ], [ %194, %162 ]
  %200 = and i32 %199, 3
  %201 = uitofp i32 %200 to double
  %202 = fadd double %t.i.i.i19.0, 0xBFE921FB54442D18
  %203 = fmul double %201, 0x3FF921FB54442D18
  %204 = fadd double %202, %203
  %205 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %206 = icmp eq i32 %205, 350
  br i1 %206, label %210, label %207

207:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i25
  %208 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %209 = icmp eq i32 %208, 370
  br i1 %209, label %210, label %269

210:                                              ; preds = %207, %__internal_trig_reduction_kerneld.exit.i.i25
  %211 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %212 = icmp eq i32 %211, 200
  br i1 %212, label %219, label %213

213:                                              ; preds = %210
  %214 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %215 = icmp eq i32 %214, 350
  br i1 %215, label %219, label %216

216:                                              ; preds = %213
  %217 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %218 = icmp eq i32 %217, 370
  br i1 %218, label %219, label %222

219:                                              ; preds = %216, %213, %210
  %220 = call double @llvm.fabs.f64(double %204)
  %221 = fcmp oeq double %220, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i

222:                                              ; preds = %216
  %223 = call i32 @llvm.nvvm.d2i.lo(double %204) #139
  %224 = call i32 @llvm.nvvm.d2i.hi(double %204) #139
  %225 = and i32 %224, 2147483647
  %226 = icmp eq i32 %225, 2146435072
  %227 = icmp eq i32 %223, 0
  %228 = select i1 %226, i1 %227, i1 false
  br label %__nv_isinfd.exit.i.i.i.i

__nv_isinfd.exit.i.i.i.i:                         ; preds = %222, %219
  %.08.in = phi i1 [ %221, %219 ], [ %228, %222 ]
  br i1 %.08.in, label %229, label %231

229:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %230 = fmul double %204, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

231:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %232 = fmul double %204, 0x3FE45F306DC9C883
  %233 = call i32 @llvm.nvvm.d2i.rn(double %232) #139
  store i32 %233, i32* %q.i.i.i.i.i, align 4, !noalias !68
  %234 = sitofp i32 %233 to double
  %235 = fneg double %234
  %236 = call double @llvm.fma.f64(double %235, double noundef 0x3FF921FB54442D18, double %204)
  %237 = call double @llvm.fma.f64(double %235, double noundef 0x3C91A62633145C00, double %236)
  %238 = call double @llvm.fma.f64(double %235, double noundef 0x397B839A252049C0, double %237)
  %239 = call double @llvm.fabs.f64(double %204)
  %240 = fcmp ult double %239, 0x41E0000000000000
  br i1 %240, label %__internal_trig_reduction_kerneld.exit.i.i.i.i, label %241

241:                                              ; preds = %231
  %242 = call fastcc double @__internal_trig_reduction_slowpathd(double %204, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i) #139
  %.pre17 = load i32, i32* %q.i.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i.i:   ; preds = %231, %241, %229
  %i.i.i.i.i.0 = phi i32 [ 0, %229 ], [ %.pre17, %241 ], [ %233, %231 ]
  %.07 = phi double [ %230, %229 ], [ %242, %241 ], [ %238, %231 ]
  %243 = fmul double %.07, %.07
  %244 = call double @llvm.fma.f64(double %243, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %245 = call double @llvm.fma.f64(double %244, double %243, double noundef 0xBE927E4F8E06E6D9)
  %246 = call double @llvm.fma.f64(double %245, double %243, double noundef 0x3EFA01A019DDBCE9)
  %247 = call double @llvm.fma.f64(double %246, double %243, double noundef 0xBF56C16C16C15D47)
  %248 = call double @llvm.fma.f64(double %247, double %243, double noundef 0x3FA5555555555551)
  %249 = call double @llvm.fma.f64(double %248, double %243, double noundef -5.000000e-01)
  %250 = call double @llvm.fma.f64(double %249, double %243, double noundef 1.000000e+00)
  %251 = call double @llvm.fma.f64(double %243, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %252 = call double @llvm.fma.f64(double %251, double %243, double noundef 0x3EC71DE369ACE392)
  %253 = call double @llvm.fma.f64(double %252, double %243, double noundef 0xBF2A01A019DB62A1)
  %254 = call double @llvm.fma.f64(double %253, double %243, double noundef 0x3F81111111110818)
  %255 = call double @llvm.fma.f64(double %254, double %243, double noundef 0xBFC5555555555554)
  %256 = call double @llvm.fma.f64(double %255, double %243, double noundef 0.000000e+00)
  %257 = call double @llvm.fma.f64(double %256, double %.07, double %.07)
  %258 = call i32 @llvm.nvvm.d2i.hi(double %257) #139
  %259 = call i32 @llvm.nvvm.d2i.lo(double %257) #139
  %260 = xor i32 %258, -2147483648
  %261 = call double @llvm.nvvm.lohi.i2d(i32 %259, i32 %260) #139
  %262 = and i32 %i.i.i.i.i.0, 1
  %.not14 = icmp eq i32 %262, 0
  %spec.select1 = select i1 %.not14, double %250, double %261
  %263 = and i32 %i.i.i.i.i.0, 2
  %.not15 = icmp eq i32 %263, 0
  br i1 %.not15, label %__internal_cos_offset.exit.i, label %264

264:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i
  %265 = call i32 @llvm.nvvm.d2i.hi(double %spec.select1) #139
  %266 = call i32 @llvm.nvvm.d2i.lo(double %spec.select1) #139
  %267 = xor i32 %265, -2147483648
  %268 = call double @llvm.nvvm.lohi.i2d(i32 %266, i32 %267) #139
  br label %__internal_cos_offset.exit.i

269:                                              ; preds = %207
  %270 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %271 = icmp eq i32 %270, 200
  br i1 %271, label %278, label %272

272:                                              ; preds = %269
  %273 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %274 = icmp eq i32 %273, 350
  br i1 %274, label %278, label %275

275:                                              ; preds = %272
  %276 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %277 = icmp eq i32 %276, 370
  br i1 %277, label %278, label %281

278:                                              ; preds = %275, %272, %269
  %279 = call double @llvm.fabs.f64(double %204)
  %280 = fcmp oeq double %279, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i26

281:                                              ; preds = %275
  %282 = call i32 @llvm.nvvm.d2i.lo(double %204) #139
  %283 = call i32 @llvm.nvvm.d2i.hi(double %204) #139
  %284 = and i32 %283, 2147483647
  %285 = icmp eq i32 %284, 2146435072
  %286 = icmp eq i32 %282, 0
  %287 = select i1 %285, i1 %286, i1 false
  br label %__nv_isinfd.exit.i.i.i26

__nv_isinfd.exit.i.i.i26:                         ; preds = %281, %278
  %.09.in = phi i1 [ %280, %278 ], [ %287, %281 ]
  br i1 %.09.in, label %288, label %290

288:                                              ; preds = %__nv_isinfd.exit.i.i.i26
  %289 = fmul double %204, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i

290:                                              ; preds = %__nv_isinfd.exit.i.i.i26
  %291 = fmul double %204, 0x3FE45F306DC9C883
  %292 = call i32 @llvm.nvvm.d2i.rn(double %291) #139
  store i32 %292, i32* %q.i.i.i.i9, align 4, !noalias !68
  %293 = sitofp i32 %292 to double
  %294 = fneg double %293
  %295 = call double @llvm.fma.f64(double %294, double noundef 0x3FF921FB54442D18, double %204)
  %296 = call double @llvm.fma.f64(double %294, double noundef 0x3C91A62633145C00, double %295)
  %297 = call double @llvm.fma.f64(double %294, double noundef 0x397B839A252049C0, double %296)
  %298 = call double @llvm.fabs.f64(double %204)
  %299 = fcmp ult double %298, 0x41E0000000000000
  br i1 %299, label %__internal_trig_reduction_kerneld.exit.i.i.i27, label %300

300:                                              ; preds = %290
  %301 = call fastcc double @__internal_trig_reduction_slowpathd(double %204, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i9) #139
  %.pre16 = load i32, i32* %q.i.i.i.i9, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i27

__internal_trig_reduction_kerneld.exit.i.i.i27:   ; preds = %300, %290
  %302 = phi i32 [ %.pre16, %300 ], [ %292, %290 ]
  %t.i1.i.i.i.0 = phi double [ %301, %300 ], [ %297, %290 ]
  %phi.bo10 = add i32 %302, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i

__internal_sin_cos_kerneld.exit.i.i.i:            ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i27, %288
  %i.i.i.i17.0 = phi i32 [ 1, %288 ], [ %phi.bo10, %__internal_trig_reduction_kerneld.exit.i.i.i27 ]
  %z.i.i.i16.0 = phi double [ %289, %288 ], [ %t.i1.i.i.i.0, %__internal_trig_reduction_kerneld.exit.i.i.i27 ]
  %303 = shl i32 %i.i.i.i17.0, 3
  %304 = and i32 %303, 8
  %305 = zext i32 %304 to i64
  %306 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %305
  %307 = fmul double %z.i.i.i16.0, %z.i.i.i16.0
  %308 = and i32 %i.i.i.i17.0, 1
  %.not11 = icmp eq i32 %308, 0
  %309 = select i1 %.not11, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %310 = getelementptr inbounds double, double addrspace(1)* %306, i64 1
  %311 = load double, double addrspace(1)* %310, align 8
  %312 = call double @llvm.fma.f64(double noundef %309, double %307, double %311)
  %313 = getelementptr inbounds double, double addrspace(1)* %306, i64 2
  %314 = load double, double addrspace(1)* %313, align 8
  %315 = call double @llvm.fma.f64(double %312, double %307, double %314)
  %316 = getelementptr inbounds double, double addrspace(1)* %306, i64 3
  %317 = load double, double addrspace(1)* %316, align 8
  %318 = call double @llvm.fma.f64(double %315, double %307, double %317)
  %319 = getelementptr inbounds double, double addrspace(1)* %306, i64 4
  %320 = load double, double addrspace(1)* %319, align 8
  %321 = call double @llvm.fma.f64(double %318, double %307, double %320)
  %322 = getelementptr inbounds double, double addrspace(1)* %306, i64 5
  %323 = load double, double addrspace(1)* %322, align 8
  %324 = call double @llvm.fma.f64(double %321, double %307, double %323)
  %325 = getelementptr inbounds double, double addrspace(1)* %306, i64 6
  %326 = load double, double addrspace(1)* %325, align 8
  %327 = call double @llvm.fma.f64(double %324, double %307, double %326)
  %328 = call double @llvm.fma.f64(double %327, double %z.i.i.i16.0, double %z.i.i.i16.0)
  %329 = call double @llvm.fma.f64(double %327, double %307, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not11, double %328, double %329
  %330 = and i32 %i.i.i.i17.0, 2
  %.not13 = icmp eq i32 %330, 0
  %331 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.111 = select i1 %.not13, double %spec.select2, double %331
  br label %__internal_cos_offset.exit.i

__internal_cos_offset.exit.i:                     ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i, %264, %__internal_sin_cos_kerneld.exit.i.i.i
  %.06 = phi double [ %.111, %__internal_sin_cos_kerneld.exit.i.i.i ], [ %268, %264 ], [ %spec.select1, %__internal_trig_reduction_kerneld.exit.i.i.i.i ]
  %332 = fmul double %187, %.06
  br label %__nv_j0.exit

__nv_j0.exit:                                     ; preds = %104, %__internal_cos_offset.exit.i, %__nv_isinfd.exit.i, %126, %82
  %r.i.3 = phi double [ %101, %82 ], [ %123, %104 ], [ %145, %126 ], [ %332, %__internal_cos_offset.exit.i ], [ 0.000000e+00, %__nv_isinfd.exit.i ]
  %333 = fmul double %80, %r.i.3
  %334 = fadd double %15, %333
  br label %604

335:                                              ; preds = %0
  %336 = fcmp ugt double %1, 0x4000347C4AB37B18
  br i1 %336, label %364, label %337

337:                                              ; preds = %335
  %338 = fadd double %1, 0xBFEC982EB8D417EA
  %339 = fadd double %338, 0xBC7EA9D270347F83
  %340 = call double @llvm.fma.f64(double %339, double noundef 0xBF01630132D75FC3, double noundef 0x3F3D054B05D3C52D)
  %341 = call double @llvm.fma.f64(double %340, double %339, double noundef 0xBF66DAC0B314B2E5)
  %342 = call double @llvm.fma.f64(double %341, double %339, double noundef 0x3F86A5D1DE76263F)
  %343 = call double @llvm.fma.f64(double %342, double %339, double noundef 0xBF9FD16652824592)
  %344 = call double @llvm.fma.f64(double %343, double %339, double noundef 0x3FB0F69A9CC79FBD)
  %345 = call double @llvm.fma.f64(double %344, double %339, double noundef 0xBFBCCE40EF15583E)
  %346 = call double @llvm.fma.f64(double %345, double %339, double noundef 0x3FC446B11780E4FC)
  %347 = call double @llvm.fma.f64(double %346, double %339, double noundef 0xBFC89AE7E19621F7)
  %348 = call double @llvm.fma.f64(double %347, double %339, double noundef 0x3FCACBA1B38EF7B8)
  %349 = call double @llvm.fma.f64(double %348, double %339, double noundef 0xBFCB4166A03BBFA5)
  %350 = call double @llvm.fma.f64(double %349, double %339, double noundef 0x3FCACCA4D5D4889A)
  %351 = call double @llvm.fma.f64(double %350, double %339, double noundef 0xBFCA1455932B9392)
  %352 = call double @llvm.fma.f64(double %351, double %339, double noundef 0x3FC96D8DB8D844EC)
  %353 = call double @llvm.fma.f64(double %352, double %339, double noundef 0xBFC8F7FB77522EDF)
  %354 = call double @llvm.fma.f64(double %353, double %339, double noundef 0x3FC8C0926ABC9AB0)
  %355 = call double @llvm.fma.f64(double %354, double %339, double noundef 0xBFC8D35B8FEA468C)
  %356 = call double @llvm.fma.f64(double %355, double %339, double noundef 0x3FC9424B8A0C8F94)
  %357 = call double @llvm.fma.f64(double %356, double %339, double noundef 0xBFCA396A7F3403EF)
  %358 = call double @llvm.fma.f64(double %357, double %339, double noundef 0x3FCC068086C37055)
  %359 = call double @llvm.fma.f64(double %358, double %339, double noundef 0xBFCCF18E6A4C5C4E)
  %360 = call double @llvm.fma.f64(double %359, double %339, double noundef 0x3FCC3B1338AF4239)
  %361 = call double @llvm.fma.f64(double %360, double %339, double noundef 0xBFDF7E38A46D70DB)
  %362 = call double @llvm.fma.f64(double %361, double %339, double noundef 0x3FEC24371844B88A)
  %363 = fmul double %339, %362
  br label %604

364:                                              ; preds = %335
  %365 = fcmp ugt double %1, 0x40161663B5D9A628
  br i1 %365, label %393, label %366

366:                                              ; preds = %364
  %367 = fadd double %1, 0xC00FA9534D98569C
  %368 = fadd double %367, 0x3C9F06AE7804384E
  %369 = call double @llvm.fma.f64(double %368, double noundef 0xBCDAEA62AC8BDA68, double noundef 0xBCD2434958151AC7)
  %370 = call double @llvm.fma.f64(double %369, double %368, double noundef 0x3D11C24A40D33FE1)
  %371 = call double @llvm.fma.f64(double %370, double %368, double noundef 0x3D237CD62FA08CA4)
  %372 = call double @llvm.fma.f64(double %371, double %368, double noundef 0xBD43902E0298C52A)
  %373 = call double @llvm.fma.f64(double %372, double %368, double noundef 0xBD1DDAAD11CAB40F)
  %374 = call double @llvm.fma.f64(double %373, double %368, double noundef 0xBD5209D9F06D7DE4)
  %375 = call double @llvm.fma.f64(double %374, double %368, double noundef 0x3D8BB9F464468E1A)
  %376 = call double @llvm.fma.f64(double %375, double %368, double noundef 0xBDA8F67B07D1B440)
  %377 = call double @llvm.fma.f64(double %376, double %368, double noundef 0x3DC7C8D60F9EAECF)
  %378 = call double @llvm.fma.f64(double %377, double %368, double noundef 0xBDE9703405B49A8D)
  %379 = call double @llvm.fma.f64(double %378, double %368, double noundef 0x3E0A6B64E76417E4)
  %380 = call double @llvm.fma.f64(double %379, double %368, double noundef 0xBE2F6B5AFB2F1359)
  %381 = call double @llvm.fma.f64(double %380, double %368, double noundef 0x3E54526B71C21EC1)
  %382 = call double @llvm.fma.f64(double %381, double %368, double noundef 0xBE5776DBCBBC8E1D)
  %383 = call double @llvm.fma.f64(double %382, double %368, double noundef 0xBE93B211FC2DF90E)
  %384 = call double @llvm.fma.f64(double %383, double %368, double noundef 0xBED486372E8562DC)
  %385 = call double @llvm.fma.f64(double %384, double %368, double noundef 0x3F0AB2C1FBC3A254)
  %386 = call double @llvm.fma.f64(double %385, double %368, double noundef 0x3F299827653353B8)
  %387 = call double @llvm.fma.f64(double %386, double %368, double noundef 0xBF61E32BC4ED7084)
  %388 = call double @llvm.fma.f64(double %387, double %368, double noundef 0xBF7C116FDC599A09)
  %389 = call double @llvm.fma.f64(double %388, double %368, double noundef 0x3FADF6D59BF50C77)
  %390 = call double @llvm.fma.f64(double %389, double %368, double noundef 0x3FAA09C92903680B)
  %391 = call double @llvm.fma.f64(double %390, double %368, double noundef 0xBFD9C34256A12A0B)
  %392 = fmul double %368, %391
  br label %604

393:                                              ; preds = %364
  %394 = fcmp ugt double %1, 0x40214EF30C0C06ED
  br i1 %394, label %414, label %395

395:                                              ; preds = %393
  %396 = fadd double %1, 0xC01C581DC4E72103
  %397 = fadd double %396, 0x3C99774A495F56CF
  %398 = call double @llvm.fma.f64(double %397, double noundef 0x3CF1CB3ABA718B8E, double noundef 0xBD3F443BB4F53D75)
  %399 = call double @llvm.fma.f64(double %398, double %397, double noundef 0xBD770F737BD6A786)
  %400 = call double @llvm.fma.f64(double %399, double %397, double noundef 0x3DBF0E9A20459E14)
  %401 = call double @llvm.fma.f64(double %400, double %397, double noundef 0x3DEFA6B137D5E108)
  %402 = call double @llvm.fma.f64(double %401, double %397, double noundef 0xBE344296729FB7FA)
  %403 = call double @llvm.fma.f64(double %402, double %397, double noundef 0xBE60A2813A80DFAA)
  %404 = call double @llvm.fma.f64(double %403, double %397, double noundef 0x3EA34AA737A83EB4)
  %405 = call double @llvm.fma.f64(double %404, double %397, double noundef 0x3EC6A9227332D03C)
  %406 = call double @llvm.fma.f64(double %405, double %397, double noundef 0xBF08177E4F93C81E)
  %407 = call double @llvm.fma.f64(double %406, double %397, double noundef 0xBF226DD71E391775)
  %408 = call double @llvm.fma.f64(double %407, double %397, double noundef 0x3F61D35E85FD7B22)
  %409 = call double @llvm.fma.f64(double %408, double %397, double noundef 0x3F6B2F14A955285C)
  %410 = call double @llvm.fma.f64(double %409, double %397, double noundef 0xBFA8969C64CBF388)
  %411 = call double @llvm.fma.f64(double %410, double %397, double noundef 0xBF95AEF611FC4D5A)
  %412 = call double @llvm.fma.f64(double %411, double %397, double noundef 0x3FD334CCA0697A5A)
  %413 = fmul double %397, %412
  br label %604

414:                                              ; preds = %393
  %415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %416 = icmp eq i32 %415, 200
  br i1 %416, label %423, label %417

417:                                              ; preds = %414
  %418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %419 = icmp eq i32 %418, 350
  br i1 %419, label %423, label %420

420:                                              ; preds = %417
  %421 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %422 = icmp eq i32 %421, 370
  br i1 %422, label %423, label %425

423:                                              ; preds = %420, %417, %414
  %424 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit

425:                                              ; preds = %420
  %426 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %427 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %428 = and i32 %427, 2147483647
  %429 = icmp eq i32 %428, 2146435072
  %430 = icmp eq i32 %426, 0
  %431 = select i1 %429, i1 %430, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %425, %423
  %.013.in = phi i1 [ %424, %423 ], [ %431, %425 ]
  br i1 %.013.in, label %604, label %432

432:                                              ; preds = %__nv_isinfd.exit
  %433 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %434 = fneg double %1
  %435 = call double @llvm.fma.f64(double %434, double %433, double noundef 1.000000e+00)
  %436 = call double @llvm.fma.f64(double %435, double %435, double %435)
  %437 = call double @llvm.fma.f64(double %436, double %433, double %433)
  %438 = fmul double %437, %437
  %439 = call double @llvm.fma.f64(double %438, double noundef 0xC0C5E91E6AC3AD03, double noundef 0x4093F56A049CDDE7)
  %440 = call double @llvm.fma.f64(double %439, double %438, double noundef 0xC05572D39DFB8433)
  %441 = call double @llvm.fma.f64(double %440, double %438, double noundef 0x4016A6041CAA59E5)
  %442 = call double @llvm.fma.f64(double %441, double %438, double noundef 0xBFE155E3A0493880)
  %443 = call double @llvm.fma.f64(double %442, double %438, double noundef 0x3FBA7FB92F417F7F)
  %444 = call double @llvm.fma.f64(double %443, double %438, double noundef 0xBFAFFFFFB12E32F5)
  %445 = call double @llvm.fma.f64(double %444, double %438, double noundef 0x3FEFFFFFFFFECED5)
  %446 = call double @llvm.fma.f64(double %438, double noundef 0x418A86A64BE101DC, double noundef 0xC15709C79AAC5813)
  %447 = call double @llvm.fma.f64(double %446, double %438, double noundef 0x41142A31C980A287)
  %448 = call double @llvm.fma.f64(double %447, double %438, double noundef 0xC0C9CBE68930485D)
  %449 = call double @llvm.fma.f64(double %448, double %438, double noundef 0x407F583E14E8A4E8)
  %450 = call double @llvm.fma.f64(double %449, double %438, double noundef 0xC0374A629C650680)
  %451 = call double @llvm.fma.f64(double %450, double %438, double noundef 0x3FFA32A7AF17FAE9)
  %452 = call double @llvm.fma.f64(double %451, double %438, double noundef 0xBFCAD32497785CD6)
  %453 = call double @llvm.fma.f64(double %452, double %438, double noundef 0x3FB0AAAA9FB75F7B)
  %454 = call double @llvm.fma.f64(double %453, double %438, double noundef 0xBFBFFFFFFFFE320F)
  %455 = call double @llvm.fma.f64(double %454, double %437, double %1)
  %456 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %457 = fmul double %456, 0x3FE9884533D43651
  %458 = fmul double %457, %445
  %459 = fmul double %455, 0x3FE45F306DC9C883
  %460 = call i32 @llvm.nvvm.d2i.rn(double %459) #139
  store i32 %460, i32* %q.i.i, align 4, !noalias !68
  %461 = sitofp i32 %460 to double
  %462 = fneg double %461
  %463 = call double @llvm.fma.f64(double %462, double noundef 0x3FF921FB54442D18, double %455)
  %464 = call double @llvm.fma.f64(double %462, double noundef 0x3C91A62633145C00, double %463)
  %465 = call double @llvm.fma.f64(double %462, double noundef 0x397B839A252049C0, double %464)
  %466 = call double @llvm.fabs.f64(double %455)
  %467 = fcmp ult double %466, 0x41E0000000000000
  br i1 %467, label %__internal_trig_reduction_kerneld.exit.i, label %468

468:                                              ; preds = %432
  %469 = call fastcc double @__internal_trig_reduction_slowpathd(double %455, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i) #139
  %.pre18 = load i32, i32* %q.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i

__internal_trig_reduction_kerneld.exit.i:         ; preds = %468, %432
  %470 = phi i32 [ %.pre18, %468 ], [ %460, %432 ]
  %t.i.i.0 = phi double [ %469, %468 ], [ %465, %432 ]
  %471 = and i32 %470, 3
  %472 = uitofp i32 %471 to double
  %473 = fadd double %t.i.i.0, 0xC002D97C7F3321D2
  %474 = fmul double %472, 0x3FF921FB54442D18
  %475 = fadd double %473, %474
  %476 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %477 = icmp eq i32 %476, 350
  br i1 %477, label %481, label %478

478:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i
  %479 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %480 = icmp eq i32 %479, 370
  br i1 %480, label %481, label %540

481:                                              ; preds = %478, %__internal_trig_reduction_kerneld.exit.i
  %482 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %483 = icmp eq i32 %482, 200
  br i1 %483, label %490, label %484

484:                                              ; preds = %481
  %485 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %486 = icmp eq i32 %485, 350
  br i1 %486, label %490, label %487

487:                                              ; preds = %484
  %488 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %489 = icmp eq i32 %488, 370
  br i1 %489, label %490, label %493

490:                                              ; preds = %487, %484, %481
  %491 = call double @llvm.fabs.f64(double %475)
  %492 = fcmp oeq double %491, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i

493:                                              ; preds = %487
  %494 = call i32 @llvm.nvvm.d2i.lo(double %475) #139
  %495 = call i32 @llvm.nvvm.d2i.hi(double %475) #139
  %496 = and i32 %495, 2147483647
  %497 = icmp eq i32 %496, 2146435072
  %498 = icmp eq i32 %494, 0
  %499 = select i1 %497, i1 %498, i1 false
  br label %__nv_isinfd.exit.i.i.i

__nv_isinfd.exit.i.i.i:                           ; preds = %493, %490
  %.02.in = phi i1 [ %492, %490 ], [ %499, %493 ]
  br i1 %.02.in, label %500, label %502

500:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %501 = fmul double %475, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

502:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %503 = fmul double %475, 0x3FE45F306DC9C883
  %504 = call i32 @llvm.nvvm.d2i.rn(double %503) #139
  store i32 %504, i32* %q.i.i.i.i, align 4, !noalias !68
  %505 = sitofp i32 %504 to double
  %506 = fneg double %505
  %507 = call double @llvm.fma.f64(double %506, double noundef 0x3FF921FB54442D18, double %475)
  %508 = call double @llvm.fma.f64(double %506, double noundef 0x3C91A62633145C00, double %507)
  %509 = call double @llvm.fma.f64(double %506, double noundef 0x397B839A252049C0, double %508)
  %510 = call double @llvm.fabs.f64(double %475)
  %511 = fcmp ult double %510, 0x41E0000000000000
  br i1 %511, label %__internal_trig_reduction_kerneld.exit.i.i.i, label %512

512:                                              ; preds = %502
  %513 = call fastcc double @__internal_trig_reduction_slowpathd(double %475, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i) #139
  %.pre20 = load i32, i32* %q.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i:     ; preds = %502, %512, %500
  %i.i.i.i.0 = phi i32 [ 0, %500 ], [ %.pre20, %512 ], [ %504, %502 ]
  %.01 = phi double [ %501, %500 ], [ %513, %512 ], [ %509, %502 ]
  %514 = fmul double %.01, %.01
  %515 = call double @llvm.fma.f64(double %514, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %516 = call double @llvm.fma.f64(double %515, double %514, double noundef 0xBE927E4F8E06E6D9)
  %517 = call double @llvm.fma.f64(double %516, double %514, double noundef 0x3EFA01A019DDBCE9)
  %518 = call double @llvm.fma.f64(double %517, double %514, double noundef 0xBF56C16C16C15D47)
  %519 = call double @llvm.fma.f64(double %518, double %514, double noundef 0x3FA5555555555551)
  %520 = call double @llvm.fma.f64(double %519, double %514, double noundef -5.000000e-01)
  %521 = call double @llvm.fma.f64(double %520, double %514, double noundef 1.000000e+00)
  %522 = call double @llvm.fma.f64(double %514, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %523 = call double @llvm.fma.f64(double %522, double %514, double noundef 0x3EC71DE369ACE392)
  %524 = call double @llvm.fma.f64(double %523, double %514, double noundef 0xBF2A01A019DB62A1)
  %525 = call double @llvm.fma.f64(double %524, double %514, double noundef 0x3F81111111110818)
  %526 = call double @llvm.fma.f64(double %525, double %514, double noundef 0xBFC5555555555554)
  %527 = call double @llvm.fma.f64(double %526, double %514, double noundef 0.000000e+00)
  %528 = call double @llvm.fma.f64(double %527, double %.01, double %.01)
  %529 = call i32 @llvm.nvvm.d2i.hi(double %528) #139
  %530 = call i32 @llvm.nvvm.d2i.lo(double %528) #139
  %531 = xor i32 %529, -2147483648
  %532 = call double @llvm.nvvm.lohi.i2d(i32 %530, i32 %531) #139
  %533 = and i32 %i.i.i.i.0, 1
  %.not8 = icmp eq i32 %533, 0
  %spec.select4 = select i1 %.not8, double %521, double %532
  %534 = and i32 %i.i.i.i.0, 2
  %.not9 = icmp eq i32 %534, 0
  br i1 %.not9, label %__internal_cos_offset.exit, label %535

535:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i
  %536 = call i32 @llvm.nvvm.d2i.hi(double %spec.select4) #139
  %537 = call i32 @llvm.nvvm.d2i.lo(double %spec.select4) #139
  %538 = xor i32 %536, -2147483648
  %539 = call double @llvm.nvvm.lohi.i2d(i32 %537, i32 %538) #139
  br label %__internal_cos_offset.exit

540:                                              ; preds = %478
  %541 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %542 = icmp eq i32 %541, 200
  br i1 %542, label %549, label %543

543:                                              ; preds = %540
  %544 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %545 = icmp eq i32 %544, 350
  br i1 %545, label %549, label %546

546:                                              ; preds = %543
  %547 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %548 = icmp eq i32 %547, 370
  br i1 %548, label %549, label %552

549:                                              ; preds = %546, %543, %540
  %550 = call double @llvm.fabs.f64(double %475)
  %551 = fcmp oeq double %550, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i

552:                                              ; preds = %546
  %553 = call i32 @llvm.nvvm.d2i.lo(double %475) #139
  %554 = call i32 @llvm.nvvm.d2i.hi(double %475) #139
  %555 = and i32 %554, 2147483647
  %556 = icmp eq i32 %555, 2146435072
  %557 = icmp eq i32 %553, 0
  %558 = select i1 %556, i1 %557, i1 false
  br label %__nv_isinfd.exit.i.i

__nv_isinfd.exit.i.i:                             ; preds = %552, %549
  %.03.in = phi i1 [ %551, %549 ], [ %558, %552 ]
  br i1 %.03.in, label %559, label %561

559:                                              ; preds = %__nv_isinfd.exit.i.i
  %560 = fmul double %475, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i

561:                                              ; preds = %__nv_isinfd.exit.i.i
  %562 = fmul double %475, 0x3FE45F306DC9C883
  %563 = call i32 @llvm.nvvm.d2i.rn(double %562) #139
  store i32 %563, i32* %q.i.i.i, align 4, !noalias !68
  %564 = sitofp i32 %563 to double
  %565 = fneg double %564
  %566 = call double @llvm.fma.f64(double %565, double noundef 0x3FF921FB54442D18, double %475)
  %567 = call double @llvm.fma.f64(double %565, double noundef 0x3C91A62633145C00, double %566)
  %568 = call double @llvm.fma.f64(double %565, double noundef 0x397B839A252049C0, double %567)
  %569 = call double @llvm.fabs.f64(double %475)
  %570 = fcmp ult double %569, 0x41E0000000000000
  br i1 %570, label %__internal_trig_reduction_kerneld.exit.i.i, label %571

571:                                              ; preds = %561
  %572 = call fastcc double @__internal_trig_reduction_slowpathd(double %475, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i) #139
  %.pre19 = load i32, i32* %q.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i

__internal_trig_reduction_kerneld.exit.i.i:       ; preds = %571, %561
  %573 = phi i32 [ %.pre19, %571 ], [ %563, %561 ]
  %t.i1.i.i.0 = phi double [ %572, %571 ], [ %568, %561 ]
  %phi.bo = add i32 %573, 1
  br label %__internal_sin_cos_kerneld.exit.i.i

__internal_sin_cos_kerneld.exit.i.i:              ; preds = %__internal_trig_reduction_kerneld.exit.i.i, %559
  %i.i.i.0 = phi i32 [ 1, %559 ], [ %phi.bo, %__internal_trig_reduction_kerneld.exit.i.i ]
  %z.i.i.0 = phi double [ %560, %559 ], [ %t.i1.i.i.0, %__internal_trig_reduction_kerneld.exit.i.i ]
  %574 = shl i32 %i.i.i.0, 3
  %575 = and i32 %574, 8
  %576 = zext i32 %575 to i64
  %577 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %576
  %578 = fmul double %z.i.i.0, %z.i.i.0
  %579 = and i32 %i.i.i.0, 1
  %.not = icmp eq i32 %579, 0
  %580 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %581 = getelementptr inbounds double, double addrspace(1)* %577, i64 1
  %582 = load double, double addrspace(1)* %581, align 8
  %583 = call double @llvm.fma.f64(double noundef %580, double %578, double %582)
  %584 = getelementptr inbounds double, double addrspace(1)* %577, i64 2
  %585 = load double, double addrspace(1)* %584, align 8
  %586 = call double @llvm.fma.f64(double %583, double %578, double %585)
  %587 = getelementptr inbounds double, double addrspace(1)* %577, i64 3
  %588 = load double, double addrspace(1)* %587, align 8
  %589 = call double @llvm.fma.f64(double %586, double %578, double %588)
  %590 = getelementptr inbounds double, double addrspace(1)* %577, i64 4
  %591 = load double, double addrspace(1)* %590, align 8
  %592 = call double @llvm.fma.f64(double %589, double %578, double %591)
  %593 = getelementptr inbounds double, double addrspace(1)* %577, i64 5
  %594 = load double, double addrspace(1)* %593, align 8
  %595 = call double @llvm.fma.f64(double %592, double %578, double %594)
  %596 = getelementptr inbounds double, double addrspace(1)* %577, i64 6
  %597 = load double, double addrspace(1)* %596, align 8
  %598 = call double @llvm.fma.f64(double %595, double %578, double %597)
  %599 = call double @llvm.fma.f64(double %598, double %z.i.i.0, double %z.i.i.0)
  %600 = call double @llvm.fma.f64(double %598, double %578, double noundef 1.000000e+00)
  %spec.select5 = select i1 %.not, double %599, double %600
  %601 = and i32 %i.i.i.0, 2
  %.not7 = icmp eq i32 %601, 0
  %602 = call double @llvm.fma.f64(double %spec.select5, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not7, double %spec.select5, double %602
  br label %__internal_cos_offset.exit

__internal_cos_offset.exit:                       ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i, %535, %__internal_sin_cos_kerneld.exit.i.i
  %.0 = phi double [ %.1, %__internal_sin_cos_kerneld.exit.i.i ], [ %539, %535 ], [ %spec.select4, %__internal_trig_reduction_kerneld.exit.i.i.i ]
  %603 = fmul double %458, %.0
  br label %604

604:                                              ; preds = %337, %395, %__nv_isinfd.exit, %__internal_cos_offset.exit, %366, %__nv_j0.exit
  %r.4 = phi double [ %334, %__nv_j0.exit ], [ %363, %337 ], [ %392, %366 ], [ %413, %395 ], [ %603, %__internal_cos_offset.exit ], [ 0.000000e+00, %__nv_isinfd.exit ]
  %605 = fcmp olt double %a, 0.000000e+00
  %r.5 = select i1 %605, double 0xFFF8000000000000, double %r.4
  ret double %r.5
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_y1(double %a) local_unnamed_addr #109 {
  %q.i.i.i.i5 = alloca i32, align 4
  %q.i.i.i.i.i = alloca i32, align 4
  %q.i.i.i16 = alloca i32, align 4
  %q.i.i.i = alloca i32, align 4
  %q.i.i.i.i = alloca i32, align 4
  %q.i.i = alloca i32, align 4
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp olt double %1, 0x730D67819E8D2
  br i1 %2, label %3, label %5

3:                                                ; preds = %0
  %4 = fdiv double 0xBFE45F306DC9C883, %1
  br label %604

5:                                                ; preds = %0
  %6 = fcmp ugt double %1, 0x3FF4C6F208132576
  br i1 %6, label %338, label %7

7:                                                ; preds = %5
  %8 = fmul double %a, %a
  %9 = call double @llvm.fma.f64(double %8, double noundef 0x3D5249F90687428C, double noundef 0xBDCF0B5B1FB7B95E) #140
  %10 = call double @llvm.fma.f64(double %9, double %8, double noundef 0x3E432E589311FA14) #140
  %11 = call double @llvm.fma.f64(double %10, double %8, double noundef 0xBEB0A780AA4A92E9) #140
  %12 = call double @llvm.fma.f64(double %11, double %8, double noundef 0x3F12C7DBFFCAEC2B) #140
  %13 = call double @llvm.fma.f64(double %12, double %8, double noundef 0xBF6835B97894BA4A) #140
  %14 = call double @llvm.fma.f64(double %13, double %8, double noundef 0x3FABD3975C75B4A3) #140
  %15 = call double @llvm.fma.f64(double %14, double %8, double noundef 0xBFC91866143CBC8A) #140
  %16 = fmul double %1, %15
  %17 = fcmp ugt double %1, 0x400353AABAD7B784
  br i1 %17, label %35, label %18

18:                                               ; preds = %7
  %19 = call double @llvm.fma.f64(double %1, double noundef 0x3D020E4ADCDE2AD3, double noundef 0xBD4DD167A0DC3F55) #140
  %20 = call double @llvm.fma.f64(double %19, double %1, double noundef 0x3D5503F5A491E487) #140
  %21 = call double @llvm.fma.f64(double %20, double %1, double noundef 0x3DC1F29940C2403A) #140
  %22 = call double @llvm.fma.f64(double %21, double %1, double noundef 0x3D84CF9302EACDEF) #140
  %23 = call double @llvm.fma.f64(double %22, double %1, double noundef 0xBE384A53DBBCA436) #140
  %24 = call double @llvm.fma.f64(double %23, double %1, double noundef 0x3D9779BEE4F63BCC) #140
  %25 = call double @llvm.fma.f64(double %24, double %1, double noundef 0x3EA6C160E414F3F0) #140
  %26 = call double @llvm.fma.f64(double %25, double %1, double noundef 0x3D8F3D2F12430699) #140
  %27 = call double @llvm.fma.f64(double %26, double %1, double noundef 0xBF0C71C72C0CED04) #140
  %28 = call double @llvm.fma.f64(double %27, double %1, double noundef 0x3D659BCA506F1128) #140
  %29 = call double @llvm.fma.f64(double %28, double %1, double noundef 0x3F65555555506982) #140
  %30 = call double @llvm.fma.f64(double %29, double %1, double noundef 0x3D15BA0B425F1BFB) #140
  %31 = call double @llvm.fma.f64(double %30, double %1, double noundef 0xBFB0000000000065) #140
  %32 = call double @llvm.fma.f64(double %31, double %1, double noundef 0x3C8729A7253FB679) #140
  %33 = call double @llvm.fma.f64(double %32, double %1, double noundef 5.000000e-01) #140
  %34 = fmul double %1, %33
  br label %__nv_j1.exit

35:                                               ; preds = %7
  %36 = fcmp ugt double %1, 0x4015B1D0574614EA
  br i1 %36, label %56, label %37

37:                                               ; preds = %35
  %38 = fadd double %1, 0xC00EA75575AF6F09
  %39 = fadd double %38, 0x3CA60155A9D1B256
  %40 = call double @llvm.fma.f64(double %39, double noundef 0xBCF8D3CDBB60175E, double noundef 0x3D41011A1DF02DAD) #140
  %41 = call double @llvm.fma.f64(double %40, double %39, double noundef 0x3D76013AC1E5E222) #140
  %42 = call double @llvm.fma.f64(double %41, double %39, double noundef 0xBDBEC315D96D5F03) #140
  %43 = call double @llvm.fma.f64(double %42, double %39, double noundef 0xBDF03BE1B4B57207) #140
  %44 = call double @llvm.fma.f64(double %43, double %39, double noundef 0x3E345695F8B660F7) #140
  %45 = call double @llvm.fma.f64(double %44, double %39, double noundef 0x3E617069FCFCFFF4) #140
  %46 = call double @llvm.fma.f64(double %45, double %39, double noundef 0xBEA33825C36745EB) #140
  %47 = call double @llvm.fma.f64(double %46, double %39, double noundef 0xBEC9799D4F90931B) #140
  %48 = call double @llvm.fma.f64(double %47, double %39, double noundef 0x3F083A06E2F7DF13) #140
  %49 = call double @llvm.fma.f64(double %48, double %39, double noundef 0x3F26E4C2D53A7CF6) #140
  %50 = call double @llvm.fma.f64(double %49, double %39, double noundef 0xBF624B3409957B1C) #140
  %51 = call double @llvm.fma.f64(double %50, double %39, double noundef 0xBF7537544C3325DF) #140
  %52 = call double @llvm.fma.f64(double %51, double %39, double noundef 0x3FAB589D1DA138E2) #140
  %53 = call double @llvm.fma.f64(double %52, double %39, double noundef 0x3FAAE8A39F51AD13) #140
  %54 = call double @llvm.fma.f64(double %53, double %39, double noundef 0xBFD9C6CF582CBF7F) #140
  %55 = fmul double %39, %54
  br label %__nv_j1.exit

56:                                               ; preds = %35
  %57 = fcmp ugt double %1, 0x40213065E54C1AA9
  br i1 %57, label %77, label %58

58:                                               ; preds = %56
  %59 = fadd double %1, 0xC01C0FF5F3B47250
  %60 = fadd double %59, 0x3C9B226D9D243827
  %61 = call double @llvm.fma.f64(double %60, double noundef 0x3CF3EB867515FAD6, double noundef 0xBD40E8363DB649A9) #140
  %62 = call double @llvm.fma.f64(double %61, double %60, double noundef 0xBD73B7DD4A6608FB) #140
  %63 = call double @llvm.fma.f64(double %62, double %60, double noundef 0x3DBEC5E01482C750) #140
  %64 = call double @llvm.fma.f64(double %63, double %60, double noundef 0x3DEC62BB9E882103) #140
  %65 = call double @llvm.fma.f64(double %64, double %60, double noundef 0xBE34462EED732A23) #140
  %66 = call double @llvm.fma.f64(double %65, double %60, double noundef 0xBE5D48DCAD7DC59B) #140
  %67 = call double @llvm.fma.f64(double %66, double %60, double noundef 0x3EA3026DF29167E9) #140
  %68 = call double @llvm.fma.f64(double %67, double %60, double noundef 0x3EC4255B0119666C) #140
  %69 = call double @llvm.fma.f64(double %68, double %60, double noundef 0xBF0796A751B32693) #140
  %70 = call double @llvm.fma.f64(double %69, double %60, double noundef 0xBF207358BBDBA284) #140
  %71 = call double @llvm.fma.f64(double %70, double %60, double noundef 0x3F613FBC7D6927B1) #140
  %72 = call double @llvm.fma.f64(double %71, double %60, double noundef 0x3F69A4B292E3DD75) #140
  %73 = call double @llvm.fma.f64(double %72, double %60, double noundef 0xBFA80C83BDEEE4FB) #140
  %74 = call double @llvm.fma.f64(double %73, double %60, double noundef 0xBF95E70DC60362BF) #140
  %75 = call double @llvm.fma.f64(double %74, double %60, double noundef 0x3FD33518B3874E8A) #140
  %76 = fmul double %60, %75
  br label %__nv_j1.exit

77:                                               ; preds = %56
  %78 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %79 = icmp eq i32 %78, 200
  br i1 %79, label %86, label %80

80:                                               ; preds = %77
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %82 = icmp eq i32 %81, 350
  br i1 %82, label %86, label %83

83:                                               ; preds = %80
  %84 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %85 = icmp eq i32 %84, 370
  br i1 %85, label %86, label %88

86:                                               ; preds = %83, %80, %77
  %87 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i

88:                                               ; preds = %83
  %89 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %90 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %91 = and i32 %90, 2147483647
  %92 = icmp eq i32 %91, 2146435072
  %93 = icmp eq i32 %89, 0
  %94 = select i1 %92, i1 %93, i1 false
  br label %__nv_isinfd.exit.i

__nv_isinfd.exit.i:                               ; preds = %88, %86
  %.011.in = phi i1 [ %87, %86 ], [ %94, %88 ]
  br i1 %.011.in, label %__nv_j1.exit, label %95

95:                                               ; preds = %__nv_isinfd.exit.i
  %96 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %97 = fneg double %1
  %98 = call double @llvm.fma.f64(double %97, double %96, double noundef 1.000000e+00)
  %99 = call double @llvm.fma.f64(double %98, double %98, double %98)
  %100 = call double @llvm.fma.f64(double %99, double %96, double %96)
  %101 = fmul double %100, %100
  %102 = call double @llvm.fma.f64(double %101, double noundef 0x40CD02EA3F2F6751, double noundef 0xC099C06322A3F8BE)
  %103 = call double @llvm.fma.f64(double %102, double %101, double noundef 0x405B89354DA77324)
  %104 = call double @llvm.fma.f64(double %103, double %101, double noundef 0xC01E352294653188)
  %105 = call double @llvm.fma.f64(double %104, double %101, double noundef 0x3FE9BC7DB16BD7A7)
  %106 = call double @llvm.fma.f64(double %105, double %101, double noundef 0xBFC8BFE1C3A4F741)
  %107 = call double @llvm.fma.f64(double %106, double %101, double noundef 0x3FC7FFFFF0D00BE2)
  %108 = call double @llvm.fma.f64(double %107, double %101, double noundef 0x3FF00000000068CC)
  %109 = call double @llvm.fma.f64(double %101, double noundef 0xC18DA26B212FDC9A, double noundef 0x415A30AC6857BEE0)
  %110 = call double @llvm.fma.f64(double %109, double %101, double noundef 0xC11764222AD7C910)
  %111 = call double @llvm.fma.f64(double %110, double %101, double noundef 0x40CEB02E0C306857)
  %112 = call double @llvm.fma.f64(double %111, double %101, double noundef 0xC08351859FA2B23B)
  %113 = call double @llvm.fma.f64(double %112, double %101, double noundef 0x403E65A07AF51F42)
  %114 = call double @llvm.fma.f64(double %113, double %101, double noundef 0xC002F2B817F77A57)
  %115 = call double @llvm.fma.f64(double %114, double %101, double noundef 0x3FD7BCC34DA069FD)
  %116 = call double @llvm.fma.f64(double %115, double %101, double noundef 0xBFC4FFFFF8A44463)
  %117 = call double @llvm.fma.f64(double %116, double %101, double noundef 0x3FD7FFFFFFFF5CD7)
  %118 = call double @llvm.fma.f64(double %117, double %100, double %1)
  %119 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %120 = fmul double %119, 0x3FE9884533D43651
  %121 = fmul double %120, %108
  %122 = fmul double %118, 0x3FE45F306DC9C883
  %123 = call i32 @llvm.nvvm.d2i.rn(double %122) #139
  store i32 %123, i32* %q.i.i.i16, align 4, !noalias !68
  %124 = sitofp i32 %123 to double
  %125 = fneg double %124
  %126 = call double @llvm.fma.f64(double %125, double noundef 0x3FF921FB54442D18, double %118)
  %127 = call double @llvm.fma.f64(double %125, double noundef 0x3C91A62633145C00, double %126)
  %128 = call double @llvm.fma.f64(double %125, double noundef 0x397B839A252049C0, double %127)
  %129 = call double @llvm.fabs.f64(double %118)
  %130 = fcmp ult double %129, 0x41E0000000000000
  br i1 %130, label %__internal_trig_reduction_kerneld.exit.i.i20, label %131

131:                                              ; preds = %95
  %132 = call fastcc double @__internal_trig_reduction_slowpathd(double %118, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i16) #139
  %.pre = load i32, i32* %q.i.i.i16, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i20

__internal_trig_reduction_kerneld.exit.i.i20:     ; preds = %131, %95
  %133 = phi i32 [ %.pre, %131 ], [ %123, %95 ]
  %t.i.i.i15.0 = phi double [ %132, %131 ], [ %128, %95 ]
  %134 = and i32 %133, 3
  %135 = uitofp i32 %134 to double
  %136 = fadd double %t.i.i.i15.0, 0xC002D97C7F3321D2
  %137 = fmul double %135, 0x3FF921FB54442D18
  %138 = fadd double %136, %137
  %139 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %140 = icmp eq i32 %139, 350
  br i1 %140, label %144, label %141

141:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i20
  %142 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %143 = icmp eq i32 %142, 370
  br i1 %143, label %144, label %203

144:                                              ; preds = %141, %__internal_trig_reduction_kerneld.exit.i.i20
  %145 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %146 = icmp eq i32 %145, 200
  br i1 %146, label %153, label %147

147:                                              ; preds = %144
  %148 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %149 = icmp eq i32 %148, 350
  br i1 %149, label %153, label %150

150:                                              ; preds = %147
  %151 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %152 = icmp eq i32 %151, 370
  br i1 %152, label %153, label %156

153:                                              ; preds = %150, %147, %144
  %154 = call double @llvm.fabs.f64(double %138)
  %155 = fcmp oeq double %154, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i

156:                                              ; preds = %150
  %157 = call i32 @llvm.nvvm.d2i.lo(double %138) #139
  %158 = call i32 @llvm.nvvm.d2i.hi(double %138) #139
  %159 = and i32 %158, 2147483647
  %160 = icmp eq i32 %159, 2146435072
  %161 = icmp eq i32 %157, 0
  %162 = select i1 %160, i1 %161, i1 false
  br label %__nv_isinfd.exit.i.i.i.i

__nv_isinfd.exit.i.i.i.i:                         ; preds = %156, %153
  %.07.in = phi i1 [ %155, %153 ], [ %162, %156 ]
  br i1 %.07.in, label %163, label %165

163:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %164 = fmul double %138, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

165:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %166 = fmul double %138, 0x3FE45F306DC9C883
  %167 = call i32 @llvm.nvvm.d2i.rn(double %166) #139
  store i32 %167, i32* %q.i.i.i.i.i, align 4, !noalias !68
  %168 = sitofp i32 %167 to double
  %169 = fneg double %168
  %170 = call double @llvm.fma.f64(double %169, double noundef 0x3FF921FB54442D18, double %138)
  %171 = call double @llvm.fma.f64(double %169, double noundef 0x3C91A62633145C00, double %170)
  %172 = call double @llvm.fma.f64(double %169, double noundef 0x397B839A252049C0, double %171)
  %173 = call double @llvm.fabs.f64(double %138)
  %174 = fcmp ult double %173, 0x41E0000000000000
  br i1 %174, label %__internal_trig_reduction_kerneld.exit.i.i.i.i, label %175

175:                                              ; preds = %165
  %176 = call fastcc double @__internal_trig_reduction_slowpathd(double %138, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i) #139
  %.pre17 = load i32, i32* %q.i.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i.i:   ; preds = %165, %175, %163
  %i.i.i.i.i.0 = phi i32 [ 0, %163 ], [ %.pre17, %175 ], [ %167, %165 ]
  %.06 = phi double [ %164, %163 ], [ %176, %175 ], [ %172, %165 ]
  %177 = fmul double %.06, %.06
  %178 = call double @llvm.fma.f64(double %177, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %179 = call double @llvm.fma.f64(double %178, double %177, double noundef 0xBE927E4F8E06E6D9)
  %180 = call double @llvm.fma.f64(double %179, double %177, double noundef 0x3EFA01A019DDBCE9)
  %181 = call double @llvm.fma.f64(double %180, double %177, double noundef 0xBF56C16C16C15D47)
  %182 = call double @llvm.fma.f64(double %181, double %177, double noundef 0x3FA5555555555551)
  %183 = call double @llvm.fma.f64(double %182, double %177, double noundef -5.000000e-01)
  %184 = call double @llvm.fma.f64(double %183, double %177, double noundef 1.000000e+00)
  %185 = call double @llvm.fma.f64(double %177, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %186 = call double @llvm.fma.f64(double %185, double %177, double noundef 0x3EC71DE369ACE392)
  %187 = call double @llvm.fma.f64(double %186, double %177, double noundef 0xBF2A01A019DB62A1)
  %188 = call double @llvm.fma.f64(double %187, double %177, double noundef 0x3F81111111110818)
  %189 = call double @llvm.fma.f64(double %188, double %177, double noundef 0xBFC5555555555554)
  %190 = call double @llvm.fma.f64(double %189, double %177, double noundef 0.000000e+00)
  %191 = call double @llvm.fma.f64(double %190, double %.06, double %.06)
  %192 = call i32 @llvm.nvvm.d2i.hi(double %191) #139
  %193 = call i32 @llvm.nvvm.d2i.lo(double %191) #139
  %194 = xor i32 %192, -2147483648
  %195 = call double @llvm.nvvm.lohi.i2d(i32 %193, i32 %194) #139
  %196 = and i32 %i.i.i.i.i.0, 1
  %.not14 = icmp eq i32 %196, 0
  %spec.select1 = select i1 %.not14, double %184, double %195
  %197 = and i32 %i.i.i.i.i.0, 2
  %.not15 = icmp eq i32 %197, 0
  br i1 %.not15, label %__internal_cos_offset.exit.i, label %198

198:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i
  %199 = call i32 @llvm.nvvm.d2i.hi(double %spec.select1) #139
  %200 = call i32 @llvm.nvvm.d2i.lo(double %spec.select1) #139
  %201 = xor i32 %199, -2147483648
  %202 = call double @llvm.nvvm.lohi.i2d(i32 %200, i32 %201) #139
  br label %__internal_cos_offset.exit.i

203:                                              ; preds = %141
  %204 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %205 = icmp eq i32 %204, 200
  br i1 %205, label %212, label %206

206:                                              ; preds = %203
  %207 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %208 = icmp eq i32 %207, 350
  br i1 %208, label %212, label %209

209:                                              ; preds = %206
  %210 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %211 = icmp eq i32 %210, 370
  br i1 %211, label %212, label %215

212:                                              ; preds = %209, %206, %203
  %213 = call double @llvm.fabs.f64(double %138)
  %214 = fcmp oeq double %213, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i21

215:                                              ; preds = %209
  %216 = call i32 @llvm.nvvm.d2i.lo(double %138) #139
  %217 = call i32 @llvm.nvvm.d2i.hi(double %138) #139
  %218 = and i32 %217, 2147483647
  %219 = icmp eq i32 %218, 2146435072
  %220 = icmp eq i32 %216, 0
  %221 = select i1 %219, i1 %220, i1 false
  br label %__nv_isinfd.exit.i.i.i21

__nv_isinfd.exit.i.i.i21:                         ; preds = %215, %212
  %.08.in = phi i1 [ %214, %212 ], [ %221, %215 ]
  br i1 %.08.in, label %222, label %224

222:                                              ; preds = %__nv_isinfd.exit.i.i.i21
  %223 = fmul double %138, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i

224:                                              ; preds = %__nv_isinfd.exit.i.i.i21
  %225 = fmul double %138, 0x3FE45F306DC9C883
  %226 = call i32 @llvm.nvvm.d2i.rn(double %225) #139
  store i32 %226, i32* %q.i.i.i.i5, align 4, !noalias !68
  %227 = sitofp i32 %226 to double
  %228 = fneg double %227
  %229 = call double @llvm.fma.f64(double %228, double noundef 0x3FF921FB54442D18, double %138)
  %230 = call double @llvm.fma.f64(double %228, double noundef 0x3C91A62633145C00, double %229)
  %231 = call double @llvm.fma.f64(double %228, double noundef 0x397B839A252049C0, double %230)
  %232 = call double @llvm.fabs.f64(double %138)
  %233 = fcmp ult double %232, 0x41E0000000000000
  br i1 %233, label %__internal_trig_reduction_kerneld.exit.i.i.i22, label %234

234:                                              ; preds = %224
  %235 = call fastcc double @__internal_trig_reduction_slowpathd(double %138, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i5) #139
  %.pre16 = load i32, i32* %q.i.i.i.i5, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i22

__internal_trig_reduction_kerneld.exit.i.i.i22:   ; preds = %234, %224
  %236 = phi i32 [ %.pre16, %234 ], [ %226, %224 ]
  %t.i1.i.i.i.0 = phi double [ %235, %234 ], [ %231, %224 ]
  %phi.bo10 = add i32 %236, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i

__internal_sin_cos_kerneld.exit.i.i.i:            ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i22, %222
  %i.i.i.i13.0 = phi i32 [ 1, %222 ], [ %phi.bo10, %__internal_trig_reduction_kerneld.exit.i.i.i22 ]
  %z.i.i.i12.0 = phi double [ %223, %222 ], [ %t.i1.i.i.i.0, %__internal_trig_reduction_kerneld.exit.i.i.i22 ]
  %237 = shl i32 %i.i.i.i13.0, 3
  %238 = and i32 %237, 8
  %239 = zext i32 %238 to i64
  %240 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %239
  %241 = fmul double %z.i.i.i12.0, %z.i.i.i12.0
  %242 = and i32 %i.i.i.i13.0, 1
  %.not11 = icmp eq i32 %242, 0
  %243 = select i1 %.not11, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %244 = getelementptr inbounds double, double addrspace(1)* %240, i64 1
  %245 = load double, double addrspace(1)* %244, align 8
  %246 = call double @llvm.fma.f64(double noundef %243, double %241, double %245)
  %247 = getelementptr inbounds double, double addrspace(1)* %240, i64 2
  %248 = load double, double addrspace(1)* %247, align 8
  %249 = call double @llvm.fma.f64(double %246, double %241, double %248)
  %250 = getelementptr inbounds double, double addrspace(1)* %240, i64 3
  %251 = load double, double addrspace(1)* %250, align 8
  %252 = call double @llvm.fma.f64(double %249, double %241, double %251)
  %253 = getelementptr inbounds double, double addrspace(1)* %240, i64 4
  %254 = load double, double addrspace(1)* %253, align 8
  %255 = call double @llvm.fma.f64(double %252, double %241, double %254)
  %256 = getelementptr inbounds double, double addrspace(1)* %240, i64 5
  %257 = load double, double addrspace(1)* %256, align 8
  %258 = call double @llvm.fma.f64(double %255, double %241, double %257)
  %259 = getelementptr inbounds double, double addrspace(1)* %240, i64 6
  %260 = load double, double addrspace(1)* %259, align 8
  %261 = call double @llvm.fma.f64(double %258, double %241, double %260)
  %262 = call double @llvm.fma.f64(double %261, double %z.i.i.i12.0, double %z.i.i.i12.0)
  %263 = call double @llvm.fma.f64(double %261, double %241, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not11, double %262, double %263
  %264 = and i32 %i.i.i.i13.0, 2
  %.not13 = icmp eq i32 %264, 0
  %265 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.110 = select i1 %.not13, double %spec.select2, double %265
  br label %__internal_cos_offset.exit.i

__internal_cos_offset.exit.i:                     ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i, %198, %__internal_sin_cos_kerneld.exit.i.i.i
  %.05 = phi double [ %.110, %__internal_sin_cos_kerneld.exit.i.i.i ], [ %202, %198 ], [ %spec.select1, %__internal_trig_reduction_kerneld.exit.i.i.i.i ]
  %266 = fmul double %121, %.05
  br label %__nv_j1.exit

__nv_j1.exit:                                     ; preds = %37, %__internal_cos_offset.exit.i, %__nv_isinfd.exit.i, %58, %18
  %r.i.3 = phi double [ %34, %18 ], [ %55, %37 ], [ %76, %58 ], [ %266, %__internal_cos_offset.exit.i ], [ 0.000000e+00, %__nv_isinfd.exit.i ]
  %267 = fcmp olt double %1, 1.000000e-30
  %268 = fmul double %1, 5.000000e-01
  %r.i.5 = select i1 %267, double %268, double %r.i.3
  %269 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %270 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %271 = icmp slt i32 %269, 1048576
  br i1 %271, label %272, label %276

272:                                              ; preds = %__nv_j1.exit
  %273 = fmul double %1, 0x4350000000000000
  %274 = call i32 @llvm.nvvm.d2i.hi(double %273) #139
  %275 = call i32 @llvm.nvvm.d2i.lo(double %273) #139
  br label %276

276:                                              ; preds = %272, %__nv_j1.exit
  %.012 = phi double [ %273, %272 ], [ %1, %__nv_j1.exit ]
  %ihi.i.0 = phi i32 [ %274, %272 ], [ %269, %__nv_j1.exit ]
  %ilo.i.0 = phi i32 [ %275, %272 ], [ %270, %__nv_j1.exit ]
  %e.i.0 = phi i32 [ -1077, %272 ], [ -1023, %__nv_j1.exit ]
  %277 = add i32 %ihi.i.0, -1
  %278 = icmp ult i32 %277, 2146435071
  br i1 %278, label %279, label %328

279:                                              ; preds = %276
  %280 = lshr i32 %ihi.i.0, 20
  %281 = add nsw i32 %e.i.0, %280
  %282 = and i32 %ihi.i.0, -2146435073
  %283 = or i32 %282, 1072693248
  %284 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %283) #139
  %285 = icmp sgt i32 %283, 1073127582
  br i1 %285, label %286, label %292

286:                                              ; preds = %279
  %287 = call i32 @llvm.nvvm.d2i.lo(double %284) #139
  %288 = call i32 @llvm.nvvm.d2i.hi(double %284) #139
  %289 = add i32 %288, -1048576
  %290 = call double @llvm.nvvm.lohi.i2d(i32 %287, i32 %289) #139
  %291 = add nsw i32 %281, 1
  br label %292

292:                                              ; preds = %286, %279
  %m.i.0 = phi double [ %290, %286 ], [ %284, %279 ]
  %e.i.1 = phi i32 [ %291, %286 ], [ %281, %279 ]
  %293 = fadd double %m.i.0, -1.000000e+00
  %294 = fadd double %m.i.0, 1.000000e+00
  %295 = call double @llvm.nvvm.rcp.approx.ftz.d(double %294) #139
  %296 = fneg double %294
  %297 = call double @llvm.fma.f64(double %296, double %295, double noundef 1.000000e+00)
  %298 = call double @llvm.fma.f64(double %297, double %297, double %297)
  %299 = call double @llvm.fma.f64(double %298, double %295, double %295)
  %300 = fmul double %293, %299
  %301 = fadd double %300, %300
  %302 = fmul double %301, %301
  %303 = call double @llvm.fma.f64(double %302, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %304 = call double @llvm.fma.f64(double %303, double %302, double noundef 0x3EF3B2669F02676F)
  %305 = call double @llvm.fma.f64(double %304, double %302, double noundef 0x3F1745CBA9AB0956)
  %306 = call double @llvm.fma.f64(double %305, double %302, double noundef 0x3F3C71C72D1B5154)
  %307 = call double @llvm.fma.f64(double %306, double %302, double noundef 0x3F624924923BE72D)
  %308 = call double @llvm.fma.f64(double %307, double %302, double noundef 0x3F8999999999A3C4)
  %309 = call double @llvm.fma.f64(double %308, double %302, double noundef 0x3FB5555555555554)
  %310 = fsub double %293, %301
  %311 = fmul double %310, 2.000000e+00
  %312 = fneg double %301
  %313 = call double @llvm.fma.f64(double %312, double %293, double %311)
  %314 = fmul double %299, %313
  %315 = fmul double %302, %309
  %316 = call double @llvm.fma.f64(double %315, double %301, double %314)
  %317 = xor i32 %e.i.1, -2147483648
  %318 = call double @llvm.nvvm.lohi.i2d(i32 %317, i32 noundef 1127219200) #139
  %319 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %320 = fsub double %318, %319
  %321 = call double @llvm.fma.f64(double %320, double noundef 0x3FE62E42FEFA39EF, double %301)
  %322 = fneg double %320
  %323 = call double @llvm.fma.f64(double %322, double noundef 0x3FE62E42FEFA39EF, double %321)
  %324 = fsub double %323, %301
  %325 = fsub double %316, %324
  %326 = call double @llvm.fma.f64(double %320, double noundef 0x3C7ABC9E3B39803F, double %325)
  %327 = fadd double %321, %326
  br label %__nv_log.exit

328:                                              ; preds = %276
  %329 = call double @llvm.fma.f64(double %.012, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %330 = call i32 @llvm.nvvm.d2i.hi(double %.012) #139
  %331 = bitcast i32 %330 to float
  %332 = fcmp oeq float %331, 0.000000e+00
  %q.i28.0 = select i1 %332, double 0xFFF0000000000000, double %329
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %328, %292
  %q.i28.1 = phi double [ %327, %292 ], [ %q.i28.0, %328 ]
  %333 = fmul double %r.i.5, %q.i28.1
  %334 = fdiv double 1.000000e+00, %1
  %335 = fsub double %333, %334
  %336 = fmul double %335, 0x3FE45F306DC9C883
  %337 = fadd double %16, %336
  br label %604

338:                                              ; preds = %5
  %339 = fcmp ugt double %1, 0x4009B510EC2ADC83
  br i1 %339, label %367, label %340

340:                                              ; preds = %338
  %341 = fadd double %1, 0xC00193BED4DFF243
  %342 = fadd double %341, 0x3C8BD1E50D219BFD
  %343 = call double @llvm.fma.f64(double %342, double noundef 0xBE2B87B0BE2AA150, double noundef 0x3E4833AAE4D8B975)
  %344 = call double @llvm.fma.f64(double %343, double %342, double noundef 0xBE44E279B423E68F)
  %345 = call double @llvm.fma.f64(double %344, double %342, double noundef 0x3E129DC6A747EB4F)
  %346 = call double @llvm.fma.f64(double %345, double %342, double noundef 0xBE61D15534496CD8)
  %347 = call double @llvm.fma.f64(double %346, double %342, double noundef 0x3E7EEC8D48FECE00)
  %348 = call double @llvm.fma.f64(double %347, double %342, double noundef 0xBE8D1180AF70A134)
  %349 = call double @llvm.fma.f64(double %348, double %342, double noundef 0x3E9C8386A0EA1388)
  %350 = call double @llvm.fma.f64(double %349, double %342, double noundef 0xBEB01A014E7F3250)
  %351 = call double @llvm.fma.f64(double %350, double %342, double noundef 0x3EC1FB752010A320)
  %352 = call double @llvm.fma.f64(double %351, double %342, double noundef 0xBED3AA0AFF4E332B)
  %353 = call double @llvm.fma.f64(double %352, double %342, double noundef 0x3EE584A6C77F6700)
  %354 = call double @llvm.fma.f64(double %353, double %342, double noundef 0xBEF794C520FC2EBB)
  %355 = call double @llvm.fma.f64(double %354, double %342, double noundef 0x3F09D18D2D35CC71)
  %356 = call double @llvm.fma.f64(double %355, double %342, double noundef 0xBF1C3FB7315C4599)
  %357 = call double @llvm.fma.f64(double %356, double %342, double noundef 0x3F2EEA7ADECCE927)
  %358 = call double @llvm.fma.f64(double %357, double %342, double noundef 0xBF40B2D85257446F)
  %359 = call double @llvm.fma.f64(double %358, double %342, double noundef 0x3F517AB4B1FE5D5B)
  %360 = call double @llvm.fma.f64(double %359, double %342, double noundef 0xBF65429DC6516C0D)
  %361 = call double @llvm.fma.f64(double %360, double %342, double noundef 0x3F7E671C7D0B090B)
  %362 = call double @llvm.fma.f64(double %361, double %342, double noundef 0xBF73A6DEC36FB27C)
  %363 = call double @llvm.fma.f64(double %362, double %342, double noundef 0xBFA0D2AF4E931FD1)
  %364 = call double @llvm.fma.f64(double %363, double %342, double noundef 0xBFBE56F82217B964)
  %365 = call double @llvm.fma.f64(double %364, double %342, double noundef 0x3FE0AA48442F014B)
  %366 = fmul double %342, %365
  br label %604

367:                                              ; preds = %338
  %368 = fcmp ugt double %1, 0x401C0D26D5A541CB
  br i1 %368, label %393, label %369

369:                                              ; preds = %367
  %370 = fadd double %1, 0xC015B7FE4E87B02E
  %371 = fadd double %370, 0xBCBDFE7BAC228E8C
  %372 = call double @llvm.fma.f64(double %371, double noundef 0x3CBA3C76069F1D8C, double noundef 0x3CC69A30996793E2)
  %373 = call double @llvm.fma.f64(double %372, double %371, double noundef 0xBCDDD8432FE756E7)
  %374 = call double @llvm.fma.f64(double %373, double %371, double noundef 0xBD143158EEE220F7)
  %375 = call double @llvm.fma.f64(double %374, double %371, double noundef 0x3D28D44491230F5A)
  %376 = call double @llvm.fma.f64(double %375, double %371, double noundef 0xBD438842EAF4EDBC)
  %377 = call double @llvm.fma.f64(double %376, double %371, double noundef 0x3D74958DAFBFAF5A)
  %378 = call double @llvm.fma.f64(double %377, double %371, double noundef 0xBD9449A60E664848)
  %379 = call double @llvm.fma.f64(double %378, double %371, double noundef 0x3D838BC8CD594A76)
  %380 = call double @llvm.fma.f64(double %379, double %371, double noundef 0xBDFA798002141323)
  %381 = call double @llvm.fma.f64(double %380, double %371, double noundef 0x3E380B4198956AAA)
  %382 = call double @llvm.fma.f64(double %381, double %371, double noundef 0x3E5B62B5F21BACD4)
  %383 = call double @llvm.fma.f64(double %382, double %371, double noundef 0xBEA255E729FB6AAE)
  %384 = call double @llvm.fma.f64(double %383, double %371, double noundef 0xBEC80618F6BAE5AA)
  %385 = call double @llvm.fma.f64(double %384, double %371, double noundef 0x3F085B940F8E8D36)
  %386 = call double @llvm.fma.f64(double %385, double %371, double noundef 0x3F2337C7E10E14E8)
  %387 = call double @llvm.fma.f64(double %386, double %371, double noundef 0xBF61BE6DB99332CA)
  %388 = call double @llvm.fma.f64(double %387, double %371, double noundef 0xBF710A329E2BE9B8)
  %389 = call double @llvm.fma.f64(double %388, double %371, double noundef 0x3FAA15D92DFE3FCF)
  %390 = call double @llvm.fma.f64(double %389, double %371, double noundef 0x3FA00B9F8571C9BE)
  %391 = call double @llvm.fma.f64(double %390, double %371, double noundef 0xBFD5C7C556F0C19A)
  %392 = fmul double %371, %391
  br label %604

393:                                              ; preds = %367
  %394 = fcmp ugt double %1, 0x4022585C739ACDDD
  br i1 %394, label %414, label %395

395:                                              ; preds = %393
  %396 = fadd double %1, 0xC0213127AE6169B4
  %397 = fadd double %396, 0xBCB479CC068D9046
  %398 = call double @llvm.fma.f64(double %397, double noundef 0x3CB09CCC22945996, double noundef 0xBD43515F67644276)
  %399 = call double @llvm.fma.f64(double %398, double %397, double noundef 0xBD72C5B978E9F5C7)
  %400 = call double @llvm.fma.f64(double %399, double %397, double noundef 0x3DBEC1151613913C)
  %401 = call double @llvm.fma.f64(double %400, double %397, double noundef 0x3DE9E38D13C4A824)
  %402 = call double @llvm.fma.f64(double %401, double %397, double noundef 0xBE341E75E1088EB5)
  %403 = call double @llvm.fma.f64(double %402, double %397, double noundef 0xBE5A384EBB13CFE1)
  %404 = call double @llvm.fma.f64(double %403, double %397, double noundef 0x3EA2BECB27F8C8F8)
  %405 = call double @llvm.fma.f64(double %404, double %397, double noundef 0x3EC176E72B989FD8)
  %406 = call double @llvm.fma.f64(double %405, double %397, double noundef 0xBF06F7BAB102F822)
  %407 = call double @llvm.fma.f64(double %406, double %397, double noundef 0xBF1B50D7E1D278E1)
  %408 = call double @llvm.fma.f64(double %407, double %397, double noundef 0x3F607A678D60004F)
  %409 = call double @llvm.fma.f64(double %408, double %397, double noundef 0x3F63CED2A2E69115)
  %410 = call double @llvm.fma.f64(double %409, double %397, double noundef 0xBFA6395DFE49FCD4)
  %411 = call double @llvm.fma.f64(double %410, double %397, double noundef 0xBF902B3933CF21B1)
  %412 = call double @llvm.fma.f64(double %411, double %397, double noundef 0x3FD15F993FCEAB5C)
  %413 = fmul double %397, %412
  br label %604

414:                                              ; preds = %393
  %415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %416 = icmp eq i32 %415, 200
  br i1 %416, label %423, label %417

417:                                              ; preds = %414
  %418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %419 = icmp eq i32 %418, 350
  br i1 %419, label %423, label %420

420:                                              ; preds = %417
  %421 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %422 = icmp eq i32 %421, 370
  br i1 %422, label %423, label %425

423:                                              ; preds = %420, %417, %414
  %424 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit

425:                                              ; preds = %420
  %426 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %427 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %428 = and i32 %427, 2147483647
  %429 = icmp eq i32 %428, 2146435072
  %430 = icmp eq i32 %426, 0
  %431 = select i1 %429, i1 %430, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %425, %423
  %.013.in = phi i1 [ %424, %423 ], [ %431, %425 ]
  br i1 %.013.in, label %604, label %432

432:                                              ; preds = %__nv_isinfd.exit
  %433 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %434 = fneg double %1
  %435 = call double @llvm.fma.f64(double %434, double %433, double noundef 1.000000e+00)
  %436 = call double @llvm.fma.f64(double %435, double %435, double %435)
  %437 = call double @llvm.fma.f64(double %436, double %433, double %433)
  %438 = fmul double %437, %437
  %439 = call double @llvm.fma.f64(double %438, double noundef 0x40D13DB326ECEBFE, double noundef 0xC09C26E89385D5B1)
  %440 = call double @llvm.fma.f64(double %439, double %438, double noundef 0x405C6AB923C6F55E)
  %441 = call double @llvm.fma.f64(double %440, double %438, double noundef 0xC01E61EAF3BD2FA1)
  %442 = call double @llvm.fma.f64(double %441, double %438, double noundef 0x3FE9BF15D9B97DD1)
  %443 = call double @llvm.fma.f64(double %442, double %438, double noundef 0xBFC8BFECF93D7D19)
  %444 = call double @llvm.fma.f64(double %443, double %438, double noundef 0x3FC7FFFFF756AA6C)
  %445 = call double @llvm.fma.f64(double %444, double %438, double noundef 0x3FF0000000003646)
  %446 = call double @llvm.fma.f64(double %438, double noundef 0xC1943281A050209C, double noundef 0x416024E99BA46E7B)
  %447 = call double @llvm.fma.f64(double %446, double %438, double noundef 0xC11A6875D7DFBD65)
  %448 = call double @llvm.fma.f64(double %447, double %438, double noundef 0x40D032C041790233)
  %449 = call double @llvm.fma.f64(double %448, double %438, double noundef 0xC0839F895BC22946)
  %450 = call double @llvm.fma.f64(double %449, double %438, double noundef 0x403E77CC78ECD2D8)
  %451 = call double @llvm.fma.f64(double %450, double %438, double noundef 0xC002F368D0117BE9)
  %452 = call double @llvm.fma.f64(double %451, double %438, double noundef 0x3FD7BCC786009A25)
  %453 = call double @llvm.fma.f64(double %452, double %438, double noundef 0xBFC4FFFFFC51BC7A)
  %454 = call double @llvm.fma.f64(double %453, double %438, double noundef 0x3FD7FFFFFFFFB5EA)
  %455 = call double @llvm.fma.f64(double %454, double %437, double %1)
  %456 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %457 = fmul double %456, 0x3FE9884533D43651
  %458 = fmul double %457, %445
  %459 = fmul double %455, 0x3FE45F306DC9C883
  %460 = call i32 @llvm.nvvm.d2i.rn(double %459) #139
  store i32 %460, i32* %q.i.i, align 4, !noalias !68
  %461 = sitofp i32 %460 to double
  %462 = fneg double %461
  %463 = call double @llvm.fma.f64(double %462, double noundef 0x3FF921FB54442D18, double %455)
  %464 = call double @llvm.fma.f64(double %462, double noundef 0x3C91A62633145C00, double %463)
  %465 = call double @llvm.fma.f64(double %462, double noundef 0x397B839A252049C0, double %464)
  %466 = call double @llvm.fabs.f64(double %455)
  %467 = fcmp ult double %466, 0x41E0000000000000
  br i1 %467, label %__internal_trig_reduction_kerneld.exit.i, label %468

468:                                              ; preds = %432
  %469 = call fastcc double @__internal_trig_reduction_slowpathd(double %455, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i) #139
  %.pre18 = load i32, i32* %q.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i

__internal_trig_reduction_kerneld.exit.i:         ; preds = %468, %432
  %470 = phi i32 [ %.pre18, %468 ], [ %460, %432 ]
  %t.i.i.0 = phi double [ %469, %468 ], [ %465, %432 ]
  %471 = and i32 %470, 3
  %472 = uitofp i32 %471 to double
  %473 = fadd double %t.i.i.0, 0xC00F6A7A2955385E
  %474 = fmul double %472, 0x3FF921FB54442D18
  %475 = fadd double %473, %474
  %476 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %477 = icmp eq i32 %476, 350
  br i1 %477, label %481, label %478

478:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i
  %479 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %480 = icmp eq i32 %479, 370
  br i1 %480, label %481, label %540

481:                                              ; preds = %478, %__internal_trig_reduction_kerneld.exit.i
  %482 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %483 = icmp eq i32 %482, 200
  br i1 %483, label %490, label %484

484:                                              ; preds = %481
  %485 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %486 = icmp eq i32 %485, 350
  br i1 %486, label %490, label %487

487:                                              ; preds = %484
  %488 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %489 = icmp eq i32 %488, 370
  br i1 %489, label %490, label %493

490:                                              ; preds = %487, %484, %481
  %491 = call double @llvm.fabs.f64(double %475)
  %492 = fcmp oeq double %491, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i

493:                                              ; preds = %487
  %494 = call i32 @llvm.nvvm.d2i.lo(double %475) #139
  %495 = call i32 @llvm.nvvm.d2i.hi(double %475) #139
  %496 = and i32 %495, 2147483647
  %497 = icmp eq i32 %496, 2146435072
  %498 = icmp eq i32 %494, 0
  %499 = select i1 %497, i1 %498, i1 false
  br label %__nv_isinfd.exit.i.i.i

__nv_isinfd.exit.i.i.i:                           ; preds = %493, %490
  %.02.in = phi i1 [ %492, %490 ], [ %499, %493 ]
  br i1 %.02.in, label %500, label %502

500:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %501 = fmul double %475, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

502:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %503 = fmul double %475, 0x3FE45F306DC9C883
  %504 = call i32 @llvm.nvvm.d2i.rn(double %503) #139
  store i32 %504, i32* %q.i.i.i.i, align 4, !noalias !68
  %505 = sitofp i32 %504 to double
  %506 = fneg double %505
  %507 = call double @llvm.fma.f64(double %506, double noundef 0x3FF921FB54442D18, double %475)
  %508 = call double @llvm.fma.f64(double %506, double noundef 0x3C91A62633145C00, double %507)
  %509 = call double @llvm.fma.f64(double %506, double noundef 0x397B839A252049C0, double %508)
  %510 = call double @llvm.fabs.f64(double %475)
  %511 = fcmp ult double %510, 0x41E0000000000000
  br i1 %511, label %__internal_trig_reduction_kerneld.exit.i.i.i, label %512

512:                                              ; preds = %502
  %513 = call fastcc double @__internal_trig_reduction_slowpathd(double %475, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i) #139
  %.pre20 = load i32, i32* %q.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i:     ; preds = %502, %512, %500
  %i.i.i.i.0 = phi i32 [ 0, %500 ], [ %.pre20, %512 ], [ %504, %502 ]
  %.01 = phi double [ %501, %500 ], [ %513, %512 ], [ %509, %502 ]
  %514 = fmul double %.01, %.01
  %515 = call double @llvm.fma.f64(double %514, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %516 = call double @llvm.fma.f64(double %515, double %514, double noundef 0xBE927E4F8E06E6D9)
  %517 = call double @llvm.fma.f64(double %516, double %514, double noundef 0x3EFA01A019DDBCE9)
  %518 = call double @llvm.fma.f64(double %517, double %514, double noundef 0xBF56C16C16C15D47)
  %519 = call double @llvm.fma.f64(double %518, double %514, double noundef 0x3FA5555555555551)
  %520 = call double @llvm.fma.f64(double %519, double %514, double noundef -5.000000e-01)
  %521 = call double @llvm.fma.f64(double %520, double %514, double noundef 1.000000e+00)
  %522 = call double @llvm.fma.f64(double %514, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %523 = call double @llvm.fma.f64(double %522, double %514, double noundef 0x3EC71DE369ACE392)
  %524 = call double @llvm.fma.f64(double %523, double %514, double noundef 0xBF2A01A019DB62A1)
  %525 = call double @llvm.fma.f64(double %524, double %514, double noundef 0x3F81111111110818)
  %526 = call double @llvm.fma.f64(double %525, double %514, double noundef 0xBFC5555555555554)
  %527 = call double @llvm.fma.f64(double %526, double %514, double noundef 0.000000e+00)
  %528 = call double @llvm.fma.f64(double %527, double %.01, double %.01)
  %529 = call i32 @llvm.nvvm.d2i.hi(double %528) #139
  %530 = call i32 @llvm.nvvm.d2i.lo(double %528) #139
  %531 = xor i32 %529, -2147483648
  %532 = call double @llvm.nvvm.lohi.i2d(i32 %530, i32 %531) #139
  %533 = and i32 %i.i.i.i.0, 1
  %.not8 = icmp eq i32 %533, 0
  %spec.select4 = select i1 %.not8, double %521, double %532
  %534 = and i32 %i.i.i.i.0, 2
  %.not9 = icmp eq i32 %534, 0
  br i1 %.not9, label %__internal_cos_offset.exit, label %535

535:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i
  %536 = call i32 @llvm.nvvm.d2i.hi(double %spec.select4) #139
  %537 = call i32 @llvm.nvvm.d2i.lo(double %spec.select4) #139
  %538 = xor i32 %536, -2147483648
  %539 = call double @llvm.nvvm.lohi.i2d(i32 %537, i32 %538) #139
  br label %__internal_cos_offset.exit

540:                                              ; preds = %478
  %541 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %542 = icmp eq i32 %541, 200
  br i1 %542, label %549, label %543

543:                                              ; preds = %540
  %544 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %545 = icmp eq i32 %544, 350
  br i1 %545, label %549, label %546

546:                                              ; preds = %543
  %547 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %548 = icmp eq i32 %547, 370
  br i1 %548, label %549, label %552

549:                                              ; preds = %546, %543, %540
  %550 = call double @llvm.fabs.f64(double %475)
  %551 = fcmp oeq double %550, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i

552:                                              ; preds = %546
  %553 = call i32 @llvm.nvvm.d2i.lo(double %475) #139
  %554 = call i32 @llvm.nvvm.d2i.hi(double %475) #139
  %555 = and i32 %554, 2147483647
  %556 = icmp eq i32 %555, 2146435072
  %557 = icmp eq i32 %553, 0
  %558 = select i1 %556, i1 %557, i1 false
  br label %__nv_isinfd.exit.i.i

__nv_isinfd.exit.i.i:                             ; preds = %552, %549
  %.03.in = phi i1 [ %551, %549 ], [ %558, %552 ]
  br i1 %.03.in, label %559, label %561

559:                                              ; preds = %__nv_isinfd.exit.i.i
  %560 = fmul double %475, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i

561:                                              ; preds = %__nv_isinfd.exit.i.i
  %562 = fmul double %475, 0x3FE45F306DC9C883
  %563 = call i32 @llvm.nvvm.d2i.rn(double %562) #139
  store i32 %563, i32* %q.i.i.i, align 4, !noalias !68
  %564 = sitofp i32 %563 to double
  %565 = fneg double %564
  %566 = call double @llvm.fma.f64(double %565, double noundef 0x3FF921FB54442D18, double %475)
  %567 = call double @llvm.fma.f64(double %565, double noundef 0x3C91A62633145C00, double %566)
  %568 = call double @llvm.fma.f64(double %565, double noundef 0x397B839A252049C0, double %567)
  %569 = call double @llvm.fabs.f64(double %475)
  %570 = fcmp ult double %569, 0x41E0000000000000
  br i1 %570, label %__internal_trig_reduction_kerneld.exit.i.i, label %571

571:                                              ; preds = %561
  %572 = call fastcc double @__internal_trig_reduction_slowpathd(double %475, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i) #139
  %.pre19 = load i32, i32* %q.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i

__internal_trig_reduction_kerneld.exit.i.i:       ; preds = %571, %561
  %573 = phi i32 [ %.pre19, %571 ], [ %563, %561 ]
  %t.i1.i.i.0 = phi double [ %572, %571 ], [ %568, %561 ]
  %phi.bo = add i32 %573, 1
  br label %__internal_sin_cos_kerneld.exit.i.i

__internal_sin_cos_kerneld.exit.i.i:              ; preds = %__internal_trig_reduction_kerneld.exit.i.i, %559
  %i.i.i.0 = phi i32 [ 1, %559 ], [ %phi.bo, %__internal_trig_reduction_kerneld.exit.i.i ]
  %z.i.i.0 = phi double [ %560, %559 ], [ %t.i1.i.i.0, %__internal_trig_reduction_kerneld.exit.i.i ]
  %574 = shl i32 %i.i.i.0, 3
  %575 = and i32 %574, 8
  %576 = zext i32 %575 to i64
  %577 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %576
  %578 = fmul double %z.i.i.0, %z.i.i.0
  %579 = and i32 %i.i.i.0, 1
  %.not = icmp eq i32 %579, 0
  %580 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %581 = getelementptr inbounds double, double addrspace(1)* %577, i64 1
  %582 = load double, double addrspace(1)* %581, align 8
  %583 = call double @llvm.fma.f64(double noundef %580, double %578, double %582)
  %584 = getelementptr inbounds double, double addrspace(1)* %577, i64 2
  %585 = load double, double addrspace(1)* %584, align 8
  %586 = call double @llvm.fma.f64(double %583, double %578, double %585)
  %587 = getelementptr inbounds double, double addrspace(1)* %577, i64 3
  %588 = load double, double addrspace(1)* %587, align 8
  %589 = call double @llvm.fma.f64(double %586, double %578, double %588)
  %590 = getelementptr inbounds double, double addrspace(1)* %577, i64 4
  %591 = load double, double addrspace(1)* %590, align 8
  %592 = call double @llvm.fma.f64(double %589, double %578, double %591)
  %593 = getelementptr inbounds double, double addrspace(1)* %577, i64 5
  %594 = load double, double addrspace(1)* %593, align 8
  %595 = call double @llvm.fma.f64(double %592, double %578, double %594)
  %596 = getelementptr inbounds double, double addrspace(1)* %577, i64 6
  %597 = load double, double addrspace(1)* %596, align 8
  %598 = call double @llvm.fma.f64(double %595, double %578, double %597)
  %599 = call double @llvm.fma.f64(double %598, double %z.i.i.0, double %z.i.i.0)
  %600 = call double @llvm.fma.f64(double %598, double %578, double noundef 1.000000e+00)
  %spec.select5 = select i1 %.not, double %599, double %600
  %601 = and i32 %i.i.i.0, 2
  %.not7 = icmp eq i32 %601, 0
  %602 = call double @llvm.fma.f64(double %spec.select5, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not7, double %spec.select5, double %602
  br label %__internal_cos_offset.exit

__internal_cos_offset.exit:                       ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i, %535, %__internal_sin_cos_kerneld.exit.i.i
  %.0 = phi double [ %.1, %__internal_sin_cos_kerneld.exit.i.i ], [ %539, %535 ], [ %spec.select4, %__internal_trig_reduction_kerneld.exit.i.i.i ]
  %603 = fmul double %458, %.0
  br label %604

604:                                              ; preds = %__nv_log.exit, %369, %__internal_cos_offset.exit, %__nv_isinfd.exit, %395, %340, %3
  %r.5 = phi double [ %4, %3 ], [ %337, %__nv_log.exit ], [ %366, %340 ], [ %392, %369 ], [ %413, %395 ], [ %603, %__internal_cos_offset.exit ], [ 0.000000e+00, %__nv_isinfd.exit ]
  %605 = fcmp ugt double %a, 0.000000e+00
  %606 = fcmp oeq double %a, 0.000000e+00
  %r.6 = select i1 %606, double 0xFFF0000000000000, double 0xFFF8000000000000
  %r.7 = select i1 %605, double %r.5, double %r.6
  ret double %r.7
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_yn(i32 %n, double %a) local_unnamed_addr #110 {
  %q.i.i.i.i5.i304 = alloca i32, align 4
  %q.i.i.i.i.i.i320 = alloca i32, align 4
  %q.i.i.i16.i334 = alloca i32, align 4
  %q.i.i.i.i351 = alloca i32, align 4
  %q.i.i.i.i.i367 = alloca i32, align 4
  %q.i.i.i381 = alloca i32, align 4
  %q.i.i.i.i9.i133 = alloca i32, align 4
  %q.i.i.i.i.i.i149 = alloca i32, align 4
  %q.i.i.i20.i163 = alloca i32, align 4
  %q.i.i.i.i202 = alloca i32, align 4
  %q.i.i.i.i.i218 = alloca i32, align 4
  %q.i.i.i232 = alloca i32, align 4
  %q.i.i.i.i9.i = alloca i32, align 4
  %q.i.i.i.i.i.i20 = alloca i32, align 4
  %q.i.i.i20.i = alloca i32, align 4
  %q.i.i.i.i58 = alloca i32, align 4
  %q.i.i.i.i.i74 = alloca i32, align 4
  %q.i.i.i88 = alloca i32, align 4
  %q.i.i.i.i5.i = alloca i32, align 4
  %q.i.i.i.i.i.i = alloca i32, align 4
  %q.i.i.i16.i = alloca i32, align 4
  %q.i.i.i.i = alloca i32, align 4
  %q.i.i.i.i.i = alloca i32, align 4
  %q.i.i.i = alloca i32, align 4
  %1 = call double @llvm.fabs.f64(double %a)
  %2 = fcmp olt double %1, 0x730D67819E8D2
  br i1 %2, label %3, label %5

3:                                                ; preds = %0
  %4 = fdiv double 0xBFE45F306DC9C883, %1
  br label %__nv_y1.exit

5:                                                ; preds = %0
  %6 = fcmp ugt double %1, 0x3FF4C6F208132576
  br i1 %6, label %338, label %7

7:                                                ; preds = %5
  %8 = fmul double %a, %a
  %9 = call double @llvm.fma.f64(double %8, double noundef 0x3D5249F90687428C, double noundef 0xBDCF0B5B1FB7B95E)
  %10 = call double @llvm.fma.f64(double %9, double %8, double noundef 0x3E432E589311FA14)
  %11 = call double @llvm.fma.f64(double %10, double %8, double noundef 0xBEB0A780AA4A92E9)
  %12 = call double @llvm.fma.f64(double %11, double %8, double noundef 0x3F12C7DBFFCAEC2B)
  %13 = call double @llvm.fma.f64(double %12, double %8, double noundef 0xBF6835B97894BA4A)
  %14 = call double @llvm.fma.f64(double %13, double %8, double noundef 0x3FABD3975C75B4A3)
  %15 = call double @llvm.fma.f64(double %14, double %8, double noundef 0xBFC91866143CBC8A)
  %16 = fmul double %1, %15
  %17 = fcmp ugt double %1, 0x400353AABAD7B784
  br i1 %17, label %35, label %18

18:                                               ; preds = %7
  %19 = call double @llvm.fma.f64(double %1, double noundef 0x3D020E4ADCDE2AD3, double noundef 0xBD4DD167A0DC3F55)
  %20 = call double @llvm.fma.f64(double %19, double %1, double noundef 0x3D5503F5A491E487)
  %21 = call double @llvm.fma.f64(double %20, double %1, double noundef 0x3DC1F29940C2403A)
  %22 = call double @llvm.fma.f64(double %21, double %1, double noundef 0x3D84CF9302EACDEF)
  %23 = call double @llvm.fma.f64(double %22, double %1, double noundef 0xBE384A53DBBCA436)
  %24 = call double @llvm.fma.f64(double %23, double %1, double noundef 0x3D9779BEE4F63BCC)
  %25 = call double @llvm.fma.f64(double %24, double %1, double noundef 0x3EA6C160E414F3F0)
  %26 = call double @llvm.fma.f64(double %25, double %1, double noundef 0x3D8F3D2F12430699)
  %27 = call double @llvm.fma.f64(double %26, double %1, double noundef 0xBF0C71C72C0CED04)
  %28 = call double @llvm.fma.f64(double %27, double %1, double noundef 0x3D659BCA506F1128)
  %29 = call double @llvm.fma.f64(double %28, double %1, double noundef 0x3F65555555506982)
  %30 = call double @llvm.fma.f64(double %29, double %1, double noundef 0x3D15BA0B425F1BFB)
  %31 = call double @llvm.fma.f64(double %30, double %1, double noundef 0xBFB0000000000065)
  %32 = call double @llvm.fma.f64(double %31, double %1, double noundef 0x3C8729A7253FB679)
  %33 = call double @llvm.fma.f64(double %32, double %1, double noundef 5.000000e-01)
  %34 = fmul double %1, %33
  br label %__nv_j1.exit.i

35:                                               ; preds = %7
  %36 = fcmp ugt double %1, 0x4015B1D0574614EA
  br i1 %36, label %56, label %37

37:                                               ; preds = %35
  %38 = fadd double %1, 0xC00EA75575AF6F09
  %39 = fadd double %38, 0x3CA60155A9D1B256
  %40 = call double @llvm.fma.f64(double %39, double noundef 0xBCF8D3CDBB60175E, double noundef 0x3D41011A1DF02DAD)
  %41 = call double @llvm.fma.f64(double %40, double %39, double noundef 0x3D76013AC1E5E222)
  %42 = call double @llvm.fma.f64(double %41, double %39, double noundef 0xBDBEC315D96D5F03)
  %43 = call double @llvm.fma.f64(double %42, double %39, double noundef 0xBDF03BE1B4B57207)
  %44 = call double @llvm.fma.f64(double %43, double %39, double noundef 0x3E345695F8B660F7)
  %45 = call double @llvm.fma.f64(double %44, double %39, double noundef 0x3E617069FCFCFFF4)
  %46 = call double @llvm.fma.f64(double %45, double %39, double noundef 0xBEA33825C36745EB)
  %47 = call double @llvm.fma.f64(double %46, double %39, double noundef 0xBEC9799D4F90931B)
  %48 = call double @llvm.fma.f64(double %47, double %39, double noundef 0x3F083A06E2F7DF13)
  %49 = call double @llvm.fma.f64(double %48, double %39, double noundef 0x3F26E4C2D53A7CF6)
  %50 = call double @llvm.fma.f64(double %49, double %39, double noundef 0xBF624B3409957B1C)
  %51 = call double @llvm.fma.f64(double %50, double %39, double noundef 0xBF7537544C3325DF)
  %52 = call double @llvm.fma.f64(double %51, double %39, double noundef 0x3FAB589D1DA138E2)
  %53 = call double @llvm.fma.f64(double %52, double %39, double noundef 0x3FAAE8A39F51AD13)
  %54 = call double @llvm.fma.f64(double %53, double %39, double noundef 0xBFD9C6CF582CBF7F)
  %55 = fmul double %39, %54
  br label %__nv_j1.exit.i

56:                                               ; preds = %35
  %57 = fcmp ugt double %1, 0x40213065E54C1AA9
  br i1 %57, label %77, label %58

58:                                               ; preds = %56
  %59 = fadd double %1, 0xC01C0FF5F3B47250
  %60 = fadd double %59, 0x3C9B226D9D243827
  %61 = call double @llvm.fma.f64(double %60, double noundef 0x3CF3EB867515FAD6, double noundef 0xBD40E8363DB649A9)
  %62 = call double @llvm.fma.f64(double %61, double %60, double noundef 0xBD73B7DD4A6608FB)
  %63 = call double @llvm.fma.f64(double %62, double %60, double noundef 0x3DBEC5E01482C750)
  %64 = call double @llvm.fma.f64(double %63, double %60, double noundef 0x3DEC62BB9E882103)
  %65 = call double @llvm.fma.f64(double %64, double %60, double noundef 0xBE34462EED732A23)
  %66 = call double @llvm.fma.f64(double %65, double %60, double noundef 0xBE5D48DCAD7DC59B)
  %67 = call double @llvm.fma.f64(double %66, double %60, double noundef 0x3EA3026DF29167E9)
  %68 = call double @llvm.fma.f64(double %67, double %60, double noundef 0x3EC4255B0119666C)
  %69 = call double @llvm.fma.f64(double %68, double %60, double noundef 0xBF0796A751B32693)
  %70 = call double @llvm.fma.f64(double %69, double %60, double noundef 0xBF207358BBDBA284)
  %71 = call double @llvm.fma.f64(double %70, double %60, double noundef 0x3F613FBC7D6927B1)
  %72 = call double @llvm.fma.f64(double %71, double %60, double noundef 0x3F69A4B292E3DD75)
  %73 = call double @llvm.fma.f64(double %72, double %60, double noundef 0xBFA80C83BDEEE4FB)
  %74 = call double @llvm.fma.f64(double %73, double %60, double noundef 0xBF95E70DC60362BF)
  %75 = call double @llvm.fma.f64(double %74, double %60, double noundef 0x3FD33518B3874E8A)
  %76 = fmul double %60, %75
  br label %__nv_j1.exit.i

77:                                               ; preds = %56
  %78 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %79 = icmp eq i32 %78, 200
  br i1 %79, label %86, label %80

80:                                               ; preds = %77
  %81 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %82 = icmp eq i32 %81, 350
  br i1 %82, label %86, label %83

83:                                               ; preds = %80
  %84 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %85 = icmp eq i32 %84, 370
  br i1 %85, label %86, label %88

86:                                               ; preds = %83, %80, %77
  %87 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i

88:                                               ; preds = %83
  %89 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %90 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %91 = and i32 %90, 2147483647
  %92 = icmp eq i32 %91, 2146435072
  %93 = icmp eq i32 %89, 0
  %94 = select i1 %92, i1 %93, i1 false
  br label %__nv_isinfd.exit.i.i

__nv_isinfd.exit.i.i:                             ; preds = %88, %86
  %.012.in = phi i1 [ %87, %86 ], [ %94, %88 ]
  br i1 %.012.in, label %__nv_j1.exit.i, label %95

95:                                               ; preds = %__nv_isinfd.exit.i.i
  %96 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %97 = fneg double %1
  %98 = call double @llvm.fma.f64(double %97, double %96, double noundef 1.000000e+00)
  %99 = call double @llvm.fma.f64(double %98, double %98, double %98)
  %100 = call double @llvm.fma.f64(double %99, double %96, double %96)
  %101 = fmul double %100, %100
  %102 = call double @llvm.fma.f64(double %101, double noundef 0x40CD02EA3F2F6751, double noundef 0xC099C06322A3F8BE)
  %103 = call double @llvm.fma.f64(double %102, double %101, double noundef 0x405B89354DA77324)
  %104 = call double @llvm.fma.f64(double %103, double %101, double noundef 0xC01E352294653188)
  %105 = call double @llvm.fma.f64(double %104, double %101, double noundef 0x3FE9BC7DB16BD7A7)
  %106 = call double @llvm.fma.f64(double %105, double %101, double noundef 0xBFC8BFE1C3A4F741)
  %107 = call double @llvm.fma.f64(double %106, double %101, double noundef 0x3FC7FFFFF0D00BE2)
  %108 = call double @llvm.fma.f64(double %107, double %101, double noundef 0x3FF00000000068CC)
  %109 = call double @llvm.fma.f64(double %101, double noundef 0xC18DA26B212FDC9A, double noundef 0x415A30AC6857BEE0)
  %110 = call double @llvm.fma.f64(double %109, double %101, double noundef 0xC11764222AD7C910)
  %111 = call double @llvm.fma.f64(double %110, double %101, double noundef 0x40CEB02E0C306857)
  %112 = call double @llvm.fma.f64(double %111, double %101, double noundef 0xC08351859FA2B23B)
  %113 = call double @llvm.fma.f64(double %112, double %101, double noundef 0x403E65A07AF51F42)
  %114 = call double @llvm.fma.f64(double %113, double %101, double noundef 0xC002F2B817F77A57)
  %115 = call double @llvm.fma.f64(double %114, double %101, double noundef 0x3FD7BCC34DA069FD)
  %116 = call double @llvm.fma.f64(double %115, double %101, double noundef 0xBFC4FFFFF8A44463)
  %117 = call double @llvm.fma.f64(double %116, double %101, double noundef 0x3FD7FFFFFFFF5CD7)
  %118 = call double @llvm.fma.f64(double %117, double %100, double %1)
  %119 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %120 = fmul double %119, 0x3FE9884533D43651
  %121 = fmul double %120, %108
  %122 = fmul double %118, 0x3FE45F306DC9C883
  %123 = call i32 @llvm.nvvm.d2i.rn(double %122) #139
  store i32 %123, i32* %q.i.i.i16.i, align 4, !noalias !68
  %124 = sitofp i32 %123 to double
  %125 = fneg double %124
  %126 = call double @llvm.fma.f64(double %125, double noundef 0x3FF921FB54442D18, double %118)
  %127 = call double @llvm.fma.f64(double %125, double noundef 0x3C91A62633145C00, double %126)
  %128 = call double @llvm.fma.f64(double %125, double noundef 0x397B839A252049C0, double %127)
  %129 = call double @llvm.fabs.f64(double %118)
  %130 = fcmp ult double %129, 0x41E0000000000000
  br i1 %130, label %__internal_trig_reduction_kerneld.exit.i.i20.i, label %131

131:                                              ; preds = %95
  %132 = call fastcc double @__internal_trig_reduction_slowpathd(double %118, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i16.i) #139
  %.pre = load i32, i32* %q.i.i.i16.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i20.i

__internal_trig_reduction_kerneld.exit.i.i20.i:   ; preds = %131, %95
  %133 = phi i32 [ %.pre, %131 ], [ %123, %95 ]
  %t.i.i.i15.i.0 = phi double [ %132, %131 ], [ %128, %95 ]
  %134 = and i32 %133, 3
  %135 = uitofp i32 %134 to double
  %136 = fadd double %t.i.i.i15.i.0, 0xC002D97C7F3321D2
  %137 = fmul double %135, 0x3FF921FB54442D18
  %138 = fadd double %136, %137
  %139 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %140 = icmp eq i32 %139, 350
  br i1 %140, label %144, label %141

141:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i20.i
  %142 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %143 = icmp eq i32 %142, 370
  br i1 %143, label %144, label %203

144:                                              ; preds = %141, %__internal_trig_reduction_kerneld.exit.i.i20.i
  %145 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %146 = icmp eq i32 %145, 200
  br i1 %146, label %153, label %147

147:                                              ; preds = %144
  %148 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %149 = icmp eq i32 %148, 350
  br i1 %149, label %153, label %150

150:                                              ; preds = %147
  %151 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %152 = icmp eq i32 %151, 370
  br i1 %152, label %153, label %156

153:                                              ; preds = %150, %147, %144
  %154 = call double @llvm.fabs.f64(double %138)
  %155 = fcmp oeq double %154, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i.i

156:                                              ; preds = %150
  %157 = call i32 @llvm.nvvm.d2i.lo(double %138) #139
  %158 = call i32 @llvm.nvvm.d2i.hi(double %138) #139
  %159 = and i32 %158, 2147483647
  %160 = icmp eq i32 %159, 2146435072
  %161 = icmp eq i32 %157, 0
  %162 = select i1 %160, i1 %161, i1 false
  br label %__nv_isinfd.exit.i.i.i.i.i

__nv_isinfd.exit.i.i.i.i.i:                       ; preds = %156, %153
  %.08.in = phi i1 [ %155, %153 ], [ %162, %156 ]
  br i1 %.08.in, label %163, label %165

163:                                              ; preds = %__nv_isinfd.exit.i.i.i.i.i
  %164 = fmul double %138, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i

165:                                              ; preds = %__nv_isinfd.exit.i.i.i.i.i
  %166 = fmul double %138, 0x3FE45F306DC9C883
  %167 = call i32 @llvm.nvvm.d2i.rn(double %166) #139
  store i32 %167, i32* %q.i.i.i.i.i.i, align 4, !noalias !68
  %168 = sitofp i32 %167 to double
  %169 = fneg double %168
  %170 = call double @llvm.fma.f64(double %169, double noundef 0x3FF921FB54442D18, double %138)
  %171 = call double @llvm.fma.f64(double %169, double noundef 0x3C91A62633145C00, double %170)
  %172 = call double @llvm.fma.f64(double %169, double noundef 0x397B839A252049C0, double %171)
  %173 = call double @llvm.fabs.f64(double %138)
  %174 = fcmp ult double %173, 0x41E0000000000000
  br i1 %174, label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i, label %175

175:                                              ; preds = %165
  %176 = call fastcc double @__internal_trig_reduction_slowpathd(double %138, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i.i) #139
  %.pre74 = load i32, i32* %q.i.i.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i.i.i: ; preds = %165, %175, %163
  %i.i.i.i.i.i.0 = phi i32 [ 0, %163 ], [ %.pre74, %175 ], [ %167, %165 ]
  %.07 = phi double [ %164, %163 ], [ %176, %175 ], [ %172, %165 ]
  %177 = fmul double %.07, %.07
  %178 = call double @llvm.fma.f64(double %177, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %179 = call double @llvm.fma.f64(double %178, double %177, double noundef 0xBE927E4F8E06E6D9)
  %180 = call double @llvm.fma.f64(double %179, double %177, double noundef 0x3EFA01A019DDBCE9)
  %181 = call double @llvm.fma.f64(double %180, double %177, double noundef 0xBF56C16C16C15D47)
  %182 = call double @llvm.fma.f64(double %181, double %177, double noundef 0x3FA5555555555551)
  %183 = call double @llvm.fma.f64(double %182, double %177, double noundef -5.000000e-01)
  %184 = call double @llvm.fma.f64(double %183, double %177, double noundef 1.000000e+00)
  %185 = call double @llvm.fma.f64(double %177, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %186 = call double @llvm.fma.f64(double %185, double %177, double noundef 0x3EC71DE369ACE392)
  %187 = call double @llvm.fma.f64(double %186, double %177, double noundef 0xBF2A01A019DB62A1)
  %188 = call double @llvm.fma.f64(double %187, double %177, double noundef 0x3F81111111110818)
  %189 = call double @llvm.fma.f64(double %188, double %177, double noundef 0xBFC5555555555554)
  %190 = call double @llvm.fma.f64(double %189, double %177, double noundef 0.000000e+00)
  %191 = call double @llvm.fma.f64(double %190, double %.07, double %.07)
  %192 = call i32 @llvm.nvvm.d2i.hi(double %191) #139
  %193 = call i32 @llvm.nvvm.d2i.lo(double %191) #139
  %194 = xor i32 %192, -2147483648
  %195 = call double @llvm.nvvm.lohi.i2d(i32 %193, i32 %194) #139
  %196 = and i32 %i.i.i.i.i.i.0, 1
  %.not67 = icmp eq i32 %196, 0
  %spec.select1 = select i1 %.not67, double %184, double %195
  %197 = and i32 %i.i.i.i.i.i.0, 2
  %.not68 = icmp eq i32 %197, 0
  br i1 %.not68, label %__internal_cos_offset.exit.i.i, label %198

198:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i
  %199 = call i32 @llvm.nvvm.d2i.hi(double %spec.select1) #139
  %200 = call i32 @llvm.nvvm.d2i.lo(double %spec.select1) #139
  %201 = xor i32 %199, -2147483648
  %202 = call double @llvm.nvvm.lohi.i2d(i32 %200, i32 %201) #139
  br label %__internal_cos_offset.exit.i.i

203:                                              ; preds = %141
  %204 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %205 = icmp eq i32 %204, 200
  br i1 %205, label %212, label %206

206:                                              ; preds = %203
  %207 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %208 = icmp eq i32 %207, 350
  br i1 %208, label %212, label %209

209:                                              ; preds = %206
  %210 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %211 = icmp eq i32 %210, 370
  br i1 %211, label %212, label %215

212:                                              ; preds = %209, %206, %203
  %213 = call double @llvm.fabs.f64(double %138)
  %214 = fcmp oeq double %213, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i21.i

215:                                              ; preds = %209
  %216 = call i32 @llvm.nvvm.d2i.lo(double %138) #139
  %217 = call i32 @llvm.nvvm.d2i.hi(double %138) #139
  %218 = and i32 %217, 2147483647
  %219 = icmp eq i32 %218, 2146435072
  %220 = icmp eq i32 %216, 0
  %221 = select i1 %219, i1 %220, i1 false
  br label %__nv_isinfd.exit.i.i.i21.i

__nv_isinfd.exit.i.i.i21.i:                       ; preds = %215, %212
  %.09.in = phi i1 [ %214, %212 ], [ %221, %215 ]
  br i1 %.09.in, label %222, label %224

222:                                              ; preds = %__nv_isinfd.exit.i.i.i21.i
  %223 = fmul double %138, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i

224:                                              ; preds = %__nv_isinfd.exit.i.i.i21.i
  %225 = fmul double %138, 0x3FE45F306DC9C883
  %226 = call i32 @llvm.nvvm.d2i.rn(double %225) #139
  store i32 %226, i32* %q.i.i.i.i5.i, align 4, !noalias !68
  %227 = sitofp i32 %226 to double
  %228 = fneg double %227
  %229 = call double @llvm.fma.f64(double %228, double noundef 0x3FF921FB54442D18, double %138)
  %230 = call double @llvm.fma.f64(double %228, double noundef 0x3C91A62633145C00, double %229)
  %231 = call double @llvm.fma.f64(double %228, double noundef 0x397B839A252049C0, double %230)
  %232 = call double @llvm.fabs.f64(double %138)
  %233 = fcmp ult double %232, 0x41E0000000000000
  br i1 %233, label %__internal_sin_cos_kerneld.exit.i.i.i.i, label %234

234:                                              ; preds = %224
  %235 = call fastcc double @__internal_trig_reduction_slowpathd(double %138, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i5.i) #139
  %.pre73 = load i32, i32* %q.i.i.i.i5.i, align 4
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i

__internal_sin_cos_kerneld.exit.i.i.i.i:          ; preds = %224, %234, %222
  %i.i.i.i13.i.0 = phi i32 [ 0, %222 ], [ %.pre73, %234 ], [ %226, %224 ]
  %z.i.i.i12.i.0 = phi double [ %223, %222 ], [ %235, %234 ], [ %231, %224 ]
  %236 = add nsw i32 %i.i.i.i13.i.0, 1
  %237 = shl i32 %236, 3
  %238 = and i32 %237, 8
  %239 = zext i32 %238 to i64
  %240 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %239
  %241 = fmul double %z.i.i.i12.i.0, %z.i.i.i12.i.0
  %242 = and i32 %236, 1
  %.not64 = icmp eq i32 %242, 0
  %243 = select i1 %.not64, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %244 = getelementptr inbounds double, double addrspace(1)* %240, i64 1
  %245 = load double, double addrspace(1)* %244, align 8
  %246 = call double @llvm.fma.f64(double noundef %243, double %241, double %245)
  %247 = getelementptr inbounds double, double addrspace(1)* %240, i64 2
  %248 = load double, double addrspace(1)* %247, align 8
  %249 = call double @llvm.fma.f64(double %246, double %241, double %248)
  %250 = getelementptr inbounds double, double addrspace(1)* %240, i64 3
  %251 = load double, double addrspace(1)* %250, align 8
  %252 = call double @llvm.fma.f64(double %249, double %241, double %251)
  %253 = getelementptr inbounds double, double addrspace(1)* %240, i64 4
  %254 = load double, double addrspace(1)* %253, align 8
  %255 = call double @llvm.fma.f64(double %252, double %241, double %254)
  %256 = getelementptr inbounds double, double addrspace(1)* %240, i64 5
  %257 = load double, double addrspace(1)* %256, align 8
  %258 = call double @llvm.fma.f64(double %255, double %241, double %257)
  %259 = getelementptr inbounds double, double addrspace(1)* %240, i64 6
  %260 = load double, double addrspace(1)* %259, align 8
  %261 = call double @llvm.fma.f64(double %258, double %241, double %260)
  %262 = call double @llvm.fma.f64(double %261, double %z.i.i.i12.i.0, double %z.i.i.i12.i.0)
  %263 = call double @llvm.fma.f64(double %261, double %241, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not64, double %262, double %263
  %264 = and i32 %236, 2
  %.not66 = icmp eq i32 %264, 0
  %265 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.111 = select i1 %.not66, double %spec.select2, double %265
  br label %__internal_cos_offset.exit.i.i

__internal_cos_offset.exit.i.i:                   ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i, %198, %__internal_sin_cos_kerneld.exit.i.i.i.i
  %.06 = phi double [ %.111, %__internal_sin_cos_kerneld.exit.i.i.i.i ], [ %202, %198 ], [ %spec.select1, %__internal_trig_reduction_kerneld.exit.i.i.i.i.i ]
  %266 = fmul double %121, %.06
  br label %__nv_j1.exit.i

__nv_j1.exit.i:                                   ; preds = %37, %__internal_cos_offset.exit.i.i, %__nv_isinfd.exit.i.i, %58, %18
  %r.i.i.3 = phi double [ %34, %18 ], [ %55, %37 ], [ %76, %58 ], [ %266, %__internal_cos_offset.exit.i.i ], [ 0.000000e+00, %__nv_isinfd.exit.i.i ]
  %267 = fcmp olt double %1, 1.000000e-30
  %268 = fmul double %1, 5.000000e-01
  %r.i.i.5 = select i1 %267, double %268, double %r.i.i.3
  %269 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %270 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %271 = icmp slt i32 %269, 1048576
  br i1 %271, label %272, label %276

272:                                              ; preds = %__nv_j1.exit.i
  %273 = fmul double %1, 0x4350000000000000
  %274 = call i32 @llvm.nvvm.d2i.hi(double %273) #139
  %275 = call i32 @llvm.nvvm.d2i.lo(double %273) #139
  br label %276

276:                                              ; preds = %272, %__nv_j1.exit.i
  %.013 = phi double [ %273, %272 ], [ %1, %__nv_j1.exit.i ]
  %ihi.i.i.0 = phi i32 [ %274, %272 ], [ %269, %__nv_j1.exit.i ]
  %ilo.i.i.0 = phi i32 [ %275, %272 ], [ %270, %__nv_j1.exit.i ]
  %e.i.i.0 = phi i32 [ -1077, %272 ], [ -1023, %__nv_j1.exit.i ]
  %277 = add i32 %ihi.i.i.0, -1
  %278 = icmp ult i32 %277, 2146435071
  br i1 %278, label %279, label %328

279:                                              ; preds = %276
  %280 = lshr i32 %ihi.i.i.0, 20
  %281 = add nsw i32 %e.i.i.0, %280
  %282 = and i32 %ihi.i.i.0, -2146435073
  %283 = or i32 %282, 1072693248
  %284 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i.0, i32 %283) #139
  %285 = icmp sgt i32 %283, 1073127582
  br i1 %285, label %286, label %292

286:                                              ; preds = %279
  %287 = call i32 @llvm.nvvm.d2i.lo(double %284) #139
  %288 = call i32 @llvm.nvvm.d2i.hi(double %284) #139
  %289 = add i32 %288, -1048576
  %290 = call double @llvm.nvvm.lohi.i2d(i32 %287, i32 %289) #139
  %291 = add nsw i32 %281, 1
  br label %292

292:                                              ; preds = %286, %279
  %m.i.i.0 = phi double [ %290, %286 ], [ %284, %279 ]
  %e.i.i.1 = phi i32 [ %291, %286 ], [ %281, %279 ]
  %293 = fadd double %m.i.i.0, -1.000000e+00
  %294 = fadd double %m.i.i.0, 1.000000e+00
  %295 = call double @llvm.nvvm.rcp.approx.ftz.d(double %294) #139
  %296 = fneg double %294
  %297 = call double @llvm.fma.f64(double %296, double %295, double noundef 1.000000e+00)
  %298 = call double @llvm.fma.f64(double %297, double %297, double %297)
  %299 = call double @llvm.fma.f64(double %298, double %295, double %295)
  %300 = fmul double %293, %299
  %301 = fadd double %300, %300
  %302 = fmul double %301, %301
  %303 = call double @llvm.fma.f64(double %302, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %304 = call double @llvm.fma.f64(double %303, double %302, double noundef 0x3EF3B2669F02676F)
  %305 = call double @llvm.fma.f64(double %304, double %302, double noundef 0x3F1745CBA9AB0956)
  %306 = call double @llvm.fma.f64(double %305, double %302, double noundef 0x3F3C71C72D1B5154)
  %307 = call double @llvm.fma.f64(double %306, double %302, double noundef 0x3F624924923BE72D)
  %308 = call double @llvm.fma.f64(double %307, double %302, double noundef 0x3F8999999999A3C4)
  %309 = call double @llvm.fma.f64(double %308, double %302, double noundef 0x3FB5555555555554)
  %310 = fsub double %293, %301
  %311 = fmul double %310, 2.000000e+00
  %312 = fneg double %301
  %313 = call double @llvm.fma.f64(double %312, double %293, double %311)
  %314 = fmul double %299, %313
  %315 = fmul double %302, %309
  %316 = call double @llvm.fma.f64(double %315, double %301, double %314)
  %317 = xor i32 %e.i.i.1, -2147483648
  %318 = call double @llvm.nvvm.lohi.i2d(i32 %317, i32 noundef 1127219200) #139
  %319 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %320 = fsub double %318, %319
  %321 = call double @llvm.fma.f64(double %320, double noundef 0x3FE62E42FEFA39EF, double %301)
  %322 = fneg double %320
  %323 = call double @llvm.fma.f64(double %322, double noundef 0x3FE62E42FEFA39EF, double %321)
  %324 = fsub double %323, %301
  %325 = fsub double %316, %324
  %326 = call double @llvm.fma.f64(double %320, double noundef 0x3C7ABC9E3B39803F, double %325)
  %327 = fadd double %321, %326
  br label %__nv_log.exit.i

328:                                              ; preds = %276
  %329 = call double @llvm.fma.f64(double %.013, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %330 = call i32 @llvm.nvvm.d2i.hi(double %.013) #139
  %331 = bitcast i32 %330 to float
  %332 = fcmp oeq float %331, 0.000000e+00
  %q.i28.i.0 = select i1 %332, double 0xFFF0000000000000, double %329
  br label %__nv_log.exit.i

__nv_log.exit.i:                                  ; preds = %328, %292
  %q.i28.i.1 = phi double [ %327, %292 ], [ %q.i28.i.0, %328 ]
  %333 = fmul double %r.i.i.5, %q.i28.i.1
  %334 = fdiv double 1.000000e+00, %1
  %335 = fsub double %333, %334
  %336 = fmul double %335, 0x3FE45F306DC9C883
  %337 = fadd double %16, %336
  br label %__nv_y1.exit

338:                                              ; preds = %5
  %339 = fcmp ugt double %1, 0x4009B510EC2ADC83
  br i1 %339, label %367, label %340

340:                                              ; preds = %338
  %341 = fadd double %1, 0xC00193BED4DFF243
  %342 = fadd double %341, 0x3C8BD1E50D219BFD
  %343 = call double @llvm.fma.f64(double %342, double noundef 0xBE2B87B0BE2AA150, double noundef 0x3E4833AAE4D8B975)
  %344 = call double @llvm.fma.f64(double %343, double %342, double noundef 0xBE44E279B423E68F)
  %345 = call double @llvm.fma.f64(double %344, double %342, double noundef 0x3E129DC6A747EB4F)
  %346 = call double @llvm.fma.f64(double %345, double %342, double noundef 0xBE61D15534496CD8)
  %347 = call double @llvm.fma.f64(double %346, double %342, double noundef 0x3E7EEC8D48FECE00)
  %348 = call double @llvm.fma.f64(double %347, double %342, double noundef 0xBE8D1180AF70A134)
  %349 = call double @llvm.fma.f64(double %348, double %342, double noundef 0x3E9C8386A0EA1388)
  %350 = call double @llvm.fma.f64(double %349, double %342, double noundef 0xBEB01A014E7F3250)
  %351 = call double @llvm.fma.f64(double %350, double %342, double noundef 0x3EC1FB752010A320)
  %352 = call double @llvm.fma.f64(double %351, double %342, double noundef 0xBED3AA0AFF4E332B)
  %353 = call double @llvm.fma.f64(double %352, double %342, double noundef 0x3EE584A6C77F6700)
  %354 = call double @llvm.fma.f64(double %353, double %342, double noundef 0xBEF794C520FC2EBB)
  %355 = call double @llvm.fma.f64(double %354, double %342, double noundef 0x3F09D18D2D35CC71)
  %356 = call double @llvm.fma.f64(double %355, double %342, double noundef 0xBF1C3FB7315C4599)
  %357 = call double @llvm.fma.f64(double %356, double %342, double noundef 0x3F2EEA7ADECCE927)
  %358 = call double @llvm.fma.f64(double %357, double %342, double noundef 0xBF40B2D85257446F)
  %359 = call double @llvm.fma.f64(double %358, double %342, double noundef 0x3F517AB4B1FE5D5B)
  %360 = call double @llvm.fma.f64(double %359, double %342, double noundef 0xBF65429DC6516C0D)
  %361 = call double @llvm.fma.f64(double %360, double %342, double noundef 0x3F7E671C7D0B090B)
  %362 = call double @llvm.fma.f64(double %361, double %342, double noundef 0xBF73A6DEC36FB27C)
  %363 = call double @llvm.fma.f64(double %362, double %342, double noundef 0xBFA0D2AF4E931FD1)
  %364 = call double @llvm.fma.f64(double %363, double %342, double noundef 0xBFBE56F82217B964)
  %365 = call double @llvm.fma.f64(double %364, double %342, double noundef 0x3FE0AA48442F014B)
  %366 = fmul double %342, %365
  br label %__nv_y1.exit

367:                                              ; preds = %338
  %368 = fcmp ugt double %1, 0x401C0D26D5A541CB
  br i1 %368, label %393, label %369

369:                                              ; preds = %367
  %370 = fadd double %1, 0xC015B7FE4E87B02E
  %371 = fadd double %370, 0xBCBDFE7BAC228E8C
  %372 = call double @llvm.fma.f64(double %371, double noundef 0x3CBA3C76069F1D8C, double noundef 0x3CC69A30996793E2)
  %373 = call double @llvm.fma.f64(double %372, double %371, double noundef 0xBCDDD8432FE756E7)
  %374 = call double @llvm.fma.f64(double %373, double %371, double noundef 0xBD143158EEE220F7)
  %375 = call double @llvm.fma.f64(double %374, double %371, double noundef 0x3D28D44491230F5A)
  %376 = call double @llvm.fma.f64(double %375, double %371, double noundef 0xBD438842EAF4EDBC)
  %377 = call double @llvm.fma.f64(double %376, double %371, double noundef 0x3D74958DAFBFAF5A)
  %378 = call double @llvm.fma.f64(double %377, double %371, double noundef 0xBD9449A60E664848)
  %379 = call double @llvm.fma.f64(double %378, double %371, double noundef 0x3D838BC8CD594A76)
  %380 = call double @llvm.fma.f64(double %379, double %371, double noundef 0xBDFA798002141323)
  %381 = call double @llvm.fma.f64(double %380, double %371, double noundef 0x3E380B4198956AAA)
  %382 = call double @llvm.fma.f64(double %381, double %371, double noundef 0x3E5B62B5F21BACD4)
  %383 = call double @llvm.fma.f64(double %382, double %371, double noundef 0xBEA255E729FB6AAE)
  %384 = call double @llvm.fma.f64(double %383, double %371, double noundef 0xBEC80618F6BAE5AA)
  %385 = call double @llvm.fma.f64(double %384, double %371, double noundef 0x3F085B940F8E8D36)
  %386 = call double @llvm.fma.f64(double %385, double %371, double noundef 0x3F2337C7E10E14E8)
  %387 = call double @llvm.fma.f64(double %386, double %371, double noundef 0xBF61BE6DB99332CA)
  %388 = call double @llvm.fma.f64(double %387, double %371, double noundef 0xBF710A329E2BE9B8)
  %389 = call double @llvm.fma.f64(double %388, double %371, double noundef 0x3FAA15D92DFE3FCF)
  %390 = call double @llvm.fma.f64(double %389, double %371, double noundef 0x3FA00B9F8571C9BE)
  %391 = call double @llvm.fma.f64(double %390, double %371, double noundef 0xBFD5C7C556F0C19A)
  %392 = fmul double %371, %391
  br label %__nv_y1.exit

393:                                              ; preds = %367
  %394 = fcmp ugt double %1, 0x4022585C739ACDDD
  br i1 %394, label %414, label %395

395:                                              ; preds = %393
  %396 = fadd double %1, 0xC0213127AE6169B4
  %397 = fadd double %396, 0xBCB479CC068D9046
  %398 = call double @llvm.fma.f64(double %397, double noundef 0x3CB09CCC22945996, double noundef 0xBD43515F67644276)
  %399 = call double @llvm.fma.f64(double %398, double %397, double noundef 0xBD72C5B978E9F5C7)
  %400 = call double @llvm.fma.f64(double %399, double %397, double noundef 0x3DBEC1151613913C)
  %401 = call double @llvm.fma.f64(double %400, double %397, double noundef 0x3DE9E38D13C4A824)
  %402 = call double @llvm.fma.f64(double %401, double %397, double noundef 0xBE341E75E1088EB5)
  %403 = call double @llvm.fma.f64(double %402, double %397, double noundef 0xBE5A384EBB13CFE1)
  %404 = call double @llvm.fma.f64(double %403, double %397, double noundef 0x3EA2BECB27F8C8F8)
  %405 = call double @llvm.fma.f64(double %404, double %397, double noundef 0x3EC176E72B989FD8)
  %406 = call double @llvm.fma.f64(double %405, double %397, double noundef 0xBF06F7BAB102F822)
  %407 = call double @llvm.fma.f64(double %406, double %397, double noundef 0xBF1B50D7E1D278E1)
  %408 = call double @llvm.fma.f64(double %407, double %397, double noundef 0x3F607A678D60004F)
  %409 = call double @llvm.fma.f64(double %408, double %397, double noundef 0x3F63CED2A2E69115)
  %410 = call double @llvm.fma.f64(double %409, double %397, double noundef 0xBFA6395DFE49FCD4)
  %411 = call double @llvm.fma.f64(double %410, double %397, double noundef 0xBF902B3933CF21B1)
  %412 = call double @llvm.fma.f64(double %411, double %397, double noundef 0x3FD15F993FCEAB5C)
  %413 = fmul double %397, %412
  br label %__nv_y1.exit

414:                                              ; preds = %393
  %415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %416 = icmp eq i32 %415, 200
  br i1 %416, label %423, label %417

417:                                              ; preds = %414
  %418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %419 = icmp eq i32 %418, 350
  br i1 %419, label %423, label %420

420:                                              ; preds = %417
  %421 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %422 = icmp eq i32 %421, 370
  br i1 %422, label %423, label %425

423:                                              ; preds = %420, %417, %414
  %424 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i

425:                                              ; preds = %420
  %426 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %427 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %428 = and i32 %427, 2147483647
  %429 = icmp eq i32 %428, 2146435072
  %430 = icmp eq i32 %426, 0
  %431 = select i1 %429, i1 %430, i1 false
  br label %__nv_isinfd.exit.i

__nv_isinfd.exit.i:                               ; preds = %425, %423
  %.014.in = phi i1 [ %424, %423 ], [ %431, %425 ]
  br i1 %.014.in, label %__nv_y1.exit, label %432

432:                                              ; preds = %__nv_isinfd.exit.i
  %433 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %434 = fneg double %1
  %435 = call double @llvm.fma.f64(double %434, double %433, double noundef 1.000000e+00)
  %436 = call double @llvm.fma.f64(double %435, double %435, double %435)
  %437 = call double @llvm.fma.f64(double %436, double %433, double %433)
  %438 = fmul double %437, %437
  %439 = call double @llvm.fma.f64(double %438, double noundef 0x40D13DB326ECEBFE, double noundef 0xC09C26E89385D5B1)
  %440 = call double @llvm.fma.f64(double %439, double %438, double noundef 0x405C6AB923C6F55E)
  %441 = call double @llvm.fma.f64(double %440, double %438, double noundef 0xC01E61EAF3BD2FA1)
  %442 = call double @llvm.fma.f64(double %441, double %438, double noundef 0x3FE9BF15D9B97DD1)
  %443 = call double @llvm.fma.f64(double %442, double %438, double noundef 0xBFC8BFECF93D7D19)
  %444 = call double @llvm.fma.f64(double %443, double %438, double noundef 0x3FC7FFFFF756AA6C)
  %445 = call double @llvm.fma.f64(double %444, double %438, double noundef 0x3FF0000000003646)
  %446 = call double @llvm.fma.f64(double %438, double noundef 0xC1943281A050209C, double noundef 0x416024E99BA46E7B)
  %447 = call double @llvm.fma.f64(double %446, double %438, double noundef 0xC11A6875D7DFBD65)
  %448 = call double @llvm.fma.f64(double %447, double %438, double noundef 0x40D032C041790233)
  %449 = call double @llvm.fma.f64(double %448, double %438, double noundef 0xC0839F895BC22946)
  %450 = call double @llvm.fma.f64(double %449, double %438, double noundef 0x403E77CC78ECD2D8)
  %451 = call double @llvm.fma.f64(double %450, double %438, double noundef 0xC002F368D0117BE9)
  %452 = call double @llvm.fma.f64(double %451, double %438, double noundef 0x3FD7BCC786009A25)
  %453 = call double @llvm.fma.f64(double %452, double %438, double noundef 0xBFC4FFFFFC51BC7A)
  %454 = call double @llvm.fma.f64(double %453, double %438, double noundef 0x3FD7FFFFFFFFB5EA)
  %455 = call double @llvm.fma.f64(double %454, double %437, double %1)
  %456 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %457 = fmul double %456, 0x3FE9884533D43651
  %458 = fmul double %457, %445
  %459 = fmul double %455, 0x3FE45F306DC9C883
  %460 = call i32 @llvm.nvvm.d2i.rn(double %459) #139
  store i32 %460, i32* %q.i.i.i, align 4, !noalias !68
  %461 = sitofp i32 %460 to double
  %462 = fneg double %461
  %463 = call double @llvm.fma.f64(double %462, double noundef 0x3FF921FB54442D18, double %455)
  %464 = call double @llvm.fma.f64(double %462, double noundef 0x3C91A62633145C00, double %463)
  %465 = call double @llvm.fma.f64(double %462, double noundef 0x397B839A252049C0, double %464)
  %466 = call double @llvm.fabs.f64(double %455)
  %467 = fcmp ult double %466, 0x41E0000000000000
  br i1 %467, label %__internal_trig_reduction_kerneld.exit.i.i, label %468

468:                                              ; preds = %432
  %469 = call fastcc double @__internal_trig_reduction_slowpathd(double %455, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i) #139
  %.pre75 = load i32, i32* %q.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i

__internal_trig_reduction_kerneld.exit.i.i:       ; preds = %468, %432
  %470 = phi i32 [ %.pre75, %468 ], [ %460, %432 ]
  %t.i.i.i.0 = phi double [ %469, %468 ], [ %465, %432 ]
  %471 = and i32 %470, 3
  %472 = uitofp i32 %471 to double
  %473 = fadd double %t.i.i.i.0, 0xC00F6A7A2955385E
  %474 = fmul double %472, 0x3FF921FB54442D18
  %475 = fadd double %473, %474
  %476 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %477 = icmp eq i32 %476, 350
  br i1 %477, label %481, label %478

478:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i
  %479 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %480 = icmp eq i32 %479, 370
  br i1 %480, label %481, label %540

481:                                              ; preds = %478, %__internal_trig_reduction_kerneld.exit.i.i
  %482 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %483 = icmp eq i32 %482, 200
  br i1 %483, label %490, label %484

484:                                              ; preds = %481
  %485 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %486 = icmp eq i32 %485, 350
  br i1 %486, label %490, label %487

487:                                              ; preds = %484
  %488 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %489 = icmp eq i32 %488, 370
  br i1 %489, label %490, label %493

490:                                              ; preds = %487, %484, %481
  %491 = call double @llvm.fabs.f64(double %475)
  %492 = fcmp oeq double %491, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i

493:                                              ; preds = %487
  %494 = call i32 @llvm.nvvm.d2i.lo(double %475) #139
  %495 = call i32 @llvm.nvvm.d2i.hi(double %475) #139
  %496 = and i32 %495, 2147483647
  %497 = icmp eq i32 %496, 2146435072
  %498 = icmp eq i32 %494, 0
  %499 = select i1 %497, i1 %498, i1 false
  br label %__nv_isinfd.exit.i.i.i.i

__nv_isinfd.exit.i.i.i.i:                         ; preds = %493, %490
  %.03.in = phi i1 [ %492, %490 ], [ %499, %493 ]
  br i1 %.03.in, label %500, label %502

500:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %501 = fmul double %475, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

502:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %503 = fmul double %475, 0x3FE45F306DC9C883
  %504 = call i32 @llvm.nvvm.d2i.rn(double %503) #139
  store i32 %504, i32* %q.i.i.i.i.i, align 4, !noalias !68
  %505 = sitofp i32 %504 to double
  %506 = fneg double %505
  %507 = call double @llvm.fma.f64(double %506, double noundef 0x3FF921FB54442D18, double %475)
  %508 = call double @llvm.fma.f64(double %506, double noundef 0x3C91A62633145C00, double %507)
  %509 = call double @llvm.fma.f64(double %506, double noundef 0x397B839A252049C0, double %508)
  %510 = call double @llvm.fabs.f64(double %475)
  %511 = fcmp ult double %510, 0x41E0000000000000
  br i1 %511, label %__internal_trig_reduction_kerneld.exit.i.i.i.i, label %512

512:                                              ; preds = %502
  %513 = call fastcc double @__internal_trig_reduction_slowpathd(double %475, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i) #139
  %.pre77 = load i32, i32* %q.i.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i.i:   ; preds = %502, %512, %500
  %i.i.i.i.i.0 = phi i32 [ 0, %500 ], [ %.pre77, %512 ], [ %504, %502 ]
  %.02 = phi double [ %501, %500 ], [ %513, %512 ], [ %509, %502 ]
  %514 = fmul double %.02, %.02
  %515 = call double @llvm.fma.f64(double %514, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %516 = call double @llvm.fma.f64(double %515, double %514, double noundef 0xBE927E4F8E06E6D9)
  %517 = call double @llvm.fma.f64(double %516, double %514, double noundef 0x3EFA01A019DDBCE9)
  %518 = call double @llvm.fma.f64(double %517, double %514, double noundef 0xBF56C16C16C15D47)
  %519 = call double @llvm.fma.f64(double %518, double %514, double noundef 0x3FA5555555555551)
  %520 = call double @llvm.fma.f64(double %519, double %514, double noundef -5.000000e-01)
  %521 = call double @llvm.fma.f64(double %520, double %514, double noundef 1.000000e+00)
  %522 = call double @llvm.fma.f64(double %514, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %523 = call double @llvm.fma.f64(double %522, double %514, double noundef 0x3EC71DE369ACE392)
  %524 = call double @llvm.fma.f64(double %523, double %514, double noundef 0xBF2A01A019DB62A1)
  %525 = call double @llvm.fma.f64(double %524, double %514, double noundef 0x3F81111111110818)
  %526 = call double @llvm.fma.f64(double %525, double %514, double noundef 0xBFC5555555555554)
  %527 = call double @llvm.fma.f64(double %526, double %514, double noundef 0.000000e+00)
  %528 = call double @llvm.fma.f64(double %527, double %.02, double %.02)
  %529 = call i32 @llvm.nvvm.d2i.hi(double %528) #139
  %530 = call i32 @llvm.nvvm.d2i.lo(double %528) #139
  %531 = xor i32 %529, -2147483648
  %532 = call double @llvm.nvvm.lohi.i2d(i32 %530, i32 %531) #139
  %533 = and i32 %i.i.i.i.i.0, 1
  %.not62 = icmp eq i32 %533, 0
  %spec.select4 = select i1 %.not62, double %521, double %532
  %534 = and i32 %i.i.i.i.i.0, 2
  %.not63 = icmp eq i32 %534, 0
  br i1 %.not63, label %__internal_cos_offset.exit.i, label %535

535:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i
  %536 = call i32 @llvm.nvvm.d2i.hi(double %spec.select4) #139
  %537 = call i32 @llvm.nvvm.d2i.lo(double %spec.select4) #139
  %538 = xor i32 %536, -2147483648
  %539 = call double @llvm.nvvm.lohi.i2d(i32 %537, i32 %538) #139
  br label %__internal_cos_offset.exit.i

540:                                              ; preds = %478
  %541 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %542 = icmp eq i32 %541, 200
  br i1 %542, label %549, label %543

543:                                              ; preds = %540
  %544 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %545 = icmp eq i32 %544, 350
  br i1 %545, label %549, label %546

546:                                              ; preds = %543
  %547 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %548 = icmp eq i32 %547, 370
  br i1 %548, label %549, label %552

549:                                              ; preds = %546, %543, %540
  %550 = call double @llvm.fabs.f64(double %475)
  %551 = fcmp oeq double %550, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i

552:                                              ; preds = %546
  %553 = call i32 @llvm.nvvm.d2i.lo(double %475) #139
  %554 = call i32 @llvm.nvvm.d2i.hi(double %475) #139
  %555 = and i32 %554, 2147483647
  %556 = icmp eq i32 %555, 2146435072
  %557 = icmp eq i32 %553, 0
  %558 = select i1 %556, i1 %557, i1 false
  br label %__nv_isinfd.exit.i.i.i

__nv_isinfd.exit.i.i.i:                           ; preds = %552, %549
  %.04.in = phi i1 [ %551, %549 ], [ %558, %552 ]
  br i1 %.04.in, label %559, label %561

559:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %560 = fmul double %475, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i

561:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %562 = fmul double %475, 0x3FE45F306DC9C883
  %563 = call i32 @llvm.nvvm.d2i.rn(double %562) #139
  store i32 %563, i32* %q.i.i.i.i, align 4, !noalias !68
  %564 = sitofp i32 %563 to double
  %565 = fneg double %564
  %566 = call double @llvm.fma.f64(double %565, double noundef 0x3FF921FB54442D18, double %475)
  %567 = call double @llvm.fma.f64(double %565, double noundef 0x3C91A62633145C00, double %566)
  %568 = call double @llvm.fma.f64(double %565, double noundef 0x397B839A252049C0, double %567)
  %569 = call double @llvm.fabs.f64(double %475)
  %570 = fcmp ult double %569, 0x41E0000000000000
  br i1 %570, label %__internal_sin_cos_kerneld.exit.i.i.i, label %571

571:                                              ; preds = %561
  %572 = call fastcc double @__internal_trig_reduction_slowpathd(double %475, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i) #139
  %.pre76 = load i32, i32* %q.i.i.i.i, align 4
  br label %__internal_sin_cos_kerneld.exit.i.i.i

__internal_sin_cos_kerneld.exit.i.i.i:            ; preds = %561, %571, %559
  %i.i.i.i.0 = phi i32 [ 0, %559 ], [ %.pre76, %571 ], [ %563, %561 ]
  %z.i.i.i.0 = phi double [ %560, %559 ], [ %572, %571 ], [ %568, %561 ]
  %573 = add nsw i32 %i.i.i.i.0, 1
  %574 = shl i32 %573, 3
  %575 = and i32 %574, 8
  %576 = zext i32 %575 to i64
  %577 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %576
  %578 = fmul double %z.i.i.i.0, %z.i.i.i.0
  %579 = and i32 %573, 1
  %.not = icmp eq i32 %579, 0
  %580 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %581 = getelementptr inbounds double, double addrspace(1)* %577, i64 1
  %582 = load double, double addrspace(1)* %581, align 8
  %583 = call double @llvm.fma.f64(double noundef %580, double %578, double %582)
  %584 = getelementptr inbounds double, double addrspace(1)* %577, i64 2
  %585 = load double, double addrspace(1)* %584, align 8
  %586 = call double @llvm.fma.f64(double %583, double %578, double %585)
  %587 = getelementptr inbounds double, double addrspace(1)* %577, i64 3
  %588 = load double, double addrspace(1)* %587, align 8
  %589 = call double @llvm.fma.f64(double %586, double %578, double %588)
  %590 = getelementptr inbounds double, double addrspace(1)* %577, i64 4
  %591 = load double, double addrspace(1)* %590, align 8
  %592 = call double @llvm.fma.f64(double %589, double %578, double %591)
  %593 = getelementptr inbounds double, double addrspace(1)* %577, i64 5
  %594 = load double, double addrspace(1)* %593, align 8
  %595 = call double @llvm.fma.f64(double %592, double %578, double %594)
  %596 = getelementptr inbounds double, double addrspace(1)* %577, i64 6
  %597 = load double, double addrspace(1)* %596, align 8
  %598 = call double @llvm.fma.f64(double %595, double %578, double %597)
  %599 = call double @llvm.fma.f64(double %598, double %z.i.i.i.0, double %z.i.i.i.0)
  %600 = call double @llvm.fma.f64(double %598, double %578, double noundef 1.000000e+00)
  %spec.select5 = select i1 %.not, double %599, double %600
  %601 = and i32 %573, 2
  %.not28 = icmp eq i32 %601, 0
  %602 = call double @llvm.fma.f64(double %spec.select5, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not28, double %spec.select5, double %602
  br label %__internal_cos_offset.exit.i

__internal_cos_offset.exit.i:                     ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i, %535, %__internal_sin_cos_kerneld.exit.i.i.i
  %.01 = phi double [ %.1, %__internal_sin_cos_kerneld.exit.i.i.i ], [ %539, %535 ], [ %spec.select4, %__internal_trig_reduction_kerneld.exit.i.i.i.i ]
  %603 = fmul double %458, %.01
  br label %__nv_y1.exit

__nv_y1.exit:                                     ; preds = %__nv_log.exit.i, %369, %__internal_cos_offset.exit.i, %__nv_isinfd.exit.i, %395, %340, %3
  %r.i.5 = phi double [ %4, %3 ], [ %337, %__nv_log.exit.i ], [ %366, %340 ], [ %392, %369 ], [ %413, %395 ], [ %603, %__internal_cos_offset.exit.i ], [ 0.000000e+00, %__nv_isinfd.exit.i ]
  %604 = fcmp ugt double %a, 0.000000e+00
  %605 = fcmp oeq double %a, 0.000000e+00
  %r.i.6 = select i1 %605, double 0xFFF0000000000000, double 0xFFF8000000000000
  %r.i.7 = select i1 %604, double %r.i.5, double %r.i.6
  %606 = fcmp ugt double %1, 0x3FE97F4A8F9D3F28
  br i1 %606, label %939, label %607

607:                                              ; preds = %__nv_y1.exit
  %608 = fmul double %a, %a
  %609 = call double @llvm.fma.f64(double %608, double noundef 0x3C8EFBD0A1B77C65, double noundef 0xBD13098C51C18514)
  %610 = call double @llvm.fma.f64(double %609, double %608, double noundef 0x3D923102D2F5F2F5)
  %611 = call double @llvm.fma.f64(double %610, double %608, double noundef 0xBE0A5F2DEE7D526E)
  %612 = call double @llvm.fma.f64(double %611, double %608, double noundef 0x3E7BB77E758B38AF)
  %613 = call double @llvm.fma.f64(double %612, double %608, double noundef 0xBEE3D1A206EC4F36)
  %614 = call double @llvm.fma.f64(double %613, double %608, double noundef 0x3F4183DCD3ED6294)
  %615 = call double @llvm.fma.f64(double %614, double %608, double noundef 0xBF903921CF04F123)
  %616 = call double @llvm.fma.f64(double %615, double %608, double noundef 0x3FC5DB69D7753176)
  %617 = fadd double %608, 0xBFDBA96740000000
  %618 = fadd double %617, 0x3E15A30C80000000
  %619 = fmul double %618, %616
  %620 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %621 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %622 = icmp slt i32 %620, 1048576
  br i1 %622, label %623, label %627

623:                                              ; preds = %607
  %624 = fmul double %1, 0x4350000000000000
  %625 = call i32 @llvm.nvvm.d2i.hi(double %624) #139
  %626 = call i32 @llvm.nvvm.d2i.lo(double %624) #139
  br label %627

627:                                              ; preds = %623, %607
  %.021 = phi double [ %624, %623 ], [ %1, %607 ]
  %ihi.i.i50.0 = phi i32 [ %625, %623 ], [ %620, %607 ]
  %ilo.i.i51.0 = phi i32 [ %626, %623 ], [ %621, %607 ]
  %e.i.i52.0 = phi i32 [ -1077, %623 ], [ -1023, %607 ]
  %628 = add i32 %ihi.i.i50.0, -1
  %629 = icmp ult i32 %628, 2146435071
  br i1 %629, label %630, label %679

630:                                              ; preds = %627
  %631 = lshr i32 %ihi.i.i50.0, 20
  %632 = add nsw i32 %e.i.i52.0, %631
  %633 = and i32 %ihi.i.i50.0, -2146435073
  %634 = or i32 %633, 1072693248
  %635 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i51.0, i32 %634) #139
  %636 = icmp sgt i32 %634, 1073127582
  br i1 %636, label %637, label %643

637:                                              ; preds = %630
  %638 = call i32 @llvm.nvvm.d2i.lo(double %635) #139
  %639 = call i32 @llvm.nvvm.d2i.hi(double %635) #139
  %640 = add i32 %639, -1048576
  %641 = call double @llvm.nvvm.lohi.i2d(i32 %638, i32 %640) #139
  %642 = add nsw i32 %632, 1
  br label %643

643:                                              ; preds = %637, %630
  %m.i.i42.0 = phi double [ %641, %637 ], [ %635, %630 ]
  %e.i.i52.1 = phi i32 [ %642, %637 ], [ %632, %630 ]
  %644 = fadd double %m.i.i42.0, -1.000000e+00
  %645 = fadd double %m.i.i42.0, 1.000000e+00
  %646 = call double @llvm.nvvm.rcp.approx.ftz.d(double %645) #139
  %647 = fneg double %645
  %648 = call double @llvm.fma.f64(double %647, double %646, double noundef 1.000000e+00)
  %649 = call double @llvm.fma.f64(double %648, double %648, double %648)
  %650 = call double @llvm.fma.f64(double %649, double %646, double %646)
  %651 = fmul double %644, %650
  %652 = fadd double %651, %651
  %653 = fmul double %652, %652
  %654 = call double @llvm.fma.f64(double %653, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %655 = call double @llvm.fma.f64(double %654, double %653, double noundef 0x3EF3B2669F02676F)
  %656 = call double @llvm.fma.f64(double %655, double %653, double noundef 0x3F1745CBA9AB0956)
  %657 = call double @llvm.fma.f64(double %656, double %653, double noundef 0x3F3C71C72D1B5154)
  %658 = call double @llvm.fma.f64(double %657, double %653, double noundef 0x3F624924923BE72D)
  %659 = call double @llvm.fma.f64(double %658, double %653, double noundef 0x3F8999999999A3C4)
  %660 = call double @llvm.fma.f64(double %659, double %653, double noundef 0x3FB5555555555554)
  %661 = fsub double %644, %652
  %662 = fmul double %661, 2.000000e+00
  %663 = fneg double %652
  %664 = call double @llvm.fma.f64(double %663, double %644, double %662)
  %665 = fmul double %650, %664
  %666 = fmul double %653, %660
  %667 = call double @llvm.fma.f64(double %666, double %652, double %665)
  %668 = xor i32 %e.i.i52.1, -2147483648
  %669 = call double @llvm.nvvm.lohi.i2d(i32 %668, i32 noundef 1127219200) #139
  %670 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %671 = fsub double %669, %670
  %672 = call double @llvm.fma.f64(double %671, double noundef 0x3FE62E42FEFA39EF, double %652)
  %673 = fneg double %671
  %674 = call double @llvm.fma.f64(double %673, double noundef 0x3FE62E42FEFA39EF, double %672)
  %675 = fsub double %674, %652
  %676 = fsub double %667, %675
  %677 = call double @llvm.fma.f64(double %671, double noundef 0x3C7ABC9E3B39803F, double %676)
  %678 = fadd double %672, %677
  br label %__nv_log.exit.i102

679:                                              ; preds = %627
  %680 = call double @llvm.fma.f64(double %.021, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %681 = call i32 @llvm.nvvm.d2i.hi(double %.021) #139
  %682 = bitcast i32 %681 to float
  %683 = fcmp oeq float %682, 0.000000e+00
  %q.i3.i.0 = select i1 %683, double 0xFFF0000000000000, double %680
  br label %__nv_log.exit.i102

__nv_log.exit.i102:                               ; preds = %679, %643
  %q.i3.i.1 = phi double [ %678, %643 ], [ %q.i3.i.0, %679 ]
  %684 = fmul double %q.i3.i.1, 0x3FE45F306DC9C883
  %685 = fcmp ugt double %1, 0x400FB319F277BBE5
  br i1 %685, label %706, label %686

686:                                              ; preds = %__nv_log.exit.i102
  %687 = fadd double %1, 0xC0033D152E971B40
  %688 = fadd double %687, 0x3CA0F539D7DA258E
  %689 = call double @llvm.fma.f64(double %688, double noundef 0xBCC0D18564C48C61, double noundef 0xBCFCF8F9A8C294BC)
  %690 = call double @llvm.fma.f64(double %689, double %688, double noundef 0x3D3FAB983CAE498B)
  %691 = call double @llvm.fma.f64(double %690, double %688, double noundef 0x3D7CD7C018579B88)
  %692 = call double @llvm.fma.f64(double %691, double %688, double noundef 0xBDBBDD2342D64FDD)
  %693 = call double @llvm.fma.f64(double %692, double %688, double noundef 0xBDF5C2D9416B1E2B)
  %694 = call double @llvm.fma.f64(double %693, double %688, double noundef 0x3E32951D73174DD5)
  %695 = call double @llvm.fma.f64(double %694, double %688, double noundef 0x3E67FF99802CAEB5)
  %696 = call double @llvm.fma.f64(double %695, double %688, double noundef 0xBEA1CCE305C4C9F7)
  %697 = call double @llvm.fma.f64(double %696, double %688, double noundef 0xBED232C77E29E1BB)
  %698 = call double @llvm.fma.f64(double %697, double %688, double noundef 0x3F06ED3B9F0EF757)
  %699 = call double @llvm.fma.f64(double %698, double %688, double noundef 0x3F315382BA096A62)
  %700 = call double @llvm.fma.f64(double %699, double %688, double noundef 0xBF61F992590D1AE4)
  %701 = call double @llvm.fma.f64(double %700, double %688, double noundef 0xBF81BB1CBE1A465F)
  %702 = call double @llvm.fma.f64(double %701, double %688, double noundef 0x3FACFAE864368D84)
  %703 = call double @llvm.fma.f64(double %702, double %688, double noundef 0x3FBBA1DEEA0294A3)
  %704 = call double @llvm.fma.f64(double %703, double %688, double noundef 0xBFE09CDB36551280)
  %705 = fmul double %688, %704
  br label %__nv_j0.exit.i

706:                                              ; preds = %__nv_log.exit.i102
  %707 = fcmp ugt double %1, 0x401C58FD1A62F5EC
  br i1 %707, label %728, label %708

708:                                              ; preds = %706
  %709 = fadd double %1, 0xC016148F5B2C2E45
  %710 = fadd double %709, 0xBC975054CD60A517
  %711 = call double @llvm.fma.f64(double %710, double noundef 0x3CBCB0A8F126B343, double noundef 0x3CF83FD1F333EB61)
  %712 = call double @llvm.fma.f64(double %711, double %710, double noundef 0xBD4100E33E3FB413)
  %713 = call double @llvm.fma.f64(double %712, double %710, double noundef 0xBD7846076D004627)
  %714 = call double @llvm.fma.f64(double %713, double %710, double noundef 0x3DBE2F1D4F90720D)
  %715 = call double @llvm.fma.f64(double %714, double %710, double noundef 0x3DF1D03B1E4A119B)
  %716 = call double @llvm.fma.f64(double %715, double %710, double noundef 0xBE341D72B1B3BCE9)
  %717 = call double @llvm.fma.f64(double %716, double %710, double noundef 0xBE62DA37CE2A9EF8)
  %718 = call double @llvm.fma.f64(double %717, double %710, double noundef 0x3EA32E6D9974F763)
  %719 = call double @llvm.fma.f64(double %718, double %710, double noundef 0x3ECAD77D744A1879)
  %720 = call double @llvm.fma.f64(double %719, double %710, double noundef 0xBF0863F481A37337)
  %721 = call double @llvm.fma.f64(double %720, double %710, double noundef 0xBF26F641F418F0F4)
  %722 = call double @llvm.fma.f64(double %721, double %710, double noundef 0x3F627E31FE9A969E)
  %723 = call double @llvm.fma.f64(double %722, double %710, double noundef 0x3F72F7FFE9025628)
  %724 = call double @llvm.fma.f64(double %723, double %710, double noundef 0xBFAB2150CB41E8BF)
  %725 = call double @llvm.fma.f64(double %724, double %710, double noundef 0xBF9F8F72E7A848DE)
  %726 = call double @llvm.fma.f64(double %725, double %710, double noundef 0x3FD5C6E60A097823)
  %727 = fmul double %710, %726
  br label %__nv_j0.exit.i

728:                                              ; preds = %706
  %729 = fcmp ugt double %1, 0x402471FCB6A7A8C0
  br i1 %729, label %750, label %730

730:                                              ; preds = %728
  %731 = fadd double %1, 0xC0214EB56CCCDECA
  %732 = fadd double %731, 0x3CB51970714C7C25
  %733 = call double @llvm.fma.f64(double %732, double noundef 0xBCBDB7FFCF659E24, double noundef 0xBCF4B3A71AAAC629)
  %734 = call double @llvm.fma.f64(double %733, double %732, double noundef 0x3D417EC150ECDCE7)
  %735 = call double @llvm.fma.f64(double %734, double %732, double noundef 0x3D7438F5EA1D10B2)
  %736 = call double @llvm.fma.f64(double %735, double %732, double noundef 0xBDBEDAE7EC2C9E87)
  %737 = call double @llvm.fma.f64(double %736, double %732, double noundef 0xBDECADD2C4B91F58)
  %738 = call double @llvm.fma.f64(double %737, double %732, double noundef 0x3E34582C8EE12204)
  %739 = call double @llvm.fma.f64(double %738, double %732, double noundef 0x3E5CEDA451DD20F8)
  %740 = call double @llvm.fma.f64(double %739, double %732, double noundef 0xBEA30E8CC3165E2F)
  %741 = call double @llvm.fma.f64(double %740, double %732, double noundef 0xBEC3324842BB1A2E)
  %742 = call double @llvm.fma.f64(double %741, double %732, double noundef 0x3F07800BC54FBDDB)
  %743 = call double @llvm.fma.f64(double %742, double %732, double noundef 0x3F1D79605276949A)
  %744 = call double @llvm.fma.f64(double %743, double %732, double noundef 0xBF60E0D60385A629)
  %745 = call double @llvm.fma.f64(double %744, double %732, double noundef 0xBF648E63600D82F3)
  %746 = call double @llvm.fma.f64(double %745, double %732, double noundef 0x3FA68B984EC6493A)
  %747 = call double @llvm.fma.f64(double %746, double %732, double noundef 0x3F900F7FCF183E0B)
  %748 = call double @llvm.fma.f64(double %747, double %732, double noundef 0xBFD15F7977A772D4)
  %749 = fmul double %732, %748
  br label %__nv_j0.exit.i

750:                                              ; preds = %728
  %751 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %752 = icmp eq i32 %751, 200
  br i1 %752, label %759, label %753

753:                                              ; preds = %750
  %754 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %755 = icmp eq i32 %754, 350
  br i1 %755, label %759, label %756

756:                                              ; preds = %753
  %757 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %758 = icmp eq i32 %757, 370
  br i1 %758, label %759, label %761

759:                                              ; preds = %756, %753, %750
  %760 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i103

761:                                              ; preds = %756
  %762 = and i32 %620, 2147483647
  %763 = icmp eq i32 %762, 2146435072
  %764 = icmp eq i32 %621, 0
  %765 = select i1 %763, i1 %764, i1 false
  br label %__nv_isinfd.exit.i.i103

__nv_isinfd.exit.i.i103:                          ; preds = %761, %759
  %.028.in = phi i1 [ %760, %759 ], [ %765, %761 ]
  br i1 %.028.in, label %__nv_j0.exit.i, label %766

766:                                              ; preds = %__nv_isinfd.exit.i.i103
  %767 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %768 = fneg double %1
  %769 = call double @llvm.fma.f64(double %768, double %767, double noundef 1.000000e+00)
  %770 = call double @llvm.fma.f64(double %769, double %769, double %769)
  %771 = call double @llvm.fma.f64(double %770, double %767, double %767)
  %772 = fmul double %771, %771
  %773 = call double @llvm.fma.f64(double %772, double noundef 0xC0D115CB8C11A9DC, double noundef 0x409927467A655012)
  %774 = call double @llvm.fma.f64(double %773, double %772, double noundef 0xC05751787E247BD4)
  %775 = call double @llvm.fma.f64(double %774, double %772, double noundef 0x401704C4E5FC36B2)
  %776 = call double @llvm.fma.f64(double %775, double %772, double noundef 0xBFE15B747A2FD531)
  %777 = call double @llvm.fma.f64(double %776, double %772, double noundef 0x3FBA7FEACF6CB79B)
  %778 = call double @llvm.fma.f64(double %777, double %772, double noundef 0xBFAFFFFFEDDCF548)
  %779 = call double @llvm.fma.f64(double %778, double %772, double noundef 0x3FEFFFFFFFFFC9E5)
  %780 = call double @llvm.fma.f64(double %772, double noundef 0xC14602FE1C34685E, double noundef 0x410ECD4523B12B84)
  %781 = call double @llvm.fma.f64(double %780, double %772, double noundef 0xC0C7A2FC1972F05A)
  %782 = call double @llvm.fma.f64(double %781, double %772, double noundef 0x407EBA131F7E5BEB)
  %783 = call double @llvm.fma.f64(double %782, double %772, double noundef 0xC0373B92E6E7CC7D)
  %784 = call double @llvm.fma.f64(double %783, double %772, double noundef 0x3FFA31BEE63A2F08)
  %785 = call double @llvm.fma.f64(double %784, double %772, double noundef 0xBFCAD320104D5D05)
  %786 = call double @llvm.fma.f64(double %785, double %772, double noundef 0x3FB0AAAA9C76D07E)
  %787 = call double @llvm.fma.f64(double %786, double %772, double noundef 0xBFBFFFFFFFFDACEC)
  %788 = call double @llvm.fma.f64(double %787, double %771, double %1)
  %789 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %790 = fmul double %789, 0x3FE9884533D43651
  %791 = fmul double %790, %779
  %792 = fmul double %788, 0x3FE45F306DC9C883
  %793 = call i32 @llvm.nvvm.d2i.rn(double %792) #139
  store i32 %793, i32* %q.i.i.i20.i, align 4, !noalias !68
  %794 = sitofp i32 %793 to double
  %795 = fneg double %794
  %796 = call double @llvm.fma.f64(double %795, double noundef 0x3FF921FB54442D18, double %788)
  %797 = call double @llvm.fma.f64(double %795, double noundef 0x3C91A62633145C00, double %796)
  %798 = call double @llvm.fma.f64(double %795, double noundef 0x397B839A252049C0, double %797)
  %799 = call double @llvm.fabs.f64(double %788)
  %800 = fcmp ult double %799, 0x41E0000000000000
  br i1 %800, label %__internal_trig_reduction_kerneld.exit.i.i25.i, label %801

801:                                              ; preds = %766
  %802 = call fastcc double @__internal_trig_reduction_slowpathd(double %788, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i20.i) #139
  %.pre78 = load i32, i32* %q.i.i.i20.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i25.i

__internal_trig_reduction_kerneld.exit.i.i25.i:   ; preds = %801, %766
  %803 = phi i32 [ %.pre78, %801 ], [ %793, %766 ]
  %t.i.i.i19.i.0 = phi double [ %802, %801 ], [ %798, %766 ]
  %804 = and i32 %803, 3
  %805 = uitofp i32 %804 to double
  %806 = fadd double %t.i.i.i19.i.0, 0xBFE921FB54442D18
  %807 = fmul double %805, 0x3FF921FB54442D18
  %808 = fadd double %806, %807
  %809 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %810 = icmp eq i32 %809, 350
  br i1 %810, label %814, label %811

811:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i25.i
  %812 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %813 = icmp eq i32 %812, 370
  br i1 %813, label %814, label %873

814:                                              ; preds = %811, %__internal_trig_reduction_kerneld.exit.i.i25.i
  %815 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %816 = icmp eq i32 %815, 200
  br i1 %816, label %823, label %817

817:                                              ; preds = %814
  %818 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %819 = icmp eq i32 %818, 350
  br i1 %819, label %823, label %820

820:                                              ; preds = %817
  %821 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %822 = icmp eq i32 %821, 370
  br i1 %822, label %823, label %826

823:                                              ; preds = %820, %817, %814
  %824 = call double @llvm.fabs.f64(double %808)
  %825 = fcmp oeq double %824, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i.i104

826:                                              ; preds = %820
  %827 = call i32 @llvm.nvvm.d2i.lo(double %808) #139
  %828 = call i32 @llvm.nvvm.d2i.hi(double %808) #139
  %829 = and i32 %828, 2147483647
  %830 = icmp eq i32 %829, 2146435072
  %831 = icmp eq i32 %827, 0
  %832 = select i1 %830, i1 %831, i1 false
  br label %__nv_isinfd.exit.i.i.i.i.i104

__nv_isinfd.exit.i.i.i.i.i104:                    ; preds = %826, %823
  %.024.in = phi i1 [ %825, %823 ], [ %832, %826 ]
  br i1 %.024.in, label %833, label %835

833:                                              ; preds = %__nv_isinfd.exit.i.i.i.i.i104
  %834 = fmul double %808, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i105

835:                                              ; preds = %__nv_isinfd.exit.i.i.i.i.i104
  %836 = fmul double %808, 0x3FE45F306DC9C883
  %837 = call i32 @llvm.nvvm.d2i.rn(double %836) #139
  store i32 %837, i32* %q.i.i.i.i.i.i20, align 4, !noalias !68
  %838 = sitofp i32 %837 to double
  %839 = fneg double %838
  %840 = call double @llvm.fma.f64(double %839, double noundef 0x3FF921FB54442D18, double %808)
  %841 = call double @llvm.fma.f64(double %839, double noundef 0x3C91A62633145C00, double %840)
  %842 = call double @llvm.fma.f64(double %839, double noundef 0x397B839A252049C0, double %841)
  %843 = call double @llvm.fabs.f64(double %808)
  %844 = fcmp ult double %843, 0x41E0000000000000
  br i1 %844, label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i105, label %845

845:                                              ; preds = %835
  %846 = call fastcc double @__internal_trig_reduction_slowpathd(double %808, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i.i20) #139
  %.pre80 = load i32, i32* %q.i.i.i.i.i.i20, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i105

__internal_trig_reduction_kerneld.exit.i.i.i.i.i105: ; preds = %835, %845, %833
  %i.i.i.i.i.i26.0 = phi i32 [ 0, %833 ], [ %.pre80, %845 ], [ %837, %835 ]
  %.023 = phi double [ %834, %833 ], [ %846, %845 ], [ %842, %835 ]
  %847 = fmul double %.023, %.023
  %848 = call double @llvm.fma.f64(double %847, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %849 = call double @llvm.fma.f64(double %848, double %847, double noundef 0xBE927E4F8E06E6D9)
  %850 = call double @llvm.fma.f64(double %849, double %847, double noundef 0x3EFA01A019DDBCE9)
  %851 = call double @llvm.fma.f64(double %850, double %847, double noundef 0xBF56C16C16C15D47)
  %852 = call double @llvm.fma.f64(double %851, double %847, double noundef 0x3FA5555555555551)
  %853 = call double @llvm.fma.f64(double %852, double %847, double noundef -5.000000e-01)
  %854 = call double @llvm.fma.f64(double %853, double %847, double noundef 1.000000e+00)
  %855 = call double @llvm.fma.f64(double %847, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %856 = call double @llvm.fma.f64(double %855, double %847, double noundef 0x3EC71DE369ACE392)
  %857 = call double @llvm.fma.f64(double %856, double %847, double noundef 0xBF2A01A019DB62A1)
  %858 = call double @llvm.fma.f64(double %857, double %847, double noundef 0x3F81111111110818)
  %859 = call double @llvm.fma.f64(double %858, double %847, double noundef 0xBFC5555555555554)
  %860 = call double @llvm.fma.f64(double %859, double %847, double noundef 0.000000e+00)
  %861 = call double @llvm.fma.f64(double %860, double %.023, double %.023)
  %862 = call i32 @llvm.nvvm.d2i.hi(double %861) #139
  %863 = call i32 @llvm.nvvm.d2i.lo(double %861) #139
  %864 = xor i32 %862, -2147483648
  %865 = call double @llvm.nvvm.lohi.i2d(i32 %863, i32 %864) #139
  %866 = and i32 %i.i.i.i.i.i26.0, 1
  %.not60 = icmp eq i32 %866, 0
  %spec.select8 = select i1 %.not60, double %854, double %865
  %867 = and i32 %i.i.i.i.i.i26.0, 2
  %.not61 = icmp eq i32 %867, 0
  br i1 %.not61, label %__internal_cos_offset.exit.i.i108, label %868

868:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i105
  %869 = call i32 @llvm.nvvm.d2i.hi(double %spec.select8) #139
  %870 = call i32 @llvm.nvvm.d2i.lo(double %spec.select8) #139
  %871 = xor i32 %869, -2147483648
  %872 = call double @llvm.nvvm.lohi.i2d(i32 %870, i32 %871) #139
  br label %__internal_cos_offset.exit.i.i108

873:                                              ; preds = %811
  %874 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %875 = icmp eq i32 %874, 200
  br i1 %875, label %882, label %876

876:                                              ; preds = %873
  %877 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %878 = icmp eq i32 %877, 350
  br i1 %878, label %882, label %879

879:                                              ; preds = %876
  %880 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %881 = icmp eq i32 %880, 370
  br i1 %881, label %882, label %885

882:                                              ; preds = %879, %876, %873
  %883 = call double @llvm.fabs.f64(double %808)
  %884 = fcmp oeq double %883, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i26.i

885:                                              ; preds = %879
  %886 = call i32 @llvm.nvvm.d2i.lo(double %808) #139
  %887 = call i32 @llvm.nvvm.d2i.hi(double %808) #139
  %888 = and i32 %887, 2147483647
  %889 = icmp eq i32 %888, 2146435072
  %890 = icmp eq i32 %886, 0
  %891 = select i1 %889, i1 %890, i1 false
  br label %__nv_isinfd.exit.i.i.i26.i

__nv_isinfd.exit.i.i.i26.i:                       ; preds = %885, %882
  %.025.in = phi i1 [ %884, %882 ], [ %891, %885 ]
  br i1 %.025.in, label %892, label %894

892:                                              ; preds = %__nv_isinfd.exit.i.i.i26.i
  %893 = fmul double %808, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i107

894:                                              ; preds = %__nv_isinfd.exit.i.i.i26.i
  %895 = fmul double %808, 0x3FE45F306DC9C883
  %896 = call i32 @llvm.nvvm.d2i.rn(double %895) #139
  store i32 %896, i32* %q.i.i.i.i9.i, align 4, !noalias !68
  %897 = sitofp i32 %896 to double
  %898 = fneg double %897
  %899 = call double @llvm.fma.f64(double %898, double noundef 0x3FF921FB54442D18, double %808)
  %900 = call double @llvm.fma.f64(double %898, double noundef 0x3C91A62633145C00, double %899)
  %901 = call double @llvm.fma.f64(double %898, double noundef 0x397B839A252049C0, double %900)
  %902 = call double @llvm.fabs.f64(double %808)
  %903 = fcmp ult double %902, 0x41E0000000000000
  br i1 %903, label %__internal_sin_cos_kerneld.exit.i.i.i.i107, label %904

904:                                              ; preds = %894
  %905 = call fastcc double @__internal_trig_reduction_slowpathd(double %808, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i9.i) #139
  %.pre79 = load i32, i32* %q.i.i.i.i9.i, align 4
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i107

__internal_sin_cos_kerneld.exit.i.i.i.i107:       ; preds = %894, %904, %892
  %i.i.i.i17.i.0 = phi i32 [ 0, %892 ], [ %.pre79, %904 ], [ %896, %894 ]
  %z.i.i.i16.i.0 = phi double [ %893, %892 ], [ %905, %904 ], [ %901, %894 ]
  %906 = add nsw i32 %i.i.i.i17.i.0, 1
  %907 = shl i32 %906, 3
  %908 = and i32 %907, 8
  %909 = zext i32 %908 to i64
  %910 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %909
  %911 = fmul double %z.i.i.i16.i.0, %z.i.i.i16.i.0
  %912 = and i32 %906, 1
  %.not57 = icmp eq i32 %912, 0
  %913 = select i1 %.not57, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %914 = getelementptr inbounds double, double addrspace(1)* %910, i64 1
  %915 = load double, double addrspace(1)* %914, align 8
  %916 = call double @llvm.fma.f64(double noundef %913, double %911, double %915)
  %917 = getelementptr inbounds double, double addrspace(1)* %910, i64 2
  %918 = load double, double addrspace(1)* %917, align 8
  %919 = call double @llvm.fma.f64(double %916, double %911, double %918)
  %920 = getelementptr inbounds double, double addrspace(1)* %910, i64 3
  %921 = load double, double addrspace(1)* %920, align 8
  %922 = call double @llvm.fma.f64(double %919, double %911, double %921)
  %923 = getelementptr inbounds double, double addrspace(1)* %910, i64 4
  %924 = load double, double addrspace(1)* %923, align 8
  %925 = call double @llvm.fma.f64(double %922, double %911, double %924)
  %926 = getelementptr inbounds double, double addrspace(1)* %910, i64 5
  %927 = load double, double addrspace(1)* %926, align 8
  %928 = call double @llvm.fma.f64(double %925, double %911, double %927)
  %929 = getelementptr inbounds double, double addrspace(1)* %910, i64 6
  %930 = load double, double addrspace(1)* %929, align 8
  %931 = call double @llvm.fma.f64(double %928, double %911, double %930)
  %932 = call double @llvm.fma.f64(double %931, double %z.i.i.i16.i.0, double %z.i.i.i16.i.0)
  %933 = call double @llvm.fma.f64(double %931, double %911, double noundef 1.000000e+00)
  %spec.select9 = select i1 %.not57, double %932, double %933
  %934 = and i32 %906, 2
  %.not59 = icmp eq i32 %934, 0
  %935 = call double @llvm.fma.f64(double %spec.select9, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.127 = select i1 %.not59, double %spec.select9, double %935
  br label %__internal_cos_offset.exit.i.i108

__internal_cos_offset.exit.i.i108:                ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i105, %868, %__internal_sin_cos_kerneld.exit.i.i.i.i107
  %.022 = phi double [ %.127, %__internal_sin_cos_kerneld.exit.i.i.i.i107 ], [ %872, %868 ], [ %spec.select8, %__internal_trig_reduction_kerneld.exit.i.i.i.i.i105 ]
  %936 = fmul double %791, %.022
  br label %__nv_j0.exit.i

__nv_j0.exit.i:                                   ; preds = %708, %__internal_cos_offset.exit.i.i108, %__nv_isinfd.exit.i.i103, %730, %686
  %r.i.i29.3 = phi double [ %705, %686 ], [ %727, %708 ], [ %749, %730 ], [ %936, %__internal_cos_offset.exit.i.i108 ], [ 0.000000e+00, %__nv_isinfd.exit.i.i103 ]
  %937 = fmul double %684, %r.i.i29.3
  %938 = fadd double %619, %937
  br label %__nv_y0.exit

939:                                              ; preds = %__nv_y1.exit
  %940 = fcmp ugt double %1, 0x4000347C4AB37B18
  br i1 %940, label %968, label %941

941:                                              ; preds = %939
  %942 = fadd double %1, 0xBFEC982EB8D417EA
  %943 = fadd double %942, 0xBC7EA9D270347F83
  %944 = call double @llvm.fma.f64(double %943, double noundef 0xBF01630132D75FC3, double noundef 0x3F3D054B05D3C52D)
  %945 = call double @llvm.fma.f64(double %944, double %943, double noundef 0xBF66DAC0B314B2E5)
  %946 = call double @llvm.fma.f64(double %945, double %943, double noundef 0x3F86A5D1DE76263F)
  %947 = call double @llvm.fma.f64(double %946, double %943, double noundef 0xBF9FD16652824592)
  %948 = call double @llvm.fma.f64(double %947, double %943, double noundef 0x3FB0F69A9CC79FBD)
  %949 = call double @llvm.fma.f64(double %948, double %943, double noundef 0xBFBCCE40EF15583E)
  %950 = call double @llvm.fma.f64(double %949, double %943, double noundef 0x3FC446B11780E4FC)
  %951 = call double @llvm.fma.f64(double %950, double %943, double noundef 0xBFC89AE7E19621F7)
  %952 = call double @llvm.fma.f64(double %951, double %943, double noundef 0x3FCACBA1B38EF7B8)
  %953 = call double @llvm.fma.f64(double %952, double %943, double noundef 0xBFCB4166A03BBFA5)
  %954 = call double @llvm.fma.f64(double %953, double %943, double noundef 0x3FCACCA4D5D4889A)
  %955 = call double @llvm.fma.f64(double %954, double %943, double noundef 0xBFCA1455932B9392)
  %956 = call double @llvm.fma.f64(double %955, double %943, double noundef 0x3FC96D8DB8D844EC)
  %957 = call double @llvm.fma.f64(double %956, double %943, double noundef 0xBFC8F7FB77522EDF)
  %958 = call double @llvm.fma.f64(double %957, double %943, double noundef 0x3FC8C0926ABC9AB0)
  %959 = call double @llvm.fma.f64(double %958, double %943, double noundef 0xBFC8D35B8FEA468C)
  %960 = call double @llvm.fma.f64(double %959, double %943, double noundef 0x3FC9424B8A0C8F94)
  %961 = call double @llvm.fma.f64(double %960, double %943, double noundef 0xBFCA396A7F3403EF)
  %962 = call double @llvm.fma.f64(double %961, double %943, double noundef 0x3FCC068086C37055)
  %963 = call double @llvm.fma.f64(double %962, double %943, double noundef 0xBFCCF18E6A4C5C4E)
  %964 = call double @llvm.fma.f64(double %963, double %943, double noundef 0x3FCC3B1338AF4239)
  %965 = call double @llvm.fma.f64(double %964, double %943, double noundef 0xBFDF7E38A46D70DB)
  %966 = call double @llvm.fma.f64(double %965, double %943, double noundef 0x3FEC24371844B88A)
  %967 = fmul double %943, %966
  br label %__nv_y0.exit

968:                                              ; preds = %939
  %969 = fcmp ugt double %1, 0x40161663B5D9A628
  br i1 %969, label %997, label %970

970:                                              ; preds = %968
  %971 = fadd double %1, 0xC00FA9534D98569C
  %972 = fadd double %971, 0x3C9F06AE7804384E
  %973 = call double @llvm.fma.f64(double %972, double noundef 0xBCDAEA62AC8BDA68, double noundef 0xBCD2434958151AC7)
  %974 = call double @llvm.fma.f64(double %973, double %972, double noundef 0x3D11C24A40D33FE1)
  %975 = call double @llvm.fma.f64(double %974, double %972, double noundef 0x3D237CD62FA08CA4)
  %976 = call double @llvm.fma.f64(double %975, double %972, double noundef 0xBD43902E0298C52A)
  %977 = call double @llvm.fma.f64(double %976, double %972, double noundef 0xBD1DDAAD11CAB40F)
  %978 = call double @llvm.fma.f64(double %977, double %972, double noundef 0xBD5209D9F06D7DE4)
  %979 = call double @llvm.fma.f64(double %978, double %972, double noundef 0x3D8BB9F464468E1A)
  %980 = call double @llvm.fma.f64(double %979, double %972, double noundef 0xBDA8F67B07D1B440)
  %981 = call double @llvm.fma.f64(double %980, double %972, double noundef 0x3DC7C8D60F9EAECF)
  %982 = call double @llvm.fma.f64(double %981, double %972, double noundef 0xBDE9703405B49A8D)
  %983 = call double @llvm.fma.f64(double %982, double %972, double noundef 0x3E0A6B64E76417E4)
  %984 = call double @llvm.fma.f64(double %983, double %972, double noundef 0xBE2F6B5AFB2F1359)
  %985 = call double @llvm.fma.f64(double %984, double %972, double noundef 0x3E54526B71C21EC1)
  %986 = call double @llvm.fma.f64(double %985, double %972, double noundef 0xBE5776DBCBBC8E1D)
  %987 = call double @llvm.fma.f64(double %986, double %972, double noundef 0xBE93B211FC2DF90E)
  %988 = call double @llvm.fma.f64(double %987, double %972, double noundef 0xBED486372E8562DC)
  %989 = call double @llvm.fma.f64(double %988, double %972, double noundef 0x3F0AB2C1FBC3A254)
  %990 = call double @llvm.fma.f64(double %989, double %972, double noundef 0x3F299827653353B8)
  %991 = call double @llvm.fma.f64(double %990, double %972, double noundef 0xBF61E32BC4ED7084)
  %992 = call double @llvm.fma.f64(double %991, double %972, double noundef 0xBF7C116FDC599A09)
  %993 = call double @llvm.fma.f64(double %992, double %972, double noundef 0x3FADF6D59BF50C77)
  %994 = call double @llvm.fma.f64(double %993, double %972, double noundef 0x3FAA09C92903680B)
  %995 = call double @llvm.fma.f64(double %994, double %972, double noundef 0xBFD9C34256A12A0B)
  %996 = fmul double %972, %995
  br label %__nv_y0.exit

997:                                              ; preds = %968
  %998 = fcmp ugt double %1, 0x40214EF30C0C06ED
  br i1 %998, label %1018, label %999

999:                                              ; preds = %997
  %1000 = fadd double %1, 0xC01C581DC4E72103
  %1001 = fadd double %1000, 0x3C99774A495F56CF
  %1002 = call double @llvm.fma.f64(double %1001, double noundef 0x3CF1CB3ABA718B8E, double noundef 0xBD3F443BB4F53D75)
  %1003 = call double @llvm.fma.f64(double %1002, double %1001, double noundef 0xBD770F737BD6A786)
  %1004 = call double @llvm.fma.f64(double %1003, double %1001, double noundef 0x3DBF0E9A20459E14)
  %1005 = call double @llvm.fma.f64(double %1004, double %1001, double noundef 0x3DEFA6B137D5E108)
  %1006 = call double @llvm.fma.f64(double %1005, double %1001, double noundef 0xBE344296729FB7FA)
  %1007 = call double @llvm.fma.f64(double %1006, double %1001, double noundef 0xBE60A2813A80DFAA)
  %1008 = call double @llvm.fma.f64(double %1007, double %1001, double noundef 0x3EA34AA737A83EB4)
  %1009 = call double @llvm.fma.f64(double %1008, double %1001, double noundef 0x3EC6A9227332D03C)
  %1010 = call double @llvm.fma.f64(double %1009, double %1001, double noundef 0xBF08177E4F93C81E)
  %1011 = call double @llvm.fma.f64(double %1010, double %1001, double noundef 0xBF226DD71E391775)
  %1012 = call double @llvm.fma.f64(double %1011, double %1001, double noundef 0x3F61D35E85FD7B22)
  %1013 = call double @llvm.fma.f64(double %1012, double %1001, double noundef 0x3F6B2F14A955285C)
  %1014 = call double @llvm.fma.f64(double %1013, double %1001, double noundef 0xBFA8969C64CBF388)
  %1015 = call double @llvm.fma.f64(double %1014, double %1001, double noundef 0xBF95AEF611FC4D5A)
  %1016 = call double @llvm.fma.f64(double %1015, double %1001, double noundef 0x3FD334CCA0697A5A)
  %1017 = fmul double %1001, %1016
  br label %__nv_y0.exit

1018:                                             ; preds = %997
  %1019 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1020 = icmp eq i32 %1019, 200
  br i1 %1020, label %1027, label %1021

1021:                                             ; preds = %1018
  %1022 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1023 = icmp eq i32 %1022, 350
  br i1 %1023, label %1027, label %1024

1024:                                             ; preds = %1021
  %1025 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1026 = icmp eq i32 %1025, 370
  br i1 %1026, label %1027, label %1029

1027:                                             ; preds = %1024, %1021, %1018
  %1028 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i109

1029:                                             ; preds = %1024
  %1030 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %1031 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %1032 = and i32 %1031, 2147483647
  %1033 = icmp eq i32 %1032, 2146435072
  %1034 = icmp eq i32 %1030, 0
  %1035 = select i1 %1033, i1 %1034, i1 false
  br label %__nv_isinfd.exit.i109

__nv_isinfd.exit.i109:                            ; preds = %1029, %1027
  %.029.in = phi i1 [ %1028, %1027 ], [ %1035, %1029 ]
  br i1 %.029.in, label %__nv_y0.exit, label %1036

1036:                                             ; preds = %__nv_isinfd.exit.i109
  %1037 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %1038 = fneg double %1
  %1039 = call double @llvm.fma.f64(double %1038, double %1037, double noundef 1.000000e+00)
  %1040 = call double @llvm.fma.f64(double %1039, double %1039, double %1039)
  %1041 = call double @llvm.fma.f64(double %1040, double %1037, double %1037)
  %1042 = fmul double %1041, %1041
  %1043 = call double @llvm.fma.f64(double %1042, double noundef 0xC0C5E91E6AC3AD03, double noundef 0x4093F56A049CDDE7)
  %1044 = call double @llvm.fma.f64(double %1043, double %1042, double noundef 0xC05572D39DFB8433)
  %1045 = call double @llvm.fma.f64(double %1044, double %1042, double noundef 0x4016A6041CAA59E5)
  %1046 = call double @llvm.fma.f64(double %1045, double %1042, double noundef 0xBFE155E3A0493880)
  %1047 = call double @llvm.fma.f64(double %1046, double %1042, double noundef 0x3FBA7FB92F417F7F)
  %1048 = call double @llvm.fma.f64(double %1047, double %1042, double noundef 0xBFAFFFFFB12E32F5)
  %1049 = call double @llvm.fma.f64(double %1048, double %1042, double noundef 0x3FEFFFFFFFFECED5)
  %1050 = call double @llvm.fma.f64(double %1042, double noundef 0x418A86A64BE101DC, double noundef 0xC15709C79AAC5813)
  %1051 = call double @llvm.fma.f64(double %1050, double %1042, double noundef 0x41142A31C980A287)
  %1052 = call double @llvm.fma.f64(double %1051, double %1042, double noundef 0xC0C9CBE68930485D)
  %1053 = call double @llvm.fma.f64(double %1052, double %1042, double noundef 0x407F583E14E8A4E8)
  %1054 = call double @llvm.fma.f64(double %1053, double %1042, double noundef 0xC0374A629C650680)
  %1055 = call double @llvm.fma.f64(double %1054, double %1042, double noundef 0x3FFA32A7AF17FAE9)
  %1056 = call double @llvm.fma.f64(double %1055, double %1042, double noundef 0xBFCAD32497785CD6)
  %1057 = call double @llvm.fma.f64(double %1056, double %1042, double noundef 0x3FB0AAAA9FB75F7B)
  %1058 = call double @llvm.fma.f64(double %1057, double %1042, double noundef 0xBFBFFFFFFFFE320F)
  %1059 = call double @llvm.fma.f64(double %1058, double %1041, double %1)
  %1060 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %1061 = fmul double %1060, 0x3FE9884533D43651
  %1062 = fmul double %1061, %1049
  %1063 = fmul double %1059, 0x3FE45F306DC9C883
  %1064 = call i32 @llvm.nvvm.d2i.rn(double %1063) #139
  store i32 %1064, i32* %q.i.i.i88, align 4, !noalias !68
  %1065 = sitofp i32 %1064 to double
  %1066 = fneg double %1065
  %1067 = call double @llvm.fma.f64(double %1066, double noundef 0x3FF921FB54442D18, double %1059)
  %1068 = call double @llvm.fma.f64(double %1066, double noundef 0x3C91A62633145C00, double %1067)
  %1069 = call double @llvm.fma.f64(double %1066, double noundef 0x397B839A252049C0, double %1068)
  %1070 = call double @llvm.fabs.f64(double %1059)
  %1071 = fcmp ult double %1070, 0x41E0000000000000
  br i1 %1071, label %__internal_trig_reduction_kerneld.exit.i.i110, label %1072

1072:                                             ; preds = %1036
  %1073 = call fastcc double @__internal_trig_reduction_slowpathd(double %1059, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i88) #139
  %.pre81 = load i32, i32* %q.i.i.i88, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i110

__internal_trig_reduction_kerneld.exit.i.i110:    ; preds = %1072, %1036
  %1074 = phi i32 [ %.pre81, %1072 ], [ %1064, %1036 ]
  %t.i.i.i87.0 = phi double [ %1073, %1072 ], [ %1069, %1036 ]
  %1075 = and i32 %1074, 3
  %1076 = uitofp i32 %1075 to double
  %1077 = fadd double %t.i.i.i87.0, 0xC002D97C7F3321D2
  %1078 = fmul double %1076, 0x3FF921FB54442D18
  %1079 = fadd double %1077, %1078
  %1080 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1081 = icmp eq i32 %1080, 350
  br i1 %1081, label %1085, label %1082

1082:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i110
  %1083 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1084 = icmp eq i32 %1083, 370
  br i1 %1084, label %1085, label %1144

1085:                                             ; preds = %1082, %__internal_trig_reduction_kerneld.exit.i.i110
  %1086 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1087 = icmp eq i32 %1086, 200
  br i1 %1087, label %1094, label %1088

1088:                                             ; preds = %1085
  %1089 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1090 = icmp eq i32 %1089, 350
  br i1 %1090, label %1094, label %1091

1091:                                             ; preds = %1088
  %1092 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1093 = icmp eq i32 %1092, 370
  br i1 %1093, label %1094, label %1097

1094:                                             ; preds = %1091, %1088, %1085
  %1095 = call double @llvm.fabs.f64(double %1079)
  %1096 = fcmp oeq double %1095, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i111

1097:                                             ; preds = %1091
  %1098 = call i32 @llvm.nvvm.d2i.lo(double %1079) #139
  %1099 = call i32 @llvm.nvvm.d2i.hi(double %1079) #139
  %1100 = and i32 %1099, 2147483647
  %1101 = icmp eq i32 %1100, 2146435072
  %1102 = icmp eq i32 %1098, 0
  %1103 = select i1 %1101, i1 %1102, i1 false
  br label %__nv_isinfd.exit.i.i.i.i111

__nv_isinfd.exit.i.i.i.i111:                      ; preds = %1097, %1094
  %.017.in = phi i1 [ %1096, %1094 ], [ %1103, %1097 ]
  br i1 %.017.in, label %1104, label %1106

1104:                                             ; preds = %__nv_isinfd.exit.i.i.i.i111
  %1105 = fmul double %1079, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i112

1106:                                             ; preds = %__nv_isinfd.exit.i.i.i.i111
  %1107 = fmul double %1079, 0x3FE45F306DC9C883
  %1108 = call i32 @llvm.nvvm.d2i.rn(double %1107) #139
  store i32 %1108, i32* %q.i.i.i.i.i74, align 4, !noalias !68
  %1109 = sitofp i32 %1108 to double
  %1110 = fneg double %1109
  %1111 = call double @llvm.fma.f64(double %1110, double noundef 0x3FF921FB54442D18, double %1079)
  %1112 = call double @llvm.fma.f64(double %1110, double noundef 0x3C91A62633145C00, double %1111)
  %1113 = call double @llvm.fma.f64(double %1110, double noundef 0x397B839A252049C0, double %1112)
  %1114 = call double @llvm.fabs.f64(double %1079)
  %1115 = fcmp ult double %1114, 0x41E0000000000000
  br i1 %1115, label %__internal_trig_reduction_kerneld.exit.i.i.i.i112, label %1116

1116:                                             ; preds = %1106
  %1117 = call fastcc double @__internal_trig_reduction_slowpathd(double %1079, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i74) #139
  %.pre83 = load i32, i32* %q.i.i.i.i.i74, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i112

__internal_trig_reduction_kerneld.exit.i.i.i.i112: ; preds = %1106, %1116, %1104
  %i.i.i.i.i81.0 = phi i32 [ 0, %1104 ], [ %.pre83, %1116 ], [ %1108, %1106 ]
  %.016 = phi double [ %1105, %1104 ], [ %1117, %1116 ], [ %1113, %1106 ]
  %1118 = fmul double %.016, %.016
  %1119 = call double @llvm.fma.f64(double %1118, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %1120 = call double @llvm.fma.f64(double %1119, double %1118, double noundef 0xBE927E4F8E06E6D9)
  %1121 = call double @llvm.fma.f64(double %1120, double %1118, double noundef 0x3EFA01A019DDBCE9)
  %1122 = call double @llvm.fma.f64(double %1121, double %1118, double noundef 0xBF56C16C16C15D47)
  %1123 = call double @llvm.fma.f64(double %1122, double %1118, double noundef 0x3FA5555555555551)
  %1124 = call double @llvm.fma.f64(double %1123, double %1118, double noundef -5.000000e-01)
  %1125 = call double @llvm.fma.f64(double %1124, double %1118, double noundef 1.000000e+00)
  %1126 = call double @llvm.fma.f64(double %1118, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %1127 = call double @llvm.fma.f64(double %1126, double %1118, double noundef 0x3EC71DE369ACE392)
  %1128 = call double @llvm.fma.f64(double %1127, double %1118, double noundef 0xBF2A01A019DB62A1)
  %1129 = call double @llvm.fma.f64(double %1128, double %1118, double noundef 0x3F81111111110818)
  %1130 = call double @llvm.fma.f64(double %1129, double %1118, double noundef 0xBFC5555555555554)
  %1131 = call double @llvm.fma.f64(double %1130, double %1118, double noundef 0.000000e+00)
  %1132 = call double @llvm.fma.f64(double %1131, double %.016, double %.016)
  %1133 = call i32 @llvm.nvvm.d2i.hi(double %1132) #139
  %1134 = call i32 @llvm.nvvm.d2i.lo(double %1132) #139
  %1135 = xor i32 %1133, -2147483648
  %1136 = call double @llvm.nvvm.lohi.i2d(i32 %1134, i32 %1135) #139
  %1137 = and i32 %i.i.i.i.i81.0, 1
  %.not55 = icmp eq i32 %1137, 0
  %spec.select11 = select i1 %.not55, double %1125, double %1136
  %1138 = and i32 %i.i.i.i.i81.0, 2
  %.not56 = icmp eq i32 %1138, 0
  br i1 %.not56, label %__internal_cos_offset.exit.i117, label %1139

1139:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i112
  %1140 = call i32 @llvm.nvvm.d2i.hi(double %spec.select11) #139
  %1141 = call i32 @llvm.nvvm.d2i.lo(double %spec.select11) #139
  %1142 = xor i32 %1140, -2147483648
  %1143 = call double @llvm.nvvm.lohi.i2d(i32 %1141, i32 %1142) #139
  br label %__internal_cos_offset.exit.i117

1144:                                             ; preds = %1082
  %1145 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1146 = icmp eq i32 %1145, 200
  br i1 %1146, label %1153, label %1147

1147:                                             ; preds = %1144
  %1148 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1149 = icmp eq i32 %1148, 350
  br i1 %1149, label %1153, label %1150

1150:                                             ; preds = %1147
  %1151 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1152 = icmp eq i32 %1151, 370
  br i1 %1152, label %1153, label %1156

1153:                                             ; preds = %1150, %1147, %1144
  %1154 = call double @llvm.fabs.f64(double %1079)
  %1155 = fcmp oeq double %1154, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i114

1156:                                             ; preds = %1150
  %1157 = call i32 @llvm.nvvm.d2i.lo(double %1079) #139
  %1158 = call i32 @llvm.nvvm.d2i.hi(double %1079) #139
  %1159 = and i32 %1158, 2147483647
  %1160 = icmp eq i32 %1159, 2146435072
  %1161 = icmp eq i32 %1157, 0
  %1162 = select i1 %1160, i1 %1161, i1 false
  br label %__nv_isinfd.exit.i.i.i114

__nv_isinfd.exit.i.i.i114:                        ; preds = %1156, %1153
  %.018.in = phi i1 [ %1155, %1153 ], [ %1162, %1156 ]
  br i1 %.018.in, label %1163, label %1165

1163:                                             ; preds = %__nv_isinfd.exit.i.i.i114
  %1164 = fmul double %1079, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i116

1165:                                             ; preds = %__nv_isinfd.exit.i.i.i114
  %1166 = fmul double %1079, 0x3FE45F306DC9C883
  %1167 = call i32 @llvm.nvvm.d2i.rn(double %1166) #139
  store i32 %1167, i32* %q.i.i.i.i58, align 4, !noalias !68
  %1168 = sitofp i32 %1167 to double
  %1169 = fneg double %1168
  %1170 = call double @llvm.fma.f64(double %1169, double noundef 0x3FF921FB54442D18, double %1079)
  %1171 = call double @llvm.fma.f64(double %1169, double noundef 0x3C91A62633145C00, double %1170)
  %1172 = call double @llvm.fma.f64(double %1169, double noundef 0x397B839A252049C0, double %1171)
  %1173 = call double @llvm.fabs.f64(double %1079)
  %1174 = fcmp ult double %1173, 0x41E0000000000000
  br i1 %1174, label %__internal_sin_cos_kerneld.exit.i.i.i116, label %1175

1175:                                             ; preds = %1165
  %1176 = call fastcc double @__internal_trig_reduction_slowpathd(double %1079, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i58) #139
  %.pre82 = load i32, i32* %q.i.i.i.i58, align 4
  br label %__internal_sin_cos_kerneld.exit.i.i.i116

__internal_sin_cos_kerneld.exit.i.i.i116:         ; preds = %1165, %1175, %1163
  %i.i.i.i85.0 = phi i32 [ 0, %1163 ], [ %.pre82, %1175 ], [ %1167, %1165 ]
  %z.i.i.i84.0 = phi double [ %1164, %1163 ], [ %1176, %1175 ], [ %1172, %1165 ]
  %1177 = add nsw i32 %i.i.i.i85.0, 1
  %1178 = shl i32 %1177, 3
  %1179 = and i32 %1178, 8
  %1180 = zext i32 %1179 to i64
  %1181 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %1180
  %1182 = fmul double %z.i.i.i84.0, %z.i.i.i84.0
  %1183 = and i32 %1177, 1
  %.not29 = icmp eq i32 %1183, 0
  %1184 = select i1 %.not29, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %1185 = getelementptr inbounds double, double addrspace(1)* %1181, i64 1
  %1186 = load double, double addrspace(1)* %1185, align 8
  %1187 = call double @llvm.fma.f64(double noundef %1184, double %1182, double %1186)
  %1188 = getelementptr inbounds double, double addrspace(1)* %1181, i64 2
  %1189 = load double, double addrspace(1)* %1188, align 8
  %1190 = call double @llvm.fma.f64(double %1187, double %1182, double %1189)
  %1191 = getelementptr inbounds double, double addrspace(1)* %1181, i64 3
  %1192 = load double, double addrspace(1)* %1191, align 8
  %1193 = call double @llvm.fma.f64(double %1190, double %1182, double %1192)
  %1194 = getelementptr inbounds double, double addrspace(1)* %1181, i64 4
  %1195 = load double, double addrspace(1)* %1194, align 8
  %1196 = call double @llvm.fma.f64(double %1193, double %1182, double %1195)
  %1197 = getelementptr inbounds double, double addrspace(1)* %1181, i64 5
  %1198 = load double, double addrspace(1)* %1197, align 8
  %1199 = call double @llvm.fma.f64(double %1196, double %1182, double %1198)
  %1200 = getelementptr inbounds double, double addrspace(1)* %1181, i64 6
  %1201 = load double, double addrspace(1)* %1200, align 8
  %1202 = call double @llvm.fma.f64(double %1199, double %1182, double %1201)
  %1203 = call double @llvm.fma.f64(double %1202, double %z.i.i.i84.0, double %z.i.i.i84.0)
  %1204 = call double @llvm.fma.f64(double %1202, double %1182, double noundef 1.000000e+00)
  %spec.select12 = select i1 %.not29, double %1203, double %1204
  %1205 = and i32 %1177, 2
  %.not31 = icmp eq i32 %1205, 0
  %1206 = call double @llvm.fma.f64(double %spec.select12, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.120 = select i1 %.not31, double %spec.select12, double %1206
  br label %__internal_cos_offset.exit.i117

__internal_cos_offset.exit.i117:                  ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i112, %1139, %__internal_sin_cos_kerneld.exit.i.i.i116
  %.015 = phi double [ %.120, %__internal_sin_cos_kerneld.exit.i.i.i116 ], [ %1143, %1139 ], [ %spec.select11, %__internal_trig_reduction_kerneld.exit.i.i.i.i112 ]
  %1207 = fmul double %1062, %.015
  br label %__nv_y0.exit

__nv_y0.exit:                                     ; preds = %941, %999, %__nv_isinfd.exit.i109, %__internal_cos_offset.exit.i117, %970, %__nv_j0.exit.i
  %r.i95.4 = phi double [ %938, %__nv_j0.exit.i ], [ %967, %941 ], [ %996, %970 ], [ %1017, %999 ], [ %1207, %__internal_cos_offset.exit.i117 ], [ 0.000000e+00, %__nv_isinfd.exit.i109 ]
  %1208 = fcmp olt double %a, 0.000000e+00
  %r.i95.5 = select i1 %1208, double 0xFFF8000000000000, double %r.i95.4
  switch i32 %n, label %2415 [
    i32 0, label %1209
    i32 1, label %1811
  ]

1209:                                             ; preds = %__nv_y0.exit
  br i1 %606, label %1542, label %1210

1210:                                             ; preds = %1209
  %1211 = fmul double %a, %a
  %1212 = call double @llvm.fma.f64(double %1211, double noundef 0x3C8EFBD0A1B77C65, double noundef 0xBD13098C51C18514)
  %1213 = call double @llvm.fma.f64(double %1212, double %1211, double noundef 0x3D923102D2F5F2F5)
  %1214 = call double @llvm.fma.f64(double %1213, double %1211, double noundef 0xBE0A5F2DEE7D526E)
  %1215 = call double @llvm.fma.f64(double %1214, double %1211, double noundef 0x3E7BB77E758B38AF)
  %1216 = call double @llvm.fma.f64(double %1215, double %1211, double noundef 0xBEE3D1A206EC4F36)
  %1217 = call double @llvm.fma.f64(double %1216, double %1211, double noundef 0x3F4183DCD3ED6294)
  %1218 = call double @llvm.fma.f64(double %1217, double %1211, double noundef 0xBF903921CF04F123)
  %1219 = call double @llvm.fma.f64(double %1218, double %1211, double noundef 0x3FC5DB69D7753176)
  %1220 = fadd double %1211, 0xBFDBA96740000000
  %1221 = fadd double %1220, 0x3E15A30C80000000
  %1222 = fmul double %1221, %1219
  %1223 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %1224 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %1225 = icmp slt i32 %1223, 1048576
  br i1 %1225, label %1226, label %1230

1226:                                             ; preds = %1210
  %1227 = fmul double %1, 0x4350000000000000
  %1228 = call i32 @llvm.nvvm.d2i.hi(double %1227) #139
  %1229 = call i32 @llvm.nvvm.d2i.lo(double %1227) #139
  br label %1230

1230:                                             ; preds = %1226, %1210
  %.036 = phi double [ %1227, %1226 ], [ %1, %1210 ]
  %ihi.i.i194.0 = phi i32 [ %1228, %1226 ], [ %1223, %1210 ]
  %ilo.i.i195.0 = phi i32 [ %1229, %1226 ], [ %1224, %1210 ]
  %e.i.i196.0 = phi i32 [ -1077, %1226 ], [ -1023, %1210 ]
  %1231 = add i32 %ihi.i.i194.0, -1
  %1232 = icmp ult i32 %1231, 2146435071
  br i1 %1232, label %1233, label %1282

1233:                                             ; preds = %1230
  %1234 = lshr i32 %ihi.i.i194.0, 20
  %1235 = add nsw i32 %e.i.i196.0, %1234
  %1236 = and i32 %ihi.i.i194.0, -2146435073
  %1237 = or i32 %1236, 1072693248
  %1238 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i195.0, i32 %1237) #139
  %1239 = icmp sgt i32 %1237, 1073127582
  br i1 %1239, label %1240, label %1246

1240:                                             ; preds = %1233
  %1241 = call i32 @llvm.nvvm.d2i.lo(double %1238) #139
  %1242 = call i32 @llvm.nvvm.d2i.hi(double %1238) #139
  %1243 = add i32 %1242, -1048576
  %1244 = call double @llvm.nvvm.lohi.i2d(i32 %1241, i32 %1243) #139
  %1245 = add nsw i32 %1235, 1
  br label %1246

1246:                                             ; preds = %1240, %1233
  %m.i.i184.0 = phi double [ %1244, %1240 ], [ %1238, %1233 ]
  %e.i.i196.1 = phi i32 [ %1245, %1240 ], [ %1235, %1233 ]
  %1247 = fadd double %m.i.i184.0, -1.000000e+00
  %1248 = fadd double %m.i.i184.0, 1.000000e+00
  %1249 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1248) #139
  %1250 = fneg double %1248
  %1251 = call double @llvm.fma.f64(double %1250, double %1249, double noundef 1.000000e+00)
  %1252 = call double @llvm.fma.f64(double %1251, double %1251, double %1251)
  %1253 = call double @llvm.fma.f64(double %1252, double %1249, double %1249)
  %1254 = fmul double %1247, %1253
  %1255 = fadd double %1254, %1254
  %1256 = fmul double %1255, %1255
  %1257 = call double @llvm.fma.f64(double %1256, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %1258 = call double @llvm.fma.f64(double %1257, double %1256, double noundef 0x3EF3B2669F02676F)
  %1259 = call double @llvm.fma.f64(double %1258, double %1256, double noundef 0x3F1745CBA9AB0956)
  %1260 = call double @llvm.fma.f64(double %1259, double %1256, double noundef 0x3F3C71C72D1B5154)
  %1261 = call double @llvm.fma.f64(double %1260, double %1256, double noundef 0x3F624924923BE72D)
  %1262 = call double @llvm.fma.f64(double %1261, double %1256, double noundef 0x3F8999999999A3C4)
  %1263 = call double @llvm.fma.f64(double %1262, double %1256, double noundef 0x3FB5555555555554)
  %1264 = fsub double %1247, %1255
  %1265 = fmul double %1264, 2.000000e+00
  %1266 = fneg double %1255
  %1267 = call double @llvm.fma.f64(double %1266, double %1247, double %1265)
  %1268 = fmul double %1253, %1267
  %1269 = fmul double %1256, %1263
  %1270 = call double @llvm.fma.f64(double %1269, double %1255, double %1268)
  %1271 = xor i32 %e.i.i196.1, -2147483648
  %1272 = call double @llvm.nvvm.lohi.i2d(i32 %1271, i32 noundef 1127219200) #139
  %1273 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %1274 = fsub double %1272, %1273
  %1275 = call double @llvm.fma.f64(double %1274, double noundef 0x3FE62E42FEFA39EF, double %1255)
  %1276 = fneg double %1274
  %1277 = call double @llvm.fma.f64(double %1276, double noundef 0x3FE62E42FEFA39EF, double %1275)
  %1278 = fsub double %1277, %1255
  %1279 = fsub double %1270, %1278
  %1280 = call double @llvm.fma.f64(double %1274, double noundef 0x3C7ABC9E3B39803F, double %1279)
  %1281 = fadd double %1275, %1280
  br label %__nv_log.exit.i246

1282:                                             ; preds = %1230
  %1283 = call double @llvm.fma.f64(double %.036, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %1284 = call i32 @llvm.nvvm.d2i.hi(double %.036) #139
  %1285 = bitcast i32 %1284 to float
  %1286 = fcmp oeq float %1285, 0.000000e+00
  %q.i3.i190.0 = select i1 %1286, double 0xFFF0000000000000, double %1283
  br label %__nv_log.exit.i246

__nv_log.exit.i246:                               ; preds = %1282, %1246
  %q.i3.i190.1 = phi double [ %1281, %1246 ], [ %q.i3.i190.0, %1282 ]
  %1287 = fmul double %q.i3.i190.1, 0x3FE45F306DC9C883
  %1288 = fcmp ugt double %1, 0x400FB319F277BBE5
  br i1 %1288, label %1309, label %1289

1289:                                             ; preds = %__nv_log.exit.i246
  %1290 = fadd double %1, 0xC0033D152E971B40
  %1291 = fadd double %1290, 0x3CA0F539D7DA258E
  %1292 = call double @llvm.fma.f64(double %1291, double noundef 0xBCC0D18564C48C61, double noundef 0xBCFCF8F9A8C294BC)
  %1293 = call double @llvm.fma.f64(double %1292, double %1291, double noundef 0x3D3FAB983CAE498B)
  %1294 = call double @llvm.fma.f64(double %1293, double %1291, double noundef 0x3D7CD7C018579B88)
  %1295 = call double @llvm.fma.f64(double %1294, double %1291, double noundef 0xBDBBDD2342D64FDD)
  %1296 = call double @llvm.fma.f64(double %1295, double %1291, double noundef 0xBDF5C2D9416B1E2B)
  %1297 = call double @llvm.fma.f64(double %1296, double %1291, double noundef 0x3E32951D73174DD5)
  %1298 = call double @llvm.fma.f64(double %1297, double %1291, double noundef 0x3E67FF99802CAEB5)
  %1299 = call double @llvm.fma.f64(double %1298, double %1291, double noundef 0xBEA1CCE305C4C9F7)
  %1300 = call double @llvm.fma.f64(double %1299, double %1291, double noundef 0xBED232C77E29E1BB)
  %1301 = call double @llvm.fma.f64(double %1300, double %1291, double noundef 0x3F06ED3B9F0EF757)
  %1302 = call double @llvm.fma.f64(double %1301, double %1291, double noundef 0x3F315382BA096A62)
  %1303 = call double @llvm.fma.f64(double %1302, double %1291, double noundef 0xBF61F992590D1AE4)
  %1304 = call double @llvm.fma.f64(double %1303, double %1291, double noundef 0xBF81BB1CBE1A465F)
  %1305 = call double @llvm.fma.f64(double %1304, double %1291, double noundef 0x3FACFAE864368D84)
  %1306 = call double @llvm.fma.f64(double %1305, double %1291, double noundef 0x3FBBA1DEEA0294A3)
  %1307 = call double @llvm.fma.f64(double %1306, double %1291, double noundef 0xBFE09CDB36551280)
  %1308 = fmul double %1291, %1307
  br label %__nv_j0.exit.i256

1309:                                             ; preds = %__nv_log.exit.i246
  %1310 = fcmp ugt double %1, 0x401C58FD1A62F5EC
  br i1 %1310, label %1331, label %1311

1311:                                             ; preds = %1309
  %1312 = fadd double %1, 0xC016148F5B2C2E45
  %1313 = fadd double %1312, 0xBC975054CD60A517
  %1314 = call double @llvm.fma.f64(double %1313, double noundef 0x3CBCB0A8F126B343, double noundef 0x3CF83FD1F333EB61)
  %1315 = call double @llvm.fma.f64(double %1314, double %1313, double noundef 0xBD4100E33E3FB413)
  %1316 = call double @llvm.fma.f64(double %1315, double %1313, double noundef 0xBD7846076D004627)
  %1317 = call double @llvm.fma.f64(double %1316, double %1313, double noundef 0x3DBE2F1D4F90720D)
  %1318 = call double @llvm.fma.f64(double %1317, double %1313, double noundef 0x3DF1D03B1E4A119B)
  %1319 = call double @llvm.fma.f64(double %1318, double %1313, double noundef 0xBE341D72B1B3BCE9)
  %1320 = call double @llvm.fma.f64(double %1319, double %1313, double noundef 0xBE62DA37CE2A9EF8)
  %1321 = call double @llvm.fma.f64(double %1320, double %1313, double noundef 0x3EA32E6D9974F763)
  %1322 = call double @llvm.fma.f64(double %1321, double %1313, double noundef 0x3ECAD77D744A1879)
  %1323 = call double @llvm.fma.f64(double %1322, double %1313, double noundef 0xBF0863F481A37337)
  %1324 = call double @llvm.fma.f64(double %1323, double %1313, double noundef 0xBF26F641F418F0F4)
  %1325 = call double @llvm.fma.f64(double %1324, double %1313, double noundef 0x3F627E31FE9A969E)
  %1326 = call double @llvm.fma.f64(double %1325, double %1313, double noundef 0x3F72F7FFE9025628)
  %1327 = call double @llvm.fma.f64(double %1326, double %1313, double noundef 0xBFAB2150CB41E8BF)
  %1328 = call double @llvm.fma.f64(double %1327, double %1313, double noundef 0xBF9F8F72E7A848DE)
  %1329 = call double @llvm.fma.f64(double %1328, double %1313, double noundef 0x3FD5C6E60A097823)
  %1330 = fmul double %1313, %1329
  br label %__nv_j0.exit.i256

1331:                                             ; preds = %1309
  %1332 = fcmp ugt double %1, 0x402471FCB6A7A8C0
  br i1 %1332, label %1353, label %1333

1333:                                             ; preds = %1331
  %1334 = fadd double %1, 0xC0214EB56CCCDECA
  %1335 = fadd double %1334, 0x3CB51970714C7C25
  %1336 = call double @llvm.fma.f64(double %1335, double noundef 0xBCBDB7FFCF659E24, double noundef 0xBCF4B3A71AAAC629)
  %1337 = call double @llvm.fma.f64(double %1336, double %1335, double noundef 0x3D417EC150ECDCE7)
  %1338 = call double @llvm.fma.f64(double %1337, double %1335, double noundef 0x3D7438F5EA1D10B2)
  %1339 = call double @llvm.fma.f64(double %1338, double %1335, double noundef 0xBDBEDAE7EC2C9E87)
  %1340 = call double @llvm.fma.f64(double %1339, double %1335, double noundef 0xBDECADD2C4B91F58)
  %1341 = call double @llvm.fma.f64(double %1340, double %1335, double noundef 0x3E34582C8EE12204)
  %1342 = call double @llvm.fma.f64(double %1341, double %1335, double noundef 0x3E5CEDA451DD20F8)
  %1343 = call double @llvm.fma.f64(double %1342, double %1335, double noundef 0xBEA30E8CC3165E2F)
  %1344 = call double @llvm.fma.f64(double %1343, double %1335, double noundef 0xBEC3324842BB1A2E)
  %1345 = call double @llvm.fma.f64(double %1344, double %1335, double noundef 0x3F07800BC54FBDDB)
  %1346 = call double @llvm.fma.f64(double %1345, double %1335, double noundef 0x3F1D79605276949A)
  %1347 = call double @llvm.fma.f64(double %1346, double %1335, double noundef 0xBF60E0D60385A629)
  %1348 = call double @llvm.fma.f64(double %1347, double %1335, double noundef 0xBF648E63600D82F3)
  %1349 = call double @llvm.fma.f64(double %1348, double %1335, double noundef 0x3FA68B984EC6493A)
  %1350 = call double @llvm.fma.f64(double %1349, double %1335, double noundef 0x3F900F7FCF183E0B)
  %1351 = call double @llvm.fma.f64(double %1350, double %1335, double noundef 0xBFD15F7977A772D4)
  %1352 = fmul double %1335, %1351
  br label %__nv_j0.exit.i256

1353:                                             ; preds = %1331
  %1354 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1355 = icmp eq i32 %1354, 200
  br i1 %1355, label %1362, label %1356

1356:                                             ; preds = %1353
  %1357 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1358 = icmp eq i32 %1357, 350
  br i1 %1358, label %1362, label %1359

1359:                                             ; preds = %1356
  %1360 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1361 = icmp eq i32 %1360, 370
  br i1 %1361, label %1362, label %1364

1362:                                             ; preds = %1359, %1356, %1353
  %1363 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i247

1364:                                             ; preds = %1359
  %1365 = and i32 %1223, 2147483647
  %1366 = icmp eq i32 %1365, 2146435072
  %1367 = icmp eq i32 %1224, 0
  %1368 = select i1 %1366, i1 %1367, i1 false
  br label %__nv_isinfd.exit.i.i247

__nv_isinfd.exit.i.i247:                          ; preds = %1364, %1362
  %.043.in = phi i1 [ %1363, %1362 ], [ %1368, %1364 ]
  br i1 %.043.in, label %__nv_j0.exit.i256, label %1369

1369:                                             ; preds = %__nv_isinfd.exit.i.i247
  %1370 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %1371 = fneg double %1
  %1372 = call double @llvm.fma.f64(double %1371, double %1370, double noundef 1.000000e+00)
  %1373 = call double @llvm.fma.f64(double %1372, double %1372, double %1372)
  %1374 = call double @llvm.fma.f64(double %1373, double %1370, double %1370)
  %1375 = fmul double %1374, %1374
  %1376 = call double @llvm.fma.f64(double %1375, double noundef 0xC0D115CB8C11A9DC, double noundef 0x409927467A655012)
  %1377 = call double @llvm.fma.f64(double %1376, double %1375, double noundef 0xC05751787E247BD4)
  %1378 = call double @llvm.fma.f64(double %1377, double %1375, double noundef 0x401704C4E5FC36B2)
  %1379 = call double @llvm.fma.f64(double %1378, double %1375, double noundef 0xBFE15B747A2FD531)
  %1380 = call double @llvm.fma.f64(double %1379, double %1375, double noundef 0x3FBA7FEACF6CB79B)
  %1381 = call double @llvm.fma.f64(double %1380, double %1375, double noundef 0xBFAFFFFFEDDCF548)
  %1382 = call double @llvm.fma.f64(double %1381, double %1375, double noundef 0x3FEFFFFFFFFFC9E5)
  %1383 = call double @llvm.fma.f64(double %1375, double noundef 0xC14602FE1C34685E, double noundef 0x410ECD4523B12B84)
  %1384 = call double @llvm.fma.f64(double %1383, double %1375, double noundef 0xC0C7A2FC1972F05A)
  %1385 = call double @llvm.fma.f64(double %1384, double %1375, double noundef 0x407EBA131F7E5BEB)
  %1386 = call double @llvm.fma.f64(double %1385, double %1375, double noundef 0xC0373B92E6E7CC7D)
  %1387 = call double @llvm.fma.f64(double %1386, double %1375, double noundef 0x3FFA31BEE63A2F08)
  %1388 = call double @llvm.fma.f64(double %1387, double %1375, double noundef 0xBFCAD320104D5D05)
  %1389 = call double @llvm.fma.f64(double %1388, double %1375, double noundef 0x3FB0AAAA9C76D07E)
  %1390 = call double @llvm.fma.f64(double %1389, double %1375, double noundef 0xBFBFFFFFFFFDACEC)
  %1391 = call double @llvm.fma.f64(double %1390, double %1374, double %1)
  %1392 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %1393 = fmul double %1392, 0x3FE9884533D43651
  %1394 = fmul double %1393, %1382
  %1395 = fmul double %1391, 0x3FE45F306DC9C883
  %1396 = call i32 @llvm.nvvm.d2i.rn(double %1395) #139
  store i32 %1396, i32* %q.i.i.i20.i163, align 4, !noalias !68
  %1397 = sitofp i32 %1396 to double
  %1398 = fneg double %1397
  %1399 = call double @llvm.fma.f64(double %1398, double noundef 0x3FF921FB54442D18, double %1391)
  %1400 = call double @llvm.fma.f64(double %1398, double noundef 0x3C91A62633145C00, double %1399)
  %1401 = call double @llvm.fma.f64(double %1398, double noundef 0x397B839A252049C0, double %1400)
  %1402 = call double @llvm.fabs.f64(double %1391)
  %1403 = fcmp ult double %1402, 0x41E0000000000000
  br i1 %1403, label %__internal_trig_reduction_kerneld.exit.i.i25.i248, label %1404

1404:                                             ; preds = %1369
  %1405 = call fastcc double @__internal_trig_reduction_slowpathd(double %1391, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i20.i163) #139
  %.pre90 = load i32, i32* %q.i.i.i20.i163, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i25.i248

__internal_trig_reduction_kerneld.exit.i.i25.i248: ; preds = %1404, %1369
  %1406 = phi i32 [ %.pre90, %1404 ], [ %1396, %1369 ]
  %t.i.i.i19.i162.0 = phi double [ %1405, %1404 ], [ %1401, %1369 ]
  %1407 = and i32 %1406, 3
  %1408 = uitofp i32 %1407 to double
  %1409 = fadd double %t.i.i.i19.i162.0, 0xBFE921FB54442D18
  %1410 = fmul double %1408, 0x3FF921FB54442D18
  %1411 = fadd double %1409, %1410
  %1412 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1413 = icmp eq i32 %1412, 350
  br i1 %1413, label %1417, label %1414

1414:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i25.i248
  %1415 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1416 = icmp eq i32 %1415, 370
  br i1 %1416, label %1417, label %1476

1417:                                             ; preds = %1414, %__internal_trig_reduction_kerneld.exit.i.i25.i248
  %1418 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1419 = icmp eq i32 %1418, 200
  br i1 %1419, label %1426, label %1420

1420:                                             ; preds = %1417
  %1421 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1422 = icmp eq i32 %1421, 350
  br i1 %1422, label %1426, label %1423

1423:                                             ; preds = %1420
  %1424 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1425 = icmp eq i32 %1424, 370
  br i1 %1425, label %1426, label %1429

1426:                                             ; preds = %1423, %1420, %1417
  %1427 = call double @llvm.fabs.f64(double %1411)
  %1428 = fcmp oeq double %1427, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i.i249

1429:                                             ; preds = %1423
  %1430 = call i32 @llvm.nvvm.d2i.lo(double %1411) #139
  %1431 = call i32 @llvm.nvvm.d2i.hi(double %1411) #139
  %1432 = and i32 %1431, 2147483647
  %1433 = icmp eq i32 %1432, 2146435072
  %1434 = icmp eq i32 %1430, 0
  %1435 = select i1 %1433, i1 %1434, i1 false
  br label %__nv_isinfd.exit.i.i.i.i.i249

__nv_isinfd.exit.i.i.i.i.i249:                    ; preds = %1429, %1426
  %.039.in = phi i1 [ %1428, %1426 ], [ %1435, %1429 ]
  br i1 %.039.in, label %1436, label %1438

1436:                                             ; preds = %__nv_isinfd.exit.i.i.i.i.i249
  %1437 = fmul double %1411, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i250

1438:                                             ; preds = %__nv_isinfd.exit.i.i.i.i.i249
  %1439 = fmul double %1411, 0x3FE45F306DC9C883
  %1440 = call i32 @llvm.nvvm.d2i.rn(double %1439) #139
  store i32 %1440, i32* %q.i.i.i.i.i.i149, align 4, !noalias !68
  %1441 = sitofp i32 %1440 to double
  %1442 = fneg double %1441
  %1443 = call double @llvm.fma.f64(double %1442, double noundef 0x3FF921FB54442D18, double %1411)
  %1444 = call double @llvm.fma.f64(double %1442, double noundef 0x3C91A62633145C00, double %1443)
  %1445 = call double @llvm.fma.f64(double %1442, double noundef 0x397B839A252049C0, double %1444)
  %1446 = call double @llvm.fabs.f64(double %1411)
  %1447 = fcmp ult double %1446, 0x41E0000000000000
  br i1 %1447, label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i250, label %1448

1448:                                             ; preds = %1438
  %1449 = call fastcc double @__internal_trig_reduction_slowpathd(double %1411, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i.i149) #139
  %.pre92 = load i32, i32* %q.i.i.i.i.i.i149, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i250

__internal_trig_reduction_kerneld.exit.i.i.i.i.i250: ; preds = %1438, %1448, %1436
  %i.i.i.i.i.i156.0 = phi i32 [ 0, %1436 ], [ %.pre92, %1448 ], [ %1440, %1438 ]
  %.038 = phi double [ %1437, %1436 ], [ %1449, %1448 ], [ %1445, %1438 ]
  %1450 = fmul double %.038, %.038
  %1451 = call double @llvm.fma.f64(double %1450, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %1452 = call double @llvm.fma.f64(double %1451, double %1450, double noundef 0xBE927E4F8E06E6D9)
  %1453 = call double @llvm.fma.f64(double %1452, double %1450, double noundef 0x3EFA01A019DDBCE9)
  %1454 = call double @llvm.fma.f64(double %1453, double %1450, double noundef 0xBF56C16C16C15D47)
  %1455 = call double @llvm.fma.f64(double %1454, double %1450, double noundef 0x3FA5555555555551)
  %1456 = call double @llvm.fma.f64(double %1455, double %1450, double noundef -5.000000e-01)
  %1457 = call double @llvm.fma.f64(double %1456, double %1450, double noundef 1.000000e+00)
  %1458 = call double @llvm.fma.f64(double %1450, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %1459 = call double @llvm.fma.f64(double %1458, double %1450, double noundef 0x3EC71DE369ACE392)
  %1460 = call double @llvm.fma.f64(double %1459, double %1450, double noundef 0xBF2A01A019DB62A1)
  %1461 = call double @llvm.fma.f64(double %1460, double %1450, double noundef 0x3F81111111110818)
  %1462 = call double @llvm.fma.f64(double %1461, double %1450, double noundef 0xBFC5555555555554)
  %1463 = call double @llvm.fma.f64(double %1462, double %1450, double noundef 0.000000e+00)
  %1464 = call double @llvm.fma.f64(double %1463, double %.038, double %.038)
  %1465 = call i32 @llvm.nvvm.d2i.hi(double %1464) #139
  %1466 = call i32 @llvm.nvvm.d2i.lo(double %1464) #139
  %1467 = xor i32 %1465, -2147483648
  %1468 = call double @llvm.nvvm.lohi.i2d(i32 %1466, i32 %1467) #139
  %1469 = and i32 %i.i.i.i.i.i156.0, 1
  %.not53 = icmp eq i32 %1469, 0
  %spec.select15 = select i1 %.not53, double %1457, double %1468
  %1470 = and i32 %i.i.i.i.i.i156.0, 2
  %.not54 = icmp eq i32 %1470, 0
  br i1 %.not54, label %__internal_cos_offset.exit.i.i255, label %1471

1471:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i250
  %1472 = call i32 @llvm.nvvm.d2i.hi(double %spec.select15) #139
  %1473 = call i32 @llvm.nvvm.d2i.lo(double %spec.select15) #139
  %1474 = xor i32 %1472, -2147483648
  %1475 = call double @llvm.nvvm.lohi.i2d(i32 %1473, i32 %1474) #139
  br label %__internal_cos_offset.exit.i.i255

1476:                                             ; preds = %1414
  %1477 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1478 = icmp eq i32 %1477, 200
  br i1 %1478, label %1485, label %1479

1479:                                             ; preds = %1476
  %1480 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1481 = icmp eq i32 %1480, 350
  br i1 %1481, label %1485, label %1482

1482:                                             ; preds = %1479
  %1483 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1484 = icmp eq i32 %1483, 370
  br i1 %1484, label %1485, label %1488

1485:                                             ; preds = %1482, %1479, %1476
  %1486 = call double @llvm.fabs.f64(double %1411)
  %1487 = fcmp oeq double %1486, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i26.i252

1488:                                             ; preds = %1482
  %1489 = call i32 @llvm.nvvm.d2i.lo(double %1411) #139
  %1490 = call i32 @llvm.nvvm.d2i.hi(double %1411) #139
  %1491 = and i32 %1490, 2147483647
  %1492 = icmp eq i32 %1491, 2146435072
  %1493 = icmp eq i32 %1489, 0
  %1494 = select i1 %1492, i1 %1493, i1 false
  br label %__nv_isinfd.exit.i.i.i26.i252

__nv_isinfd.exit.i.i.i26.i252:                    ; preds = %1488, %1485
  %.040.in = phi i1 [ %1487, %1485 ], [ %1494, %1488 ]
  br i1 %.040.in, label %1495, label %1497

1495:                                             ; preds = %__nv_isinfd.exit.i.i.i26.i252
  %1496 = fmul double %1411, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i254

1497:                                             ; preds = %__nv_isinfd.exit.i.i.i26.i252
  %1498 = fmul double %1411, 0x3FE45F306DC9C883
  %1499 = call i32 @llvm.nvvm.d2i.rn(double %1498) #139
  store i32 %1499, i32* %q.i.i.i.i9.i133, align 4, !noalias !68
  %1500 = sitofp i32 %1499 to double
  %1501 = fneg double %1500
  %1502 = call double @llvm.fma.f64(double %1501, double noundef 0x3FF921FB54442D18, double %1411)
  %1503 = call double @llvm.fma.f64(double %1501, double noundef 0x3C91A62633145C00, double %1502)
  %1504 = call double @llvm.fma.f64(double %1501, double noundef 0x397B839A252049C0, double %1503)
  %1505 = call double @llvm.fabs.f64(double %1411)
  %1506 = fcmp ult double %1505, 0x41E0000000000000
  br i1 %1506, label %__internal_trig_reduction_kerneld.exit.i.i.i27.i253, label %1507

1507:                                             ; preds = %1497
  %1508 = call fastcc double @__internal_trig_reduction_slowpathd(double %1411, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i9.i133) #139
  %.pre91 = load i32, i32* %q.i.i.i.i9.i133, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i27.i253

__internal_trig_reduction_kerneld.exit.i.i.i27.i253: ; preds = %1507, %1497
  %1509 = phi i32 [ %.pre91, %1507 ], [ %1499, %1497 ]
  %t.i1.i.i.i.i132.0 = phi double [ %1508, %1507 ], [ %1504, %1497 ]
  %phi.bo49 = add i32 %1509, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i254

__internal_sin_cos_kerneld.exit.i.i.i.i254:       ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i27.i253, %1495
  %i.i.i.i17.i160.0 = phi i32 [ 1, %1495 ], [ %phi.bo49, %__internal_trig_reduction_kerneld.exit.i.i.i27.i253 ]
  %z.i.i.i16.i159.0 = phi double [ %1496, %1495 ], [ %t.i1.i.i.i.i132.0, %__internal_trig_reduction_kerneld.exit.i.i.i27.i253 ]
  %1510 = shl i32 %i.i.i.i17.i160.0, 3
  %1511 = and i32 %1510, 8
  %1512 = zext i32 %1511 to i64
  %1513 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %1512
  %1514 = fmul double %z.i.i.i16.i159.0, %z.i.i.i16.i159.0
  %1515 = and i32 %i.i.i.i17.i160.0, 1
  %.not50 = icmp eq i32 %1515, 0
  %1516 = select i1 %.not50, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %1517 = getelementptr inbounds double, double addrspace(1)* %1513, i64 1
  %1518 = load double, double addrspace(1)* %1517, align 8
  %1519 = call double @llvm.fma.f64(double noundef %1516, double %1514, double %1518)
  %1520 = getelementptr inbounds double, double addrspace(1)* %1513, i64 2
  %1521 = load double, double addrspace(1)* %1520, align 8
  %1522 = call double @llvm.fma.f64(double %1519, double %1514, double %1521)
  %1523 = getelementptr inbounds double, double addrspace(1)* %1513, i64 3
  %1524 = load double, double addrspace(1)* %1523, align 8
  %1525 = call double @llvm.fma.f64(double %1522, double %1514, double %1524)
  %1526 = getelementptr inbounds double, double addrspace(1)* %1513, i64 4
  %1527 = load double, double addrspace(1)* %1526, align 8
  %1528 = call double @llvm.fma.f64(double %1525, double %1514, double %1527)
  %1529 = getelementptr inbounds double, double addrspace(1)* %1513, i64 5
  %1530 = load double, double addrspace(1)* %1529, align 8
  %1531 = call double @llvm.fma.f64(double %1528, double %1514, double %1530)
  %1532 = getelementptr inbounds double, double addrspace(1)* %1513, i64 6
  %1533 = load double, double addrspace(1)* %1532, align 8
  %1534 = call double @llvm.fma.f64(double %1531, double %1514, double %1533)
  %1535 = call double @llvm.fma.f64(double %1534, double %z.i.i.i16.i159.0, double %z.i.i.i16.i159.0)
  %1536 = call double @llvm.fma.f64(double %1534, double %1514, double noundef 1.000000e+00)
  %spec.select16 = select i1 %.not50, double %1535, double %1536
  %1537 = and i32 %i.i.i.i17.i160.0, 2
  %.not52 = icmp eq i32 %1537, 0
  %1538 = call double @llvm.fma.f64(double %spec.select16, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.142 = select i1 %.not52, double %spec.select16, double %1538
  br label %__internal_cos_offset.exit.i.i255

__internal_cos_offset.exit.i.i255:                ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i250, %1471, %__internal_sin_cos_kerneld.exit.i.i.i.i254
  %.037 = phi double [ %.142, %__internal_sin_cos_kerneld.exit.i.i.i.i254 ], [ %1475, %1471 ], [ %spec.select15, %__internal_trig_reduction_kerneld.exit.i.i.i.i.i250 ]
  %1539 = fmul double %1394, %.037
  br label %__nv_j0.exit.i256

__nv_j0.exit.i256:                                ; preds = %1311, %__internal_cos_offset.exit.i.i255, %__nv_isinfd.exit.i.i247, %1333, %1289
  %r.i.i169.3 = phi double [ %1308, %1289 ], [ %1330, %1311 ], [ %1352, %1333 ], [ %1539, %__internal_cos_offset.exit.i.i255 ], [ 0.000000e+00, %__nv_isinfd.exit.i.i247 ]
  %1540 = fmul double %1287, %r.i.i169.3
  %1541 = fadd double %1222, %1540
  br label %__nv_y0.exit266

1542:                                             ; preds = %1209
  %1543 = fcmp ugt double %1, 0x4000347C4AB37B18
  br i1 %1543, label %1571, label %1544

1544:                                             ; preds = %1542
  %1545 = fadd double %1, 0xBFEC982EB8D417EA
  %1546 = fadd double %1545, 0xBC7EA9D270347F83
  %1547 = call double @llvm.fma.f64(double %1546, double noundef 0xBF01630132D75FC3, double noundef 0x3F3D054B05D3C52D)
  %1548 = call double @llvm.fma.f64(double %1547, double %1546, double noundef 0xBF66DAC0B314B2E5)
  %1549 = call double @llvm.fma.f64(double %1548, double %1546, double noundef 0x3F86A5D1DE76263F)
  %1550 = call double @llvm.fma.f64(double %1549, double %1546, double noundef 0xBF9FD16652824592)
  %1551 = call double @llvm.fma.f64(double %1550, double %1546, double noundef 0x3FB0F69A9CC79FBD)
  %1552 = call double @llvm.fma.f64(double %1551, double %1546, double noundef 0xBFBCCE40EF15583E)
  %1553 = call double @llvm.fma.f64(double %1552, double %1546, double noundef 0x3FC446B11780E4FC)
  %1554 = call double @llvm.fma.f64(double %1553, double %1546, double noundef 0xBFC89AE7E19621F7)
  %1555 = call double @llvm.fma.f64(double %1554, double %1546, double noundef 0x3FCACBA1B38EF7B8)
  %1556 = call double @llvm.fma.f64(double %1555, double %1546, double noundef 0xBFCB4166A03BBFA5)
  %1557 = call double @llvm.fma.f64(double %1556, double %1546, double noundef 0x3FCACCA4D5D4889A)
  %1558 = call double @llvm.fma.f64(double %1557, double %1546, double noundef 0xBFCA1455932B9392)
  %1559 = call double @llvm.fma.f64(double %1558, double %1546, double noundef 0x3FC96D8DB8D844EC)
  %1560 = call double @llvm.fma.f64(double %1559, double %1546, double noundef 0xBFC8F7FB77522EDF)
  %1561 = call double @llvm.fma.f64(double %1560, double %1546, double noundef 0x3FC8C0926ABC9AB0)
  %1562 = call double @llvm.fma.f64(double %1561, double %1546, double noundef 0xBFC8D35B8FEA468C)
  %1563 = call double @llvm.fma.f64(double %1562, double %1546, double noundef 0x3FC9424B8A0C8F94)
  %1564 = call double @llvm.fma.f64(double %1563, double %1546, double noundef 0xBFCA396A7F3403EF)
  %1565 = call double @llvm.fma.f64(double %1564, double %1546, double noundef 0x3FCC068086C37055)
  %1566 = call double @llvm.fma.f64(double %1565, double %1546, double noundef 0xBFCCF18E6A4C5C4E)
  %1567 = call double @llvm.fma.f64(double %1566, double %1546, double noundef 0x3FCC3B1338AF4239)
  %1568 = call double @llvm.fma.f64(double %1567, double %1546, double noundef 0xBFDF7E38A46D70DB)
  %1569 = call double @llvm.fma.f64(double %1568, double %1546, double noundef 0x3FEC24371844B88A)
  %1570 = fmul double %1546, %1569
  br label %__nv_y0.exit266

1571:                                             ; preds = %1542
  %1572 = fcmp ugt double %1, 0x40161663B5D9A628
  br i1 %1572, label %1600, label %1573

1573:                                             ; preds = %1571
  %1574 = fadd double %1, 0xC00FA9534D98569C
  %1575 = fadd double %1574, 0x3C9F06AE7804384E
  %1576 = call double @llvm.fma.f64(double %1575, double noundef 0xBCDAEA62AC8BDA68, double noundef 0xBCD2434958151AC7)
  %1577 = call double @llvm.fma.f64(double %1576, double %1575, double noundef 0x3D11C24A40D33FE1)
  %1578 = call double @llvm.fma.f64(double %1577, double %1575, double noundef 0x3D237CD62FA08CA4)
  %1579 = call double @llvm.fma.f64(double %1578, double %1575, double noundef 0xBD43902E0298C52A)
  %1580 = call double @llvm.fma.f64(double %1579, double %1575, double noundef 0xBD1DDAAD11CAB40F)
  %1581 = call double @llvm.fma.f64(double %1580, double %1575, double noundef 0xBD5209D9F06D7DE4)
  %1582 = call double @llvm.fma.f64(double %1581, double %1575, double noundef 0x3D8BB9F464468E1A)
  %1583 = call double @llvm.fma.f64(double %1582, double %1575, double noundef 0xBDA8F67B07D1B440)
  %1584 = call double @llvm.fma.f64(double %1583, double %1575, double noundef 0x3DC7C8D60F9EAECF)
  %1585 = call double @llvm.fma.f64(double %1584, double %1575, double noundef 0xBDE9703405B49A8D)
  %1586 = call double @llvm.fma.f64(double %1585, double %1575, double noundef 0x3E0A6B64E76417E4)
  %1587 = call double @llvm.fma.f64(double %1586, double %1575, double noundef 0xBE2F6B5AFB2F1359)
  %1588 = call double @llvm.fma.f64(double %1587, double %1575, double noundef 0x3E54526B71C21EC1)
  %1589 = call double @llvm.fma.f64(double %1588, double %1575, double noundef 0xBE5776DBCBBC8E1D)
  %1590 = call double @llvm.fma.f64(double %1589, double %1575, double noundef 0xBE93B211FC2DF90E)
  %1591 = call double @llvm.fma.f64(double %1590, double %1575, double noundef 0xBED486372E8562DC)
  %1592 = call double @llvm.fma.f64(double %1591, double %1575, double noundef 0x3F0AB2C1FBC3A254)
  %1593 = call double @llvm.fma.f64(double %1592, double %1575, double noundef 0x3F299827653353B8)
  %1594 = call double @llvm.fma.f64(double %1593, double %1575, double noundef 0xBF61E32BC4ED7084)
  %1595 = call double @llvm.fma.f64(double %1594, double %1575, double noundef 0xBF7C116FDC599A09)
  %1596 = call double @llvm.fma.f64(double %1595, double %1575, double noundef 0x3FADF6D59BF50C77)
  %1597 = call double @llvm.fma.f64(double %1596, double %1575, double noundef 0x3FAA09C92903680B)
  %1598 = call double @llvm.fma.f64(double %1597, double %1575, double noundef 0xBFD9C34256A12A0B)
  %1599 = fmul double %1575, %1598
  br label %__nv_y0.exit266

1600:                                             ; preds = %1571
  %1601 = fcmp ugt double %1, 0x40214EF30C0C06ED
  br i1 %1601, label %1621, label %1602

1602:                                             ; preds = %1600
  %1603 = fadd double %1, 0xC01C581DC4E72103
  %1604 = fadd double %1603, 0x3C99774A495F56CF
  %1605 = call double @llvm.fma.f64(double %1604, double noundef 0x3CF1CB3ABA718B8E, double noundef 0xBD3F443BB4F53D75)
  %1606 = call double @llvm.fma.f64(double %1605, double %1604, double noundef 0xBD770F737BD6A786)
  %1607 = call double @llvm.fma.f64(double %1606, double %1604, double noundef 0x3DBF0E9A20459E14)
  %1608 = call double @llvm.fma.f64(double %1607, double %1604, double noundef 0x3DEFA6B137D5E108)
  %1609 = call double @llvm.fma.f64(double %1608, double %1604, double noundef 0xBE344296729FB7FA)
  %1610 = call double @llvm.fma.f64(double %1609, double %1604, double noundef 0xBE60A2813A80DFAA)
  %1611 = call double @llvm.fma.f64(double %1610, double %1604, double noundef 0x3EA34AA737A83EB4)
  %1612 = call double @llvm.fma.f64(double %1611, double %1604, double noundef 0x3EC6A9227332D03C)
  %1613 = call double @llvm.fma.f64(double %1612, double %1604, double noundef 0xBF08177E4F93C81E)
  %1614 = call double @llvm.fma.f64(double %1613, double %1604, double noundef 0xBF226DD71E391775)
  %1615 = call double @llvm.fma.f64(double %1614, double %1604, double noundef 0x3F61D35E85FD7B22)
  %1616 = call double @llvm.fma.f64(double %1615, double %1604, double noundef 0x3F6B2F14A955285C)
  %1617 = call double @llvm.fma.f64(double %1616, double %1604, double noundef 0xBFA8969C64CBF388)
  %1618 = call double @llvm.fma.f64(double %1617, double %1604, double noundef 0xBF95AEF611FC4D5A)
  %1619 = call double @llvm.fma.f64(double %1618, double %1604, double noundef 0x3FD334CCA0697A5A)
  %1620 = fmul double %1604, %1619
  br label %__nv_y0.exit266

1621:                                             ; preds = %1600
  %1622 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1623 = icmp eq i32 %1622, 200
  br i1 %1623, label %1630, label %1624

1624:                                             ; preds = %1621
  %1625 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1626 = icmp eq i32 %1625, 350
  br i1 %1626, label %1630, label %1627

1627:                                             ; preds = %1624
  %1628 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1629 = icmp eq i32 %1628, 370
  br i1 %1629, label %1630, label %1632

1630:                                             ; preds = %1627, %1624, %1621
  %1631 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i257

1632:                                             ; preds = %1627
  %1633 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %1634 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %1635 = and i32 %1634, 2147483647
  %1636 = icmp eq i32 %1635, 2146435072
  %1637 = icmp eq i32 %1633, 0
  %1638 = select i1 %1636, i1 %1637, i1 false
  br label %__nv_isinfd.exit.i257

__nv_isinfd.exit.i257:                            ; preds = %1632, %1630
  %.044.in = phi i1 [ %1631, %1630 ], [ %1638, %1632 ]
  br i1 %.044.in, label %__nv_y0.exit266, label %1639

1639:                                             ; preds = %__nv_isinfd.exit.i257
  %1640 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %1641 = fneg double %1
  %1642 = call double @llvm.fma.f64(double %1641, double %1640, double noundef 1.000000e+00)
  %1643 = call double @llvm.fma.f64(double %1642, double %1642, double %1642)
  %1644 = call double @llvm.fma.f64(double %1643, double %1640, double %1640)
  %1645 = fmul double %1644, %1644
  %1646 = call double @llvm.fma.f64(double %1645, double noundef 0xC0C5E91E6AC3AD03, double noundef 0x4093F56A049CDDE7)
  %1647 = call double @llvm.fma.f64(double %1646, double %1645, double noundef 0xC05572D39DFB8433)
  %1648 = call double @llvm.fma.f64(double %1647, double %1645, double noundef 0x4016A6041CAA59E5)
  %1649 = call double @llvm.fma.f64(double %1648, double %1645, double noundef 0xBFE155E3A0493880)
  %1650 = call double @llvm.fma.f64(double %1649, double %1645, double noundef 0x3FBA7FB92F417F7F)
  %1651 = call double @llvm.fma.f64(double %1650, double %1645, double noundef 0xBFAFFFFFB12E32F5)
  %1652 = call double @llvm.fma.f64(double %1651, double %1645, double noundef 0x3FEFFFFFFFFECED5)
  %1653 = call double @llvm.fma.f64(double %1645, double noundef 0x418A86A64BE101DC, double noundef 0xC15709C79AAC5813)
  %1654 = call double @llvm.fma.f64(double %1653, double %1645, double noundef 0x41142A31C980A287)
  %1655 = call double @llvm.fma.f64(double %1654, double %1645, double noundef 0xC0C9CBE68930485D)
  %1656 = call double @llvm.fma.f64(double %1655, double %1645, double noundef 0x407F583E14E8A4E8)
  %1657 = call double @llvm.fma.f64(double %1656, double %1645, double noundef 0xC0374A629C650680)
  %1658 = call double @llvm.fma.f64(double %1657, double %1645, double noundef 0x3FFA32A7AF17FAE9)
  %1659 = call double @llvm.fma.f64(double %1658, double %1645, double noundef 0xBFCAD32497785CD6)
  %1660 = call double @llvm.fma.f64(double %1659, double %1645, double noundef 0x3FB0AAAA9FB75F7B)
  %1661 = call double @llvm.fma.f64(double %1660, double %1645, double noundef 0xBFBFFFFFFFFE320F)
  %1662 = call double @llvm.fma.f64(double %1661, double %1644, double %1)
  %1663 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %1664 = fmul double %1663, 0x3FE9884533D43651
  %1665 = fmul double %1664, %1652
  %1666 = fmul double %1662, 0x3FE45F306DC9C883
  %1667 = call i32 @llvm.nvvm.d2i.rn(double %1666) #139
  store i32 %1667, i32* %q.i.i.i232, align 4, !noalias !68
  %1668 = sitofp i32 %1667 to double
  %1669 = fneg double %1668
  %1670 = call double @llvm.fma.f64(double %1669, double noundef 0x3FF921FB54442D18, double %1662)
  %1671 = call double @llvm.fma.f64(double %1669, double noundef 0x3C91A62633145C00, double %1670)
  %1672 = call double @llvm.fma.f64(double %1669, double noundef 0x397B839A252049C0, double %1671)
  %1673 = call double @llvm.fabs.f64(double %1662)
  %1674 = fcmp ult double %1673, 0x41E0000000000000
  br i1 %1674, label %__internal_trig_reduction_kerneld.exit.i.i258, label %1675

1675:                                             ; preds = %1639
  %1676 = call fastcc double @__internal_trig_reduction_slowpathd(double %1662, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i232) #139
  %.pre93 = load i32, i32* %q.i.i.i232, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i258

__internal_trig_reduction_kerneld.exit.i.i258:    ; preds = %1675, %1639
  %1677 = phi i32 [ %.pre93, %1675 ], [ %1667, %1639 ]
  %t.i.i.i231.0 = phi double [ %1676, %1675 ], [ %1672, %1639 ]
  %1678 = and i32 %1677, 3
  %1679 = uitofp i32 %1678 to double
  %1680 = fadd double %t.i.i.i231.0, 0xC002D97C7F3321D2
  %1681 = fmul double %1679, 0x3FF921FB54442D18
  %1682 = fadd double %1680, %1681
  %1683 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1684 = icmp eq i32 %1683, 350
  br i1 %1684, label %1688, label %1685

1685:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i258
  %1686 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1687 = icmp eq i32 %1686, 370
  br i1 %1687, label %1688, label %1747

1688:                                             ; preds = %1685, %__internal_trig_reduction_kerneld.exit.i.i258
  %1689 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1690 = icmp eq i32 %1689, 200
  br i1 %1690, label %1697, label %1691

1691:                                             ; preds = %1688
  %1692 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1693 = icmp eq i32 %1692, 350
  br i1 %1693, label %1697, label %1694

1694:                                             ; preds = %1691
  %1695 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1696 = icmp eq i32 %1695, 370
  br i1 %1696, label %1697, label %1700

1697:                                             ; preds = %1694, %1691, %1688
  %1698 = call double @llvm.fabs.f64(double %1682)
  %1699 = fcmp oeq double %1698, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i259

1700:                                             ; preds = %1694
  %1701 = call i32 @llvm.nvvm.d2i.lo(double %1682) #139
  %1702 = call i32 @llvm.nvvm.d2i.hi(double %1682) #139
  %1703 = and i32 %1702, 2147483647
  %1704 = icmp eq i32 %1703, 2146435072
  %1705 = icmp eq i32 %1701, 0
  %1706 = select i1 %1704, i1 %1705, i1 false
  br label %__nv_isinfd.exit.i.i.i.i259

__nv_isinfd.exit.i.i.i.i259:                      ; preds = %1700, %1697
  %.032.in = phi i1 [ %1699, %1697 ], [ %1706, %1700 ]
  br i1 %.032.in, label %1707, label %1709

1707:                                             ; preds = %__nv_isinfd.exit.i.i.i.i259
  %1708 = fmul double %1682, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i260

1709:                                             ; preds = %__nv_isinfd.exit.i.i.i.i259
  %1710 = fmul double %1682, 0x3FE45F306DC9C883
  %1711 = call i32 @llvm.nvvm.d2i.rn(double %1710) #139
  store i32 %1711, i32* %q.i.i.i.i.i218, align 4, !noalias !68
  %1712 = sitofp i32 %1711 to double
  %1713 = fneg double %1712
  %1714 = call double @llvm.fma.f64(double %1713, double noundef 0x3FF921FB54442D18, double %1682)
  %1715 = call double @llvm.fma.f64(double %1713, double noundef 0x3C91A62633145C00, double %1714)
  %1716 = call double @llvm.fma.f64(double %1713, double noundef 0x397B839A252049C0, double %1715)
  %1717 = call double @llvm.fabs.f64(double %1682)
  %1718 = fcmp ult double %1717, 0x41E0000000000000
  br i1 %1718, label %__internal_trig_reduction_kerneld.exit.i.i.i.i260, label %1719

1719:                                             ; preds = %1709
  %1720 = call fastcc double @__internal_trig_reduction_slowpathd(double %1682, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i218) #139
  %.pre95 = load i32, i32* %q.i.i.i.i.i218, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i260

__internal_trig_reduction_kerneld.exit.i.i.i.i260: ; preds = %1709, %1719, %1707
  %i.i.i.i.i225.0 = phi i32 [ 0, %1707 ], [ %.pre95, %1719 ], [ %1711, %1709 ]
  %.031 = phi double [ %1708, %1707 ], [ %1720, %1719 ], [ %1716, %1709 ]
  %1721 = fmul double %.031, %.031
  %1722 = call double @llvm.fma.f64(double %1721, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %1723 = call double @llvm.fma.f64(double %1722, double %1721, double noundef 0xBE927E4F8E06E6D9)
  %1724 = call double @llvm.fma.f64(double %1723, double %1721, double noundef 0x3EFA01A019DDBCE9)
  %1725 = call double @llvm.fma.f64(double %1724, double %1721, double noundef 0xBF56C16C16C15D47)
  %1726 = call double @llvm.fma.f64(double %1725, double %1721, double noundef 0x3FA5555555555551)
  %1727 = call double @llvm.fma.f64(double %1726, double %1721, double noundef -5.000000e-01)
  %1728 = call double @llvm.fma.f64(double %1727, double %1721, double noundef 1.000000e+00)
  %1729 = call double @llvm.fma.f64(double %1721, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %1730 = call double @llvm.fma.f64(double %1729, double %1721, double noundef 0x3EC71DE369ACE392)
  %1731 = call double @llvm.fma.f64(double %1730, double %1721, double noundef 0xBF2A01A019DB62A1)
  %1732 = call double @llvm.fma.f64(double %1731, double %1721, double noundef 0x3F81111111110818)
  %1733 = call double @llvm.fma.f64(double %1732, double %1721, double noundef 0xBFC5555555555554)
  %1734 = call double @llvm.fma.f64(double %1733, double %1721, double noundef 0.000000e+00)
  %1735 = call double @llvm.fma.f64(double %1734, double %.031, double %.031)
  %1736 = call i32 @llvm.nvvm.d2i.hi(double %1735) #139
  %1737 = call i32 @llvm.nvvm.d2i.lo(double %1735) #139
  %1738 = xor i32 %1736, -2147483648
  %1739 = call double @llvm.nvvm.lohi.i2d(i32 %1737, i32 %1738) #139
  %1740 = and i32 %i.i.i.i.i225.0, 1
  %.not47 = icmp eq i32 %1740, 0
  %spec.select18 = select i1 %.not47, double %1728, double %1739
  %1741 = and i32 %i.i.i.i.i225.0, 2
  %.not48 = icmp eq i32 %1741, 0
  br i1 %.not48, label %__internal_cos_offset.exit.i265, label %1742

1742:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i260
  %1743 = call i32 @llvm.nvvm.d2i.hi(double %spec.select18) #139
  %1744 = call i32 @llvm.nvvm.d2i.lo(double %spec.select18) #139
  %1745 = xor i32 %1743, -2147483648
  %1746 = call double @llvm.nvvm.lohi.i2d(i32 %1744, i32 %1745) #139
  br label %__internal_cos_offset.exit.i265

1747:                                             ; preds = %1685
  %1748 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1749 = icmp eq i32 %1748, 200
  br i1 %1749, label %1756, label %1750

1750:                                             ; preds = %1747
  %1751 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1752 = icmp eq i32 %1751, 350
  br i1 %1752, label %1756, label %1753

1753:                                             ; preds = %1750
  %1754 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1755 = icmp eq i32 %1754, 370
  br i1 %1755, label %1756, label %1759

1756:                                             ; preds = %1753, %1750, %1747
  %1757 = call double @llvm.fabs.f64(double %1682)
  %1758 = fcmp oeq double %1757, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i262

1759:                                             ; preds = %1753
  %1760 = call i32 @llvm.nvvm.d2i.lo(double %1682) #139
  %1761 = call i32 @llvm.nvvm.d2i.hi(double %1682) #139
  %1762 = and i32 %1761, 2147483647
  %1763 = icmp eq i32 %1762, 2146435072
  %1764 = icmp eq i32 %1760, 0
  %1765 = select i1 %1763, i1 %1764, i1 false
  br label %__nv_isinfd.exit.i.i.i262

__nv_isinfd.exit.i.i.i262:                        ; preds = %1759, %1756
  %.033.in = phi i1 [ %1758, %1756 ], [ %1765, %1759 ]
  br i1 %.033.in, label %1766, label %1768

1766:                                             ; preds = %__nv_isinfd.exit.i.i.i262
  %1767 = fmul double %1682, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i264

1768:                                             ; preds = %__nv_isinfd.exit.i.i.i262
  %1769 = fmul double %1682, 0x3FE45F306DC9C883
  %1770 = call i32 @llvm.nvvm.d2i.rn(double %1769) #139
  store i32 %1770, i32* %q.i.i.i.i202, align 4, !noalias !68
  %1771 = sitofp i32 %1770 to double
  %1772 = fneg double %1771
  %1773 = call double @llvm.fma.f64(double %1772, double noundef 0x3FF921FB54442D18, double %1682)
  %1774 = call double @llvm.fma.f64(double %1772, double noundef 0x3C91A62633145C00, double %1773)
  %1775 = call double @llvm.fma.f64(double %1772, double noundef 0x397B839A252049C0, double %1774)
  %1776 = call double @llvm.fabs.f64(double %1682)
  %1777 = fcmp ult double %1776, 0x41E0000000000000
  br i1 %1777, label %__internal_trig_reduction_kerneld.exit.i.i.i263, label %1778

1778:                                             ; preds = %1768
  %1779 = call fastcc double @__internal_trig_reduction_slowpathd(double %1682, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i202) #139
  %.pre94 = load i32, i32* %q.i.i.i.i202, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i263

__internal_trig_reduction_kerneld.exit.i.i.i263:  ; preds = %1778, %1768
  %1780 = phi i32 [ %.pre94, %1778 ], [ %1770, %1768 ]
  %t.i1.i.i.i201.0 = phi double [ %1779, %1778 ], [ %1775, %1768 ]
  %phi.bo43 = add i32 %1780, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i264

__internal_sin_cos_kerneld.exit.i.i.i264:         ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i263, %1766
  %i.i.i.i229.0 = phi i32 [ 1, %1766 ], [ %phi.bo43, %__internal_trig_reduction_kerneld.exit.i.i.i263 ]
  %z.i.i.i228.0 = phi double [ %1767, %1766 ], [ %t.i1.i.i.i201.0, %__internal_trig_reduction_kerneld.exit.i.i.i263 ]
  %1781 = shl i32 %i.i.i.i229.0, 3
  %1782 = and i32 %1781, 8
  %1783 = zext i32 %1782 to i64
  %1784 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %1783
  %1785 = fmul double %z.i.i.i228.0, %z.i.i.i228.0
  %1786 = and i32 %i.i.i.i229.0, 1
  %.not44 = icmp eq i32 %1786, 0
  %1787 = select i1 %.not44, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %1788 = getelementptr inbounds double, double addrspace(1)* %1784, i64 1
  %1789 = load double, double addrspace(1)* %1788, align 8
  %1790 = call double @llvm.fma.f64(double noundef %1787, double %1785, double %1789)
  %1791 = getelementptr inbounds double, double addrspace(1)* %1784, i64 2
  %1792 = load double, double addrspace(1)* %1791, align 8
  %1793 = call double @llvm.fma.f64(double %1790, double %1785, double %1792)
  %1794 = getelementptr inbounds double, double addrspace(1)* %1784, i64 3
  %1795 = load double, double addrspace(1)* %1794, align 8
  %1796 = call double @llvm.fma.f64(double %1793, double %1785, double %1795)
  %1797 = getelementptr inbounds double, double addrspace(1)* %1784, i64 4
  %1798 = load double, double addrspace(1)* %1797, align 8
  %1799 = call double @llvm.fma.f64(double %1796, double %1785, double %1798)
  %1800 = getelementptr inbounds double, double addrspace(1)* %1784, i64 5
  %1801 = load double, double addrspace(1)* %1800, align 8
  %1802 = call double @llvm.fma.f64(double %1799, double %1785, double %1801)
  %1803 = getelementptr inbounds double, double addrspace(1)* %1784, i64 6
  %1804 = load double, double addrspace(1)* %1803, align 8
  %1805 = call double @llvm.fma.f64(double %1802, double %1785, double %1804)
  %1806 = call double @llvm.fma.f64(double %1805, double %z.i.i.i228.0, double %z.i.i.i228.0)
  %1807 = call double @llvm.fma.f64(double %1805, double %1785, double noundef 1.000000e+00)
  %spec.select19 = select i1 %.not44, double %1806, double %1807
  %1808 = and i32 %i.i.i.i229.0, 2
  %.not46 = icmp eq i32 %1808, 0
  %1809 = call double @llvm.fma.f64(double %spec.select19, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.135 = select i1 %.not46, double %spec.select19, double %1809
  br label %__internal_cos_offset.exit.i265

__internal_cos_offset.exit.i265:                  ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i260, %1742, %__internal_sin_cos_kerneld.exit.i.i.i264
  %.030 = phi double [ %.135, %__internal_sin_cos_kerneld.exit.i.i.i264 ], [ %1746, %1742 ], [ %spec.select18, %__internal_trig_reduction_kerneld.exit.i.i.i.i260 ]
  %1810 = fmul double %1665, %.030
  br label %__nv_y0.exit266

__nv_y0.exit266:                                  ; preds = %1544, %1602, %__nv_isinfd.exit.i257, %__internal_cos_offset.exit.i265, %1573, %__nv_j0.exit.i256
  %r.i239.4 = phi double [ %1541, %__nv_j0.exit.i256 ], [ %1570, %1544 ], [ %1599, %1573 ], [ %1620, %1602 ], [ %1810, %__internal_cos_offset.exit.i265 ], [ 0.000000e+00, %__nv_isinfd.exit.i257 ]
  %r.i239.5 = select i1 %1208, double 0xFFF8000000000000, double %r.i239.4
  br label %__nv_y1.exit416

1811:                                             ; preds = %__nv_y0.exit
  br i1 %2, label %1812, label %1814

1812:                                             ; preds = %1811
  %1813 = fdiv double 0xBFE45F306DC9C883, %1
  br label %2413

1814:                                             ; preds = %1811
  %1815 = fcmp ugt double %1, 0x3FF4C6F208132576
  br i1 %1815, label %2147, label %1816

1816:                                             ; preds = %1814
  %1817 = fmul double %a, %a
  %1818 = call double @llvm.fma.f64(double %1817, double noundef 0x3D5249F90687428C, double noundef 0xBDCF0B5B1FB7B95E)
  %1819 = call double @llvm.fma.f64(double %1818, double %1817, double noundef 0x3E432E589311FA14)
  %1820 = call double @llvm.fma.f64(double %1819, double %1817, double noundef 0xBEB0A780AA4A92E9)
  %1821 = call double @llvm.fma.f64(double %1820, double %1817, double noundef 0x3F12C7DBFFCAEC2B)
  %1822 = call double @llvm.fma.f64(double %1821, double %1817, double noundef 0xBF6835B97894BA4A)
  %1823 = call double @llvm.fma.f64(double %1822, double %1817, double noundef 0x3FABD3975C75B4A3)
  %1824 = call double @llvm.fma.f64(double %1823, double %1817, double noundef 0xBFC91866143CBC8A)
  %1825 = fmul double %1, %1824
  %1826 = fcmp ugt double %1, 0x400353AABAD7B784
  br i1 %1826, label %1844, label %1827

1827:                                             ; preds = %1816
  %1828 = call double @llvm.fma.f64(double %1, double noundef 0x3D020E4ADCDE2AD3, double noundef 0xBD4DD167A0DC3F55)
  %1829 = call double @llvm.fma.f64(double %1828, double %1, double noundef 0x3D5503F5A491E487)
  %1830 = call double @llvm.fma.f64(double %1829, double %1, double noundef 0x3DC1F29940C2403A)
  %1831 = call double @llvm.fma.f64(double %1830, double %1, double noundef 0x3D84CF9302EACDEF)
  %1832 = call double @llvm.fma.f64(double %1831, double %1, double noundef 0xBE384A53DBBCA436)
  %1833 = call double @llvm.fma.f64(double %1832, double %1, double noundef 0x3D9779BEE4F63BCC)
  %1834 = call double @llvm.fma.f64(double %1833, double %1, double noundef 0x3EA6C160E414F3F0)
  %1835 = call double @llvm.fma.f64(double %1834, double %1, double noundef 0x3D8F3D2F12430699)
  %1836 = call double @llvm.fma.f64(double %1835, double %1, double noundef 0xBF0C71C72C0CED04)
  %1837 = call double @llvm.fma.f64(double %1836, double %1, double noundef 0x3D659BCA506F1128)
  %1838 = call double @llvm.fma.f64(double %1837, double %1, double noundef 0x3F65555555506982)
  %1839 = call double @llvm.fma.f64(double %1838, double %1, double noundef 0x3D15BA0B425F1BFB)
  %1840 = call double @llvm.fma.f64(double %1839, double %1, double noundef 0xBFB0000000000065)
  %1841 = call double @llvm.fma.f64(double %1840, double %1, double noundef 0x3C8729A7253FB679)
  %1842 = call double @llvm.fma.f64(double %1841, double %1, double noundef 5.000000e-01)
  %1843 = fmul double %1, %1842
  br label %__nv_j1.exit.i405

1844:                                             ; preds = %1816
  %1845 = fcmp ugt double %1, 0x4015B1D0574614EA
  br i1 %1845, label %1865, label %1846

1846:                                             ; preds = %1844
  %1847 = fadd double %1, 0xC00EA75575AF6F09
  %1848 = fadd double %1847, 0x3CA60155A9D1B256
  %1849 = call double @llvm.fma.f64(double %1848, double noundef 0xBCF8D3CDBB60175E, double noundef 0x3D41011A1DF02DAD)
  %1850 = call double @llvm.fma.f64(double %1849, double %1848, double noundef 0x3D76013AC1E5E222)
  %1851 = call double @llvm.fma.f64(double %1850, double %1848, double noundef 0xBDBEC315D96D5F03)
  %1852 = call double @llvm.fma.f64(double %1851, double %1848, double noundef 0xBDF03BE1B4B57207)
  %1853 = call double @llvm.fma.f64(double %1852, double %1848, double noundef 0x3E345695F8B660F7)
  %1854 = call double @llvm.fma.f64(double %1853, double %1848, double noundef 0x3E617069FCFCFFF4)
  %1855 = call double @llvm.fma.f64(double %1854, double %1848, double noundef 0xBEA33825C36745EB)
  %1856 = call double @llvm.fma.f64(double %1855, double %1848, double noundef 0xBEC9799D4F90931B)
  %1857 = call double @llvm.fma.f64(double %1856, double %1848, double noundef 0x3F083A06E2F7DF13)
  %1858 = call double @llvm.fma.f64(double %1857, double %1848, double noundef 0x3F26E4C2D53A7CF6)
  %1859 = call double @llvm.fma.f64(double %1858, double %1848, double noundef 0xBF624B3409957B1C)
  %1860 = call double @llvm.fma.f64(double %1859, double %1848, double noundef 0xBF7537544C3325DF)
  %1861 = call double @llvm.fma.f64(double %1860, double %1848, double noundef 0x3FAB589D1DA138E2)
  %1862 = call double @llvm.fma.f64(double %1861, double %1848, double noundef 0x3FAAE8A39F51AD13)
  %1863 = call double @llvm.fma.f64(double %1862, double %1848, double noundef 0xBFD9C6CF582CBF7F)
  %1864 = fmul double %1848, %1863
  br label %__nv_j1.exit.i405

1865:                                             ; preds = %1844
  %1866 = fcmp ugt double %1, 0x40213065E54C1AA9
  br i1 %1866, label %1886, label %1867

1867:                                             ; preds = %1865
  %1868 = fadd double %1, 0xC01C0FF5F3B47250
  %1869 = fadd double %1868, 0x3C9B226D9D243827
  %1870 = call double @llvm.fma.f64(double %1869, double noundef 0x3CF3EB867515FAD6, double noundef 0xBD40E8363DB649A9)
  %1871 = call double @llvm.fma.f64(double %1870, double %1869, double noundef 0xBD73B7DD4A6608FB)
  %1872 = call double @llvm.fma.f64(double %1871, double %1869, double noundef 0x3DBEC5E01482C750)
  %1873 = call double @llvm.fma.f64(double %1872, double %1869, double noundef 0x3DEC62BB9E882103)
  %1874 = call double @llvm.fma.f64(double %1873, double %1869, double noundef 0xBE34462EED732A23)
  %1875 = call double @llvm.fma.f64(double %1874, double %1869, double noundef 0xBE5D48DCAD7DC59B)
  %1876 = call double @llvm.fma.f64(double %1875, double %1869, double noundef 0x3EA3026DF29167E9)
  %1877 = call double @llvm.fma.f64(double %1876, double %1869, double noundef 0x3EC4255B0119666C)
  %1878 = call double @llvm.fma.f64(double %1877, double %1869, double noundef 0xBF0796A751B32693)
  %1879 = call double @llvm.fma.f64(double %1878, double %1869, double noundef 0xBF207358BBDBA284)
  %1880 = call double @llvm.fma.f64(double %1879, double %1869, double noundef 0x3F613FBC7D6927B1)
  %1881 = call double @llvm.fma.f64(double %1880, double %1869, double noundef 0x3F69A4B292E3DD75)
  %1882 = call double @llvm.fma.f64(double %1881, double %1869, double noundef 0xBFA80C83BDEEE4FB)
  %1883 = call double @llvm.fma.f64(double %1882, double %1869, double noundef 0xBF95E70DC60362BF)
  %1884 = call double @llvm.fma.f64(double %1883, double %1869, double noundef 0x3FD33518B3874E8A)
  %1885 = fmul double %1869, %1884
  br label %__nv_j1.exit.i405

1886:                                             ; preds = %1865
  %1887 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1888 = icmp eq i32 %1887, 200
  br i1 %1888, label %1895, label %1889

1889:                                             ; preds = %1886
  %1890 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1891 = icmp eq i32 %1890, 350
  br i1 %1891, label %1895, label %1892

1892:                                             ; preds = %1889
  %1893 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1894 = icmp eq i32 %1893, 370
  br i1 %1894, label %1895, label %1897

1895:                                             ; preds = %1892, %1889, %1886
  %1896 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i396

1897:                                             ; preds = %1892
  %1898 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %1899 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %1900 = and i32 %1899, 2147483647
  %1901 = icmp eq i32 %1900, 2146435072
  %1902 = icmp eq i32 %1898, 0
  %1903 = select i1 %1901, i1 %1902, i1 false
  br label %__nv_isinfd.exit.i.i396

__nv_isinfd.exit.i.i396:                          ; preds = %1897, %1895
  %.058.in = phi i1 [ %1896, %1895 ], [ %1903, %1897 ]
  br i1 %.058.in, label %__nv_j1.exit.i405, label %1904

1904:                                             ; preds = %__nv_isinfd.exit.i.i396
  %1905 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %1906 = fneg double %1
  %1907 = call double @llvm.fma.f64(double %1906, double %1905, double noundef 1.000000e+00)
  %1908 = call double @llvm.fma.f64(double %1907, double %1907, double %1907)
  %1909 = call double @llvm.fma.f64(double %1908, double %1905, double %1905)
  %1910 = fmul double %1909, %1909
  %1911 = call double @llvm.fma.f64(double %1910, double noundef 0x40CD02EA3F2F6751, double noundef 0xC099C06322A3F8BE)
  %1912 = call double @llvm.fma.f64(double %1911, double %1910, double noundef 0x405B89354DA77324)
  %1913 = call double @llvm.fma.f64(double %1912, double %1910, double noundef 0xC01E352294653188)
  %1914 = call double @llvm.fma.f64(double %1913, double %1910, double noundef 0x3FE9BC7DB16BD7A7)
  %1915 = call double @llvm.fma.f64(double %1914, double %1910, double noundef 0xBFC8BFE1C3A4F741)
  %1916 = call double @llvm.fma.f64(double %1915, double %1910, double noundef 0x3FC7FFFFF0D00BE2)
  %1917 = call double @llvm.fma.f64(double %1916, double %1910, double noundef 0x3FF00000000068CC)
  %1918 = call double @llvm.fma.f64(double %1910, double noundef 0xC18DA26B212FDC9A, double noundef 0x415A30AC6857BEE0)
  %1919 = call double @llvm.fma.f64(double %1918, double %1910, double noundef 0xC11764222AD7C910)
  %1920 = call double @llvm.fma.f64(double %1919, double %1910, double noundef 0x40CEB02E0C306857)
  %1921 = call double @llvm.fma.f64(double %1920, double %1910, double noundef 0xC08351859FA2B23B)
  %1922 = call double @llvm.fma.f64(double %1921, double %1910, double noundef 0x403E65A07AF51F42)
  %1923 = call double @llvm.fma.f64(double %1922, double %1910, double noundef 0xC002F2B817F77A57)
  %1924 = call double @llvm.fma.f64(double %1923, double %1910, double noundef 0x3FD7BCC34DA069FD)
  %1925 = call double @llvm.fma.f64(double %1924, double %1910, double noundef 0xBFC4FFFFF8A44463)
  %1926 = call double @llvm.fma.f64(double %1925, double %1910, double noundef 0x3FD7FFFFFFFF5CD7)
  %1927 = call double @llvm.fma.f64(double %1926, double %1909, double %1)
  %1928 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %1929 = fmul double %1928, 0x3FE9884533D43651
  %1930 = fmul double %1929, %1917
  %1931 = fmul double %1927, 0x3FE45F306DC9C883
  %1932 = call i32 @llvm.nvvm.d2i.rn(double %1931) #139
  store i32 %1932, i32* %q.i.i.i16.i334, align 4, !noalias !68
  %1933 = sitofp i32 %1932 to double
  %1934 = fneg double %1933
  %1935 = call double @llvm.fma.f64(double %1934, double noundef 0x3FF921FB54442D18, double %1927)
  %1936 = call double @llvm.fma.f64(double %1934, double noundef 0x3C91A62633145C00, double %1935)
  %1937 = call double @llvm.fma.f64(double %1934, double noundef 0x397B839A252049C0, double %1936)
  %1938 = call double @llvm.fabs.f64(double %1927)
  %1939 = fcmp ult double %1938, 0x41E0000000000000
  br i1 %1939, label %__internal_trig_reduction_kerneld.exit.i.i20.i397, label %1940

1940:                                             ; preds = %1904
  %1941 = call fastcc double @__internal_trig_reduction_slowpathd(double %1927, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i16.i334) #139
  %.pre84 = load i32, i32* %q.i.i.i16.i334, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i20.i397

__internal_trig_reduction_kerneld.exit.i.i20.i397: ; preds = %1940, %1904
  %1942 = phi i32 [ %.pre84, %1940 ], [ %1932, %1904 ]
  %t.i.i.i15.i333.0 = phi double [ %1941, %1940 ], [ %1937, %1904 ]
  %1943 = and i32 %1942, 3
  %1944 = uitofp i32 %1943 to double
  %1945 = fadd double %t.i.i.i15.i333.0, 0xC002D97C7F3321D2
  %1946 = fmul double %1944, 0x3FF921FB54442D18
  %1947 = fadd double %1945, %1946
  %1948 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1949 = icmp eq i32 %1948, 350
  br i1 %1949, label %1953, label %1950

1950:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i20.i397
  %1951 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1952 = icmp eq i32 %1951, 370
  br i1 %1952, label %1953, label %2012

1953:                                             ; preds = %1950, %__internal_trig_reduction_kerneld.exit.i.i20.i397
  %1954 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1955 = icmp eq i32 %1954, 200
  br i1 %1955, label %1962, label %1956

1956:                                             ; preds = %1953
  %1957 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1958 = icmp eq i32 %1957, 350
  br i1 %1958, label %1962, label %1959

1959:                                             ; preds = %1956
  %1960 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %1961 = icmp eq i32 %1960, 370
  br i1 %1961, label %1962, label %1965

1962:                                             ; preds = %1959, %1956, %1953
  %1963 = call double @llvm.fabs.f64(double %1947)
  %1964 = fcmp oeq double %1963, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i.i398

1965:                                             ; preds = %1959
  %1966 = call i32 @llvm.nvvm.d2i.lo(double %1947) #139
  %1967 = call i32 @llvm.nvvm.d2i.hi(double %1947) #139
  %1968 = and i32 %1967, 2147483647
  %1969 = icmp eq i32 %1968, 2146435072
  %1970 = icmp eq i32 %1966, 0
  %1971 = select i1 %1969, i1 %1970, i1 false
  br label %__nv_isinfd.exit.i.i.i.i.i398

__nv_isinfd.exit.i.i.i.i.i398:                    ; preds = %1965, %1962
  %.054.in = phi i1 [ %1964, %1962 ], [ %1971, %1965 ]
  br i1 %.054.in, label %1972, label %1974

1972:                                             ; preds = %__nv_isinfd.exit.i.i.i.i.i398
  %1973 = fmul double %1947, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i399

1974:                                             ; preds = %__nv_isinfd.exit.i.i.i.i.i398
  %1975 = fmul double %1947, 0x3FE45F306DC9C883
  %1976 = call i32 @llvm.nvvm.d2i.rn(double %1975) #139
  store i32 %1976, i32* %q.i.i.i.i.i.i320, align 4, !noalias !68
  %1977 = sitofp i32 %1976 to double
  %1978 = fneg double %1977
  %1979 = call double @llvm.fma.f64(double %1978, double noundef 0x3FF921FB54442D18, double %1947)
  %1980 = call double @llvm.fma.f64(double %1978, double noundef 0x3C91A62633145C00, double %1979)
  %1981 = call double @llvm.fma.f64(double %1978, double noundef 0x397B839A252049C0, double %1980)
  %1982 = call double @llvm.fabs.f64(double %1947)
  %1983 = fcmp ult double %1982, 0x41E0000000000000
  br i1 %1983, label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i399, label %1984

1984:                                             ; preds = %1974
  %1985 = call fastcc double @__internal_trig_reduction_slowpathd(double %1947, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i.i320) #139
  %.pre86 = load i32, i32* %q.i.i.i.i.i.i320, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i.i399

__internal_trig_reduction_kerneld.exit.i.i.i.i.i399: ; preds = %1974, %1984, %1972
  %i.i.i.i.i.i327.0 = phi i32 [ 0, %1972 ], [ %.pre86, %1984 ], [ %1976, %1974 ]
  %.053 = phi double [ %1973, %1972 ], [ %1985, %1984 ], [ %1981, %1974 ]
  %1986 = fmul double %.053, %.053
  %1987 = call double @llvm.fma.f64(double %1986, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %1988 = call double @llvm.fma.f64(double %1987, double %1986, double noundef 0xBE927E4F8E06E6D9)
  %1989 = call double @llvm.fma.f64(double %1988, double %1986, double noundef 0x3EFA01A019DDBCE9)
  %1990 = call double @llvm.fma.f64(double %1989, double %1986, double noundef 0xBF56C16C16C15D47)
  %1991 = call double @llvm.fma.f64(double %1990, double %1986, double noundef 0x3FA5555555555551)
  %1992 = call double @llvm.fma.f64(double %1991, double %1986, double noundef -5.000000e-01)
  %1993 = call double @llvm.fma.f64(double %1992, double %1986, double noundef 1.000000e+00)
  %1994 = call double @llvm.fma.f64(double %1986, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %1995 = call double @llvm.fma.f64(double %1994, double %1986, double noundef 0x3EC71DE369ACE392)
  %1996 = call double @llvm.fma.f64(double %1995, double %1986, double noundef 0xBF2A01A019DB62A1)
  %1997 = call double @llvm.fma.f64(double %1996, double %1986, double noundef 0x3F81111111110818)
  %1998 = call double @llvm.fma.f64(double %1997, double %1986, double noundef 0xBFC5555555555554)
  %1999 = call double @llvm.fma.f64(double %1998, double %1986, double noundef 0.000000e+00)
  %2000 = call double @llvm.fma.f64(double %1999, double %.053, double %.053)
  %2001 = call i32 @llvm.nvvm.d2i.hi(double %2000) #139
  %2002 = call i32 @llvm.nvvm.d2i.lo(double %2000) #139
  %2003 = xor i32 %2001, -2147483648
  %2004 = call double @llvm.nvvm.lohi.i2d(i32 %2002, i32 %2003) #139
  %2005 = and i32 %i.i.i.i.i.i327.0, 1
  %.not41 = icmp eq i32 %2005, 0
  %spec.select21 = select i1 %.not41, double %1993, double %2004
  %2006 = and i32 %i.i.i.i.i.i327.0, 2
  %.not42 = icmp eq i32 %2006, 0
  br i1 %.not42, label %__internal_cos_offset.exit.i.i404, label %2007

2007:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i399
  %2008 = call i32 @llvm.nvvm.d2i.hi(double %spec.select21) #139
  %2009 = call i32 @llvm.nvvm.d2i.lo(double %spec.select21) #139
  %2010 = xor i32 %2008, -2147483648
  %2011 = call double @llvm.nvvm.lohi.i2d(i32 %2009, i32 %2010) #139
  br label %__internal_cos_offset.exit.i.i404

2012:                                             ; preds = %1950
  %2013 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2014 = icmp eq i32 %2013, 200
  br i1 %2014, label %2021, label %2015

2015:                                             ; preds = %2012
  %2016 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2017 = icmp eq i32 %2016, 350
  br i1 %2017, label %2021, label %2018

2018:                                             ; preds = %2015
  %2019 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2020 = icmp eq i32 %2019, 370
  br i1 %2020, label %2021, label %2024

2021:                                             ; preds = %2018, %2015, %2012
  %2022 = call double @llvm.fabs.f64(double %1947)
  %2023 = fcmp oeq double %2022, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i21.i401

2024:                                             ; preds = %2018
  %2025 = call i32 @llvm.nvvm.d2i.lo(double %1947) #139
  %2026 = call i32 @llvm.nvvm.d2i.hi(double %1947) #139
  %2027 = and i32 %2026, 2147483647
  %2028 = icmp eq i32 %2027, 2146435072
  %2029 = icmp eq i32 %2025, 0
  %2030 = select i1 %2028, i1 %2029, i1 false
  br label %__nv_isinfd.exit.i.i.i21.i401

__nv_isinfd.exit.i.i.i21.i401:                    ; preds = %2024, %2021
  %.055.in = phi i1 [ %2023, %2021 ], [ %2030, %2024 ]
  br i1 %.055.in, label %2031, label %2033

2031:                                             ; preds = %__nv_isinfd.exit.i.i.i21.i401
  %2032 = fmul double %1947, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i403

2033:                                             ; preds = %__nv_isinfd.exit.i.i.i21.i401
  %2034 = fmul double %1947, 0x3FE45F306DC9C883
  %2035 = call i32 @llvm.nvvm.d2i.rn(double %2034) #139
  store i32 %2035, i32* %q.i.i.i.i5.i304, align 4, !noalias !68
  %2036 = sitofp i32 %2035 to double
  %2037 = fneg double %2036
  %2038 = call double @llvm.fma.f64(double %2037, double noundef 0x3FF921FB54442D18, double %1947)
  %2039 = call double @llvm.fma.f64(double %2037, double noundef 0x3C91A62633145C00, double %2038)
  %2040 = call double @llvm.fma.f64(double %2037, double noundef 0x397B839A252049C0, double %2039)
  %2041 = call double @llvm.fabs.f64(double %1947)
  %2042 = fcmp ult double %2041, 0x41E0000000000000
  br i1 %2042, label %__internal_trig_reduction_kerneld.exit.i.i.i22.i402, label %2043

2043:                                             ; preds = %2033
  %2044 = call fastcc double @__internal_trig_reduction_slowpathd(double %1947, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i5.i304) #139
  %.pre85 = load i32, i32* %q.i.i.i.i5.i304, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i22.i402

__internal_trig_reduction_kerneld.exit.i.i.i22.i402: ; preds = %2043, %2033
  %2045 = phi i32 [ %.pre85, %2043 ], [ %2035, %2033 ]
  %t.i1.i.i.i.i303.0 = phi double [ %2044, %2043 ], [ %2040, %2033 ]
  %phi.bo37 = add i32 %2045, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i.i403

__internal_sin_cos_kerneld.exit.i.i.i.i403:       ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i22.i402, %2031
  %i.i.i.i13.i331.0 = phi i32 [ 1, %2031 ], [ %phi.bo37, %__internal_trig_reduction_kerneld.exit.i.i.i22.i402 ]
  %z.i.i.i12.i330.0 = phi double [ %2032, %2031 ], [ %t.i1.i.i.i.i303.0, %__internal_trig_reduction_kerneld.exit.i.i.i22.i402 ]
  %2046 = shl i32 %i.i.i.i13.i331.0, 3
  %2047 = and i32 %2046, 8
  %2048 = zext i32 %2047 to i64
  %2049 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %2048
  %2050 = fmul double %z.i.i.i12.i330.0, %z.i.i.i12.i330.0
  %2051 = and i32 %i.i.i.i13.i331.0, 1
  %.not38 = icmp eq i32 %2051, 0
  %2052 = select i1 %.not38, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %2053 = getelementptr inbounds double, double addrspace(1)* %2049, i64 1
  %2054 = load double, double addrspace(1)* %2053, align 8
  %2055 = call double @llvm.fma.f64(double noundef %2052, double %2050, double %2054)
  %2056 = getelementptr inbounds double, double addrspace(1)* %2049, i64 2
  %2057 = load double, double addrspace(1)* %2056, align 8
  %2058 = call double @llvm.fma.f64(double %2055, double %2050, double %2057)
  %2059 = getelementptr inbounds double, double addrspace(1)* %2049, i64 3
  %2060 = load double, double addrspace(1)* %2059, align 8
  %2061 = call double @llvm.fma.f64(double %2058, double %2050, double %2060)
  %2062 = getelementptr inbounds double, double addrspace(1)* %2049, i64 4
  %2063 = load double, double addrspace(1)* %2062, align 8
  %2064 = call double @llvm.fma.f64(double %2061, double %2050, double %2063)
  %2065 = getelementptr inbounds double, double addrspace(1)* %2049, i64 5
  %2066 = load double, double addrspace(1)* %2065, align 8
  %2067 = call double @llvm.fma.f64(double %2064, double %2050, double %2066)
  %2068 = getelementptr inbounds double, double addrspace(1)* %2049, i64 6
  %2069 = load double, double addrspace(1)* %2068, align 8
  %2070 = call double @llvm.fma.f64(double %2067, double %2050, double %2069)
  %2071 = call double @llvm.fma.f64(double %2070, double %z.i.i.i12.i330.0, double %z.i.i.i12.i330.0)
  %2072 = call double @llvm.fma.f64(double %2070, double %2050, double noundef 1.000000e+00)
  %spec.select22 = select i1 %.not38, double %2071, double %2072
  %2073 = and i32 %i.i.i.i13.i331.0, 2
  %.not40 = icmp eq i32 %2073, 0
  %2074 = call double @llvm.fma.f64(double %spec.select22, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.157 = select i1 %.not40, double %spec.select22, double %2074
  br label %__internal_cos_offset.exit.i.i404

__internal_cos_offset.exit.i.i404:                ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i.i399, %2007, %__internal_sin_cos_kerneld.exit.i.i.i.i403
  %.051 = phi double [ %.157, %__internal_sin_cos_kerneld.exit.i.i.i.i403 ], [ %2011, %2007 ], [ %spec.select21, %__internal_trig_reduction_kerneld.exit.i.i.i.i.i399 ]
  %2075 = fmul double %1930, %.051
  br label %__nv_j1.exit.i405

__nv_j1.exit.i405:                                ; preds = %1846, %__internal_cos_offset.exit.i.i404, %__nv_isinfd.exit.i.i396, %1867, %1827
  %r.i.i340.3 = phi double [ %1843, %1827 ], [ %1864, %1846 ], [ %1885, %1867 ], [ %2075, %__internal_cos_offset.exit.i.i404 ], [ 0.000000e+00, %__nv_isinfd.exit.i.i396 ]
  %2076 = fcmp olt double %1, 1.000000e-30
  %2077 = fmul double %1, 5.000000e-01
  %r.i.i340.5 = select i1 %2076, double %2077, double %r.i.i340.3
  %2078 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %2079 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %2080 = icmp slt i32 %2078, 1048576
  br i1 %2080, label %2081, label %2085

2081:                                             ; preds = %__nv_j1.exit.i405
  %2082 = fmul double %1, 0x4350000000000000
  %2083 = call i32 @llvm.nvvm.d2i.hi(double %2082) #139
  %2084 = call i32 @llvm.nvvm.d2i.lo(double %2082) #139
  br label %2085

2085:                                             ; preds = %2081, %__nv_j1.exit.i405
  %.059 = phi double [ %2082, %2081 ], [ %1, %__nv_j1.exit.i405 ]
  %ihi.i.i291.0 = phi i32 [ %2083, %2081 ], [ %2078, %__nv_j1.exit.i405 ]
  %ilo.i.i292.0 = phi i32 [ %2084, %2081 ], [ %2079, %__nv_j1.exit.i405 ]
  %e.i.i293.0 = phi i32 [ -1077, %2081 ], [ -1023, %__nv_j1.exit.i405 ]
  %2086 = add i32 %ihi.i.i291.0, -1
  %2087 = icmp ult i32 %2086, 2146435071
  br i1 %2087, label %2088, label %2137

2088:                                             ; preds = %2085
  %2089 = lshr i32 %ihi.i.i291.0, 20
  %2090 = add nsw i32 %e.i.i293.0, %2089
  %2091 = and i32 %ihi.i.i291.0, -2146435073
  %2092 = or i32 %2091, 1072693248
  %2093 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i292.0, i32 %2092) #139
  %2094 = icmp sgt i32 %2092, 1073127582
  br i1 %2094, label %2095, label %2101

2095:                                             ; preds = %2088
  %2096 = call i32 @llvm.nvvm.d2i.lo(double %2093) #139
  %2097 = call i32 @llvm.nvvm.d2i.hi(double %2093) #139
  %2098 = add i32 %2097, -1048576
  %2099 = call double @llvm.nvvm.lohi.i2d(i32 %2096, i32 %2098) #139
  %2100 = add nsw i32 %2090, 1
  br label %2101

2101:                                             ; preds = %2095, %2088
  %m.i.i281.0 = phi double [ %2099, %2095 ], [ %2093, %2088 ]
  %e.i.i293.1 = phi i32 [ %2100, %2095 ], [ %2090, %2088 ]
  %2102 = fadd double %m.i.i281.0, -1.000000e+00
  %2103 = fadd double %m.i.i281.0, 1.000000e+00
  %2104 = call double @llvm.nvvm.rcp.approx.ftz.d(double %2103) #139
  %2105 = fneg double %2103
  %2106 = call double @llvm.fma.f64(double %2105, double %2104, double noundef 1.000000e+00)
  %2107 = call double @llvm.fma.f64(double %2106, double %2106, double %2106)
  %2108 = call double @llvm.fma.f64(double %2107, double %2104, double %2104)
  %2109 = fmul double %2102, %2108
  %2110 = fadd double %2109, %2109
  %2111 = fmul double %2110, %2110
  %2112 = call double @llvm.fma.f64(double %2111, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %2113 = call double @llvm.fma.f64(double %2112, double %2111, double noundef 0x3EF3B2669F02676F)
  %2114 = call double @llvm.fma.f64(double %2113, double %2111, double noundef 0x3F1745CBA9AB0956)
  %2115 = call double @llvm.fma.f64(double %2114, double %2111, double noundef 0x3F3C71C72D1B5154)
  %2116 = call double @llvm.fma.f64(double %2115, double %2111, double noundef 0x3F624924923BE72D)
  %2117 = call double @llvm.fma.f64(double %2116, double %2111, double noundef 0x3F8999999999A3C4)
  %2118 = call double @llvm.fma.f64(double %2117, double %2111, double noundef 0x3FB5555555555554)
  %2119 = fsub double %2102, %2110
  %2120 = fmul double %2119, 2.000000e+00
  %2121 = fneg double %2110
  %2122 = call double @llvm.fma.f64(double %2121, double %2102, double %2120)
  %2123 = fmul double %2108, %2122
  %2124 = fmul double %2111, %2118
  %2125 = call double @llvm.fma.f64(double %2124, double %2110, double %2123)
  %2126 = xor i32 %e.i.i293.1, -2147483648
  %2127 = call double @llvm.nvvm.lohi.i2d(i32 %2126, i32 noundef 1127219200) #139
  %2128 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %2129 = fsub double %2127, %2128
  %2130 = call double @llvm.fma.f64(double %2129, double noundef 0x3FE62E42FEFA39EF, double %2110)
  %2131 = fneg double %2129
  %2132 = call double @llvm.fma.f64(double %2131, double noundef 0x3FE62E42FEFA39EF, double %2130)
  %2133 = fsub double %2132, %2110
  %2134 = fsub double %2125, %2133
  %2135 = call double @llvm.fma.f64(double %2129, double noundef 0x3C7ABC9E3B39803F, double %2134)
  %2136 = fadd double %2130, %2135
  br label %__nv_log.exit.i406

2137:                                             ; preds = %2085
  %2138 = call double @llvm.fma.f64(double %.059, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %2139 = call i32 @llvm.nvvm.d2i.hi(double %.059) #139
  %2140 = bitcast i32 %2139 to float
  %2141 = fcmp oeq float %2140, 0.000000e+00
  %q.i28.i287.0 = select i1 %2141, double 0xFFF0000000000000, double %2138
  br label %__nv_log.exit.i406

__nv_log.exit.i406:                               ; preds = %2137, %2101
  %q.i28.i287.1 = phi double [ %2136, %2101 ], [ %q.i28.i287.0, %2137 ]
  %2142 = fmul double %r.i.i340.5, %q.i28.i287.1
  %2143 = fdiv double 1.000000e+00, %1
  %2144 = fsub double %2142, %2143
  %2145 = fmul double %2144, 0x3FE45F306DC9C883
  %2146 = fadd double %1825, %2145
  br label %2413

2147:                                             ; preds = %1814
  %2148 = fcmp ugt double %1, 0x4009B510EC2ADC83
  br i1 %2148, label %2176, label %2149

2149:                                             ; preds = %2147
  %2150 = fadd double %1, 0xC00193BED4DFF243
  %2151 = fadd double %2150, 0x3C8BD1E50D219BFD
  %2152 = call double @llvm.fma.f64(double %2151, double noundef 0xBE2B87B0BE2AA150, double noundef 0x3E4833AAE4D8B975)
  %2153 = call double @llvm.fma.f64(double %2152, double %2151, double noundef 0xBE44E279B423E68F)
  %2154 = call double @llvm.fma.f64(double %2153, double %2151, double noundef 0x3E129DC6A747EB4F)
  %2155 = call double @llvm.fma.f64(double %2154, double %2151, double noundef 0xBE61D15534496CD8)
  %2156 = call double @llvm.fma.f64(double %2155, double %2151, double noundef 0x3E7EEC8D48FECE00)
  %2157 = call double @llvm.fma.f64(double %2156, double %2151, double noundef 0xBE8D1180AF70A134)
  %2158 = call double @llvm.fma.f64(double %2157, double %2151, double noundef 0x3E9C8386A0EA1388)
  %2159 = call double @llvm.fma.f64(double %2158, double %2151, double noundef 0xBEB01A014E7F3250)
  %2160 = call double @llvm.fma.f64(double %2159, double %2151, double noundef 0x3EC1FB752010A320)
  %2161 = call double @llvm.fma.f64(double %2160, double %2151, double noundef 0xBED3AA0AFF4E332B)
  %2162 = call double @llvm.fma.f64(double %2161, double %2151, double noundef 0x3EE584A6C77F6700)
  %2163 = call double @llvm.fma.f64(double %2162, double %2151, double noundef 0xBEF794C520FC2EBB)
  %2164 = call double @llvm.fma.f64(double %2163, double %2151, double noundef 0x3F09D18D2D35CC71)
  %2165 = call double @llvm.fma.f64(double %2164, double %2151, double noundef 0xBF1C3FB7315C4599)
  %2166 = call double @llvm.fma.f64(double %2165, double %2151, double noundef 0x3F2EEA7ADECCE927)
  %2167 = call double @llvm.fma.f64(double %2166, double %2151, double noundef 0xBF40B2D85257446F)
  %2168 = call double @llvm.fma.f64(double %2167, double %2151, double noundef 0x3F517AB4B1FE5D5B)
  %2169 = call double @llvm.fma.f64(double %2168, double %2151, double noundef 0xBF65429DC6516C0D)
  %2170 = call double @llvm.fma.f64(double %2169, double %2151, double noundef 0x3F7E671C7D0B090B)
  %2171 = call double @llvm.fma.f64(double %2170, double %2151, double noundef 0xBF73A6DEC36FB27C)
  %2172 = call double @llvm.fma.f64(double %2171, double %2151, double noundef 0xBFA0D2AF4E931FD1)
  %2173 = call double @llvm.fma.f64(double %2172, double %2151, double noundef 0xBFBE56F82217B964)
  %2174 = call double @llvm.fma.f64(double %2173, double %2151, double noundef 0x3FE0AA48442F014B)
  %2175 = fmul double %2151, %2174
  br label %2413

2176:                                             ; preds = %2147
  %2177 = fcmp ugt double %1, 0x401C0D26D5A541CB
  br i1 %2177, label %2202, label %2178

2178:                                             ; preds = %2176
  %2179 = fadd double %1, 0xC015B7FE4E87B02E
  %2180 = fadd double %2179, 0xBCBDFE7BAC228E8C
  %2181 = call double @llvm.fma.f64(double %2180, double noundef 0x3CBA3C76069F1D8C, double noundef 0x3CC69A30996793E2)
  %2182 = call double @llvm.fma.f64(double %2181, double %2180, double noundef 0xBCDDD8432FE756E7)
  %2183 = call double @llvm.fma.f64(double %2182, double %2180, double noundef 0xBD143158EEE220F7)
  %2184 = call double @llvm.fma.f64(double %2183, double %2180, double noundef 0x3D28D44491230F5A)
  %2185 = call double @llvm.fma.f64(double %2184, double %2180, double noundef 0xBD438842EAF4EDBC)
  %2186 = call double @llvm.fma.f64(double %2185, double %2180, double noundef 0x3D74958DAFBFAF5A)
  %2187 = call double @llvm.fma.f64(double %2186, double %2180, double noundef 0xBD9449A60E664848)
  %2188 = call double @llvm.fma.f64(double %2187, double %2180, double noundef 0x3D838BC8CD594A76)
  %2189 = call double @llvm.fma.f64(double %2188, double %2180, double noundef 0xBDFA798002141323)
  %2190 = call double @llvm.fma.f64(double %2189, double %2180, double noundef 0x3E380B4198956AAA)
  %2191 = call double @llvm.fma.f64(double %2190, double %2180, double noundef 0x3E5B62B5F21BACD4)
  %2192 = call double @llvm.fma.f64(double %2191, double %2180, double noundef 0xBEA255E729FB6AAE)
  %2193 = call double @llvm.fma.f64(double %2192, double %2180, double noundef 0xBEC80618F6BAE5AA)
  %2194 = call double @llvm.fma.f64(double %2193, double %2180, double noundef 0x3F085B940F8E8D36)
  %2195 = call double @llvm.fma.f64(double %2194, double %2180, double noundef 0x3F2337C7E10E14E8)
  %2196 = call double @llvm.fma.f64(double %2195, double %2180, double noundef 0xBF61BE6DB99332CA)
  %2197 = call double @llvm.fma.f64(double %2196, double %2180, double noundef 0xBF710A329E2BE9B8)
  %2198 = call double @llvm.fma.f64(double %2197, double %2180, double noundef 0x3FAA15D92DFE3FCF)
  %2199 = call double @llvm.fma.f64(double %2198, double %2180, double noundef 0x3FA00B9F8571C9BE)
  %2200 = call double @llvm.fma.f64(double %2199, double %2180, double noundef 0xBFD5C7C556F0C19A)
  %2201 = fmul double %2180, %2200
  br label %2413

2202:                                             ; preds = %2176
  %2203 = fcmp ugt double %1, 0x4022585C739ACDDD
  br i1 %2203, label %2223, label %2204

2204:                                             ; preds = %2202
  %2205 = fadd double %1, 0xC0213127AE6169B4
  %2206 = fadd double %2205, 0xBCB479CC068D9046
  %2207 = call double @llvm.fma.f64(double %2206, double noundef 0x3CB09CCC22945996, double noundef 0xBD43515F67644276)
  %2208 = call double @llvm.fma.f64(double %2207, double %2206, double noundef 0xBD72C5B978E9F5C7)
  %2209 = call double @llvm.fma.f64(double %2208, double %2206, double noundef 0x3DBEC1151613913C)
  %2210 = call double @llvm.fma.f64(double %2209, double %2206, double noundef 0x3DE9E38D13C4A824)
  %2211 = call double @llvm.fma.f64(double %2210, double %2206, double noundef 0xBE341E75E1088EB5)
  %2212 = call double @llvm.fma.f64(double %2211, double %2206, double noundef 0xBE5A384EBB13CFE1)
  %2213 = call double @llvm.fma.f64(double %2212, double %2206, double noundef 0x3EA2BECB27F8C8F8)
  %2214 = call double @llvm.fma.f64(double %2213, double %2206, double noundef 0x3EC176E72B989FD8)
  %2215 = call double @llvm.fma.f64(double %2214, double %2206, double noundef 0xBF06F7BAB102F822)
  %2216 = call double @llvm.fma.f64(double %2215, double %2206, double noundef 0xBF1B50D7E1D278E1)
  %2217 = call double @llvm.fma.f64(double %2216, double %2206, double noundef 0x3F607A678D60004F)
  %2218 = call double @llvm.fma.f64(double %2217, double %2206, double noundef 0x3F63CED2A2E69115)
  %2219 = call double @llvm.fma.f64(double %2218, double %2206, double noundef 0xBFA6395DFE49FCD4)
  %2220 = call double @llvm.fma.f64(double %2219, double %2206, double noundef 0xBF902B3933CF21B1)
  %2221 = call double @llvm.fma.f64(double %2220, double %2206, double noundef 0x3FD15F993FCEAB5C)
  %2222 = fmul double %2206, %2221
  br label %2413

2223:                                             ; preds = %2202
  %2224 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2225 = icmp eq i32 %2224, 200
  br i1 %2225, label %2232, label %2226

2226:                                             ; preds = %2223
  %2227 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2228 = icmp eq i32 %2227, 350
  br i1 %2228, label %2232, label %2229

2229:                                             ; preds = %2226
  %2230 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2231 = icmp eq i32 %2230, 370
  br i1 %2231, label %2232, label %2234

2232:                                             ; preds = %2229, %2226, %2223
  %2233 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i407

2234:                                             ; preds = %2229
  %2235 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %2236 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %2237 = and i32 %2236, 2147483647
  %2238 = icmp eq i32 %2237, 2146435072
  %2239 = icmp eq i32 %2235, 0
  %2240 = select i1 %2238, i1 %2239, i1 false
  br label %__nv_isinfd.exit.i407

__nv_isinfd.exit.i407:                            ; preds = %2234, %2232
  %.052.in = phi i1 [ %2233, %2232 ], [ %2240, %2234 ]
  br i1 %.052.in, label %2413, label %2241

2241:                                             ; preds = %__nv_isinfd.exit.i407
  %2242 = call double @llvm.nvvm.rcp.approx.ftz.d(double %1) #139
  %2243 = fneg double %1
  %2244 = call double @llvm.fma.f64(double %2243, double %2242, double noundef 1.000000e+00)
  %2245 = call double @llvm.fma.f64(double %2244, double %2244, double %2244)
  %2246 = call double @llvm.fma.f64(double %2245, double %2242, double %2242)
  %2247 = fmul double %2246, %2246
  %2248 = call double @llvm.fma.f64(double %2247, double noundef 0x40D13DB326ECEBFE, double noundef 0xC09C26E89385D5B1)
  %2249 = call double @llvm.fma.f64(double %2248, double %2247, double noundef 0x405C6AB923C6F55E)
  %2250 = call double @llvm.fma.f64(double %2249, double %2247, double noundef 0xC01E61EAF3BD2FA1)
  %2251 = call double @llvm.fma.f64(double %2250, double %2247, double noundef 0x3FE9BF15D9B97DD1)
  %2252 = call double @llvm.fma.f64(double %2251, double %2247, double noundef 0xBFC8BFECF93D7D19)
  %2253 = call double @llvm.fma.f64(double %2252, double %2247, double noundef 0x3FC7FFFFF756AA6C)
  %2254 = call double @llvm.fma.f64(double %2253, double %2247, double noundef 0x3FF0000000003646)
  %2255 = call double @llvm.fma.f64(double %2247, double noundef 0xC1943281A050209C, double noundef 0x416024E99BA46E7B)
  %2256 = call double @llvm.fma.f64(double %2255, double %2247, double noundef 0xC11A6875D7DFBD65)
  %2257 = call double @llvm.fma.f64(double %2256, double %2247, double noundef 0x40D032C041790233)
  %2258 = call double @llvm.fma.f64(double %2257, double %2247, double noundef 0xC0839F895BC22946)
  %2259 = call double @llvm.fma.f64(double %2258, double %2247, double noundef 0x403E77CC78ECD2D8)
  %2260 = call double @llvm.fma.f64(double %2259, double %2247, double noundef 0xC002F368D0117BE9)
  %2261 = call double @llvm.fma.f64(double %2260, double %2247, double noundef 0x3FD7BCC786009A25)
  %2262 = call double @llvm.fma.f64(double %2261, double %2247, double noundef 0xBFC4FFFFFC51BC7A)
  %2263 = call double @llvm.fma.f64(double %2262, double %2247, double noundef 0x3FD7FFFFFFFFB5EA)
  %2264 = call double @llvm.fma.f64(double %2263, double %2246, double %1)
  %2265 = call double @llvm.nvvm.rsqrt.approx.d(double %1) #139
  %2266 = fmul double %2265, 0x3FE9884533D43651
  %2267 = fmul double %2266, %2254
  %2268 = fmul double %2264, 0x3FE45F306DC9C883
  %2269 = call i32 @llvm.nvvm.d2i.rn(double %2268) #139
  store i32 %2269, i32* %q.i.i.i381, align 4, !noalias !68
  %2270 = sitofp i32 %2269 to double
  %2271 = fneg double %2270
  %2272 = call double @llvm.fma.f64(double %2271, double noundef 0x3FF921FB54442D18, double %2264)
  %2273 = call double @llvm.fma.f64(double %2271, double noundef 0x3C91A62633145C00, double %2272)
  %2274 = call double @llvm.fma.f64(double %2271, double noundef 0x397B839A252049C0, double %2273)
  %2275 = call double @llvm.fabs.f64(double %2264)
  %2276 = fcmp ult double %2275, 0x41E0000000000000
  br i1 %2276, label %__internal_trig_reduction_kerneld.exit.i.i408, label %2277

2277:                                             ; preds = %2241
  %2278 = call fastcc double @__internal_trig_reduction_slowpathd(double %2264, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i381) #139
  %.pre87 = load i32, i32* %q.i.i.i381, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i408

__internal_trig_reduction_kerneld.exit.i.i408:    ; preds = %2277, %2241
  %2279 = phi i32 [ %.pre87, %2277 ], [ %2269, %2241 ]
  %t.i.i.i380.0 = phi double [ %2278, %2277 ], [ %2274, %2241 ]
  %2280 = and i32 %2279, 3
  %2281 = uitofp i32 %2280 to double
  %2282 = fadd double %t.i.i.i380.0, 0xC00F6A7A2955385E
  %2283 = fmul double %2281, 0x3FF921FB54442D18
  %2284 = fadd double %2282, %2283
  %2285 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2286 = icmp eq i32 %2285, 350
  br i1 %2286, label %2290, label %2287

2287:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i408
  %2288 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2289 = icmp eq i32 %2288, 370
  br i1 %2289, label %2290, label %2349

2290:                                             ; preds = %2287, %__internal_trig_reduction_kerneld.exit.i.i408
  %2291 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2292 = icmp eq i32 %2291, 200
  br i1 %2292, label %2299, label %2293

2293:                                             ; preds = %2290
  %2294 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2295 = icmp eq i32 %2294, 350
  br i1 %2295, label %2299, label %2296

2296:                                             ; preds = %2293
  %2297 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2298 = icmp eq i32 %2297, 370
  br i1 %2298, label %2299, label %2302

2299:                                             ; preds = %2296, %2293, %2290
  %2300 = call double @llvm.fabs.f64(double %2284)
  %2301 = fcmp oeq double %2300, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i409

2302:                                             ; preds = %2296
  %2303 = call i32 @llvm.nvvm.d2i.lo(double %2284) #139
  %2304 = call i32 @llvm.nvvm.d2i.hi(double %2284) #139
  %2305 = and i32 %2304, 2147483647
  %2306 = icmp eq i32 %2305, 2146435072
  %2307 = icmp eq i32 %2303, 0
  %2308 = select i1 %2306, i1 %2307, i1 false
  br label %__nv_isinfd.exit.i.i.i.i409

__nv_isinfd.exit.i.i.i.i409:                      ; preds = %2302, %2299
  %.047.in = phi i1 [ %2301, %2299 ], [ %2308, %2302 ]
  br i1 %.047.in, label %2309, label %2311

2309:                                             ; preds = %__nv_isinfd.exit.i.i.i.i409
  %2310 = fmul double %2284, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i410

2311:                                             ; preds = %__nv_isinfd.exit.i.i.i.i409
  %2312 = fmul double %2284, 0x3FE45F306DC9C883
  %2313 = call i32 @llvm.nvvm.d2i.rn(double %2312) #139
  store i32 %2313, i32* %q.i.i.i.i.i367, align 4, !noalias !68
  %2314 = sitofp i32 %2313 to double
  %2315 = fneg double %2314
  %2316 = call double @llvm.fma.f64(double %2315, double noundef 0x3FF921FB54442D18, double %2284)
  %2317 = call double @llvm.fma.f64(double %2315, double noundef 0x3C91A62633145C00, double %2316)
  %2318 = call double @llvm.fma.f64(double %2315, double noundef 0x397B839A252049C0, double %2317)
  %2319 = call double @llvm.fabs.f64(double %2284)
  %2320 = fcmp ult double %2319, 0x41E0000000000000
  br i1 %2320, label %__internal_trig_reduction_kerneld.exit.i.i.i.i410, label %2321

2321:                                             ; preds = %2311
  %2322 = call fastcc double @__internal_trig_reduction_slowpathd(double %2284, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i367) #139
  %.pre89 = load i32, i32* %q.i.i.i.i.i367, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i410

__internal_trig_reduction_kerneld.exit.i.i.i.i410: ; preds = %2311, %2321, %2309
  %i.i.i.i.i374.0 = phi i32 [ 0, %2309 ], [ %.pre89, %2321 ], [ %2313, %2311 ]
  %.046 = phi double [ %2310, %2309 ], [ %2322, %2321 ], [ %2318, %2311 ]
  %2323 = fmul double %.046, %.046
  %2324 = call double @llvm.fma.f64(double %2323, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %2325 = call double @llvm.fma.f64(double %2324, double %2323, double noundef 0xBE927E4F8E06E6D9)
  %2326 = call double @llvm.fma.f64(double %2325, double %2323, double noundef 0x3EFA01A019DDBCE9)
  %2327 = call double @llvm.fma.f64(double %2326, double %2323, double noundef 0xBF56C16C16C15D47)
  %2328 = call double @llvm.fma.f64(double %2327, double %2323, double noundef 0x3FA5555555555551)
  %2329 = call double @llvm.fma.f64(double %2328, double %2323, double noundef -5.000000e-01)
  %2330 = call double @llvm.fma.f64(double %2329, double %2323, double noundef 1.000000e+00)
  %2331 = call double @llvm.fma.f64(double %2323, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %2332 = call double @llvm.fma.f64(double %2331, double %2323, double noundef 0x3EC71DE369ACE392)
  %2333 = call double @llvm.fma.f64(double %2332, double %2323, double noundef 0xBF2A01A019DB62A1)
  %2334 = call double @llvm.fma.f64(double %2333, double %2323, double noundef 0x3F81111111110818)
  %2335 = call double @llvm.fma.f64(double %2334, double %2323, double noundef 0xBFC5555555555554)
  %2336 = call double @llvm.fma.f64(double %2335, double %2323, double noundef 0.000000e+00)
  %2337 = call double @llvm.fma.f64(double %2336, double %.046, double %.046)
  %2338 = call i32 @llvm.nvvm.d2i.hi(double %2337) #139
  %2339 = call i32 @llvm.nvvm.d2i.lo(double %2337) #139
  %2340 = xor i32 %2338, -2147483648
  %2341 = call double @llvm.nvvm.lohi.i2d(i32 %2339, i32 %2340) #139
  %2342 = and i32 %i.i.i.i.i374.0, 1
  %.not35 = icmp eq i32 %2342, 0
  %spec.select25 = select i1 %.not35, double %2330, double %2341
  %2343 = and i32 %i.i.i.i.i374.0, 2
  %.not36 = icmp eq i32 %2343, 0
  br i1 %.not36, label %__internal_cos_offset.exit.i415, label %2344

2344:                                             ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i410
  %2345 = call i32 @llvm.nvvm.d2i.hi(double %spec.select25) #139
  %2346 = call i32 @llvm.nvvm.d2i.lo(double %spec.select25) #139
  %2347 = xor i32 %2345, -2147483648
  %2348 = call double @llvm.nvvm.lohi.i2d(i32 %2346, i32 %2347) #139
  br label %__internal_cos_offset.exit.i415

2349:                                             ; preds = %2287
  %2350 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2351 = icmp eq i32 %2350, 200
  br i1 %2351, label %2358, label %2352

2352:                                             ; preds = %2349
  %2353 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2354 = icmp eq i32 %2353, 350
  br i1 %2354, label %2358, label %2355

2355:                                             ; preds = %2352
  %2356 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2357 = icmp eq i32 %2356, 370
  br i1 %2357, label %2358, label %2361

2358:                                             ; preds = %2355, %2352, %2349
  %2359 = call double @llvm.fabs.f64(double %2284)
  %2360 = fcmp oeq double %2359, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i412

2361:                                             ; preds = %2355
  %2362 = call i32 @llvm.nvvm.d2i.lo(double %2284) #139
  %2363 = call i32 @llvm.nvvm.d2i.hi(double %2284) #139
  %2364 = and i32 %2363, 2147483647
  %2365 = icmp eq i32 %2364, 2146435072
  %2366 = icmp eq i32 %2362, 0
  %2367 = select i1 %2365, i1 %2366, i1 false
  br label %__nv_isinfd.exit.i.i.i412

__nv_isinfd.exit.i.i.i412:                        ; preds = %2361, %2358
  %.048.in = phi i1 [ %2360, %2358 ], [ %2367, %2361 ]
  br i1 %.048.in, label %2368, label %2370

2368:                                             ; preds = %__nv_isinfd.exit.i.i.i412
  %2369 = fmul double %2284, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i414

2370:                                             ; preds = %__nv_isinfd.exit.i.i.i412
  %2371 = fmul double %2284, 0x3FE45F306DC9C883
  %2372 = call i32 @llvm.nvvm.d2i.rn(double %2371) #139
  store i32 %2372, i32* %q.i.i.i.i351, align 4, !noalias !68
  %2373 = sitofp i32 %2372 to double
  %2374 = fneg double %2373
  %2375 = call double @llvm.fma.f64(double %2374, double noundef 0x3FF921FB54442D18, double %2284)
  %2376 = call double @llvm.fma.f64(double %2374, double noundef 0x3C91A62633145C00, double %2375)
  %2377 = call double @llvm.fma.f64(double %2374, double noundef 0x397B839A252049C0, double %2376)
  %2378 = call double @llvm.fabs.f64(double %2284)
  %2379 = fcmp ult double %2378, 0x41E0000000000000
  br i1 %2379, label %__internal_trig_reduction_kerneld.exit.i.i.i413, label %2380

2380:                                             ; preds = %2370
  %2381 = call fastcc double @__internal_trig_reduction_slowpathd(double %2284, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i351) #139
  %.pre88 = load i32, i32* %q.i.i.i.i351, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i413

__internal_trig_reduction_kerneld.exit.i.i.i413:  ; preds = %2380, %2370
  %2382 = phi i32 [ %.pre88, %2380 ], [ %2372, %2370 ]
  %t.i1.i.i.i350.0 = phi double [ %2381, %2380 ], [ %2377, %2370 ]
  %phi.bo = add i32 %2382, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i414

__internal_sin_cos_kerneld.exit.i.i.i414:         ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i413, %2368
  %i.i.i.i378.0 = phi i32 [ 1, %2368 ], [ %phi.bo, %__internal_trig_reduction_kerneld.exit.i.i.i413 ]
  %z.i.i.i377.0 = phi double [ %2369, %2368 ], [ %t.i1.i.i.i350.0, %__internal_trig_reduction_kerneld.exit.i.i.i413 ]
  %2383 = shl i32 %i.i.i.i378.0, 3
  %2384 = and i32 %2383, 8
  %2385 = zext i32 %2384 to i64
  %2386 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %2385
  %2387 = fmul double %z.i.i.i377.0, %z.i.i.i377.0
  %2388 = and i32 %i.i.i.i378.0, 1
  %.not32 = icmp eq i32 %2388, 0
  %2389 = select i1 %.not32, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %2390 = getelementptr inbounds double, double addrspace(1)* %2386, i64 1
  %2391 = load double, double addrspace(1)* %2390, align 8
  %2392 = call double @llvm.fma.f64(double noundef %2389, double %2387, double %2391)
  %2393 = getelementptr inbounds double, double addrspace(1)* %2386, i64 2
  %2394 = load double, double addrspace(1)* %2393, align 8
  %2395 = call double @llvm.fma.f64(double %2392, double %2387, double %2394)
  %2396 = getelementptr inbounds double, double addrspace(1)* %2386, i64 3
  %2397 = load double, double addrspace(1)* %2396, align 8
  %2398 = call double @llvm.fma.f64(double %2395, double %2387, double %2397)
  %2399 = getelementptr inbounds double, double addrspace(1)* %2386, i64 4
  %2400 = load double, double addrspace(1)* %2399, align 8
  %2401 = call double @llvm.fma.f64(double %2398, double %2387, double %2400)
  %2402 = getelementptr inbounds double, double addrspace(1)* %2386, i64 5
  %2403 = load double, double addrspace(1)* %2402, align 8
  %2404 = call double @llvm.fma.f64(double %2401, double %2387, double %2403)
  %2405 = getelementptr inbounds double, double addrspace(1)* %2386, i64 6
  %2406 = load double, double addrspace(1)* %2405, align 8
  %2407 = call double @llvm.fma.f64(double %2404, double %2387, double %2406)
  %2408 = call double @llvm.fma.f64(double %2407, double %z.i.i.i377.0, double %z.i.i.i377.0)
  %2409 = call double @llvm.fma.f64(double %2407, double %2387, double noundef 1.000000e+00)
  %spec.select26 = select i1 %.not32, double %2408, double %2409
  %2410 = and i32 %i.i.i.i378.0, 2
  %.not34 = icmp eq i32 %2410, 0
  %2411 = call double @llvm.fma.f64(double %spec.select26, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.150 = select i1 %.not34, double %spec.select26, double %2411
  br label %__internal_cos_offset.exit.i415

__internal_cos_offset.exit.i415:                  ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i410, %2344, %__internal_sin_cos_kerneld.exit.i.i.i414
  %.045 = phi double [ %.150, %__internal_sin_cos_kerneld.exit.i.i.i414 ], [ %2348, %2344 ], [ %spec.select25, %__internal_trig_reduction_kerneld.exit.i.i.i.i410 ]
  %2412 = fmul double %2267, %.045
  br label %2413

2413:                                             ; preds = %__nv_log.exit.i406, %2178, %__internal_cos_offset.exit.i415, %__nv_isinfd.exit.i407, %2204, %2149, %1812
  %r.i389.5 = phi double [ %1813, %1812 ], [ %2146, %__nv_log.exit.i406 ], [ %2175, %2149 ], [ %2201, %2178 ], [ %2222, %2204 ], [ %2412, %__internal_cos_offset.exit.i415 ], [ 0.000000e+00, %__nv_isinfd.exit.i407 ]
  br i1 %604, label %__nv_y1.exit416, label %2414

2414:                                             ; preds = %2413
  br label %__nv_y1.exit416

2415:                                             ; preds = %__nv_y0.exit
  %2416 = icmp slt i32 %n, 0
  %2417 = fcmp ult double %a, 0.000000e+00
  %or.cond = select i1 %2416, i1 true, i1 %2417
  br i1 %or.cond, label %__nv_y1.exit416, label %2418

2418:                                             ; preds = %2415
  br i1 %2, label %2419, label %.lr.ph

2419:                                             ; preds = %2418
  %2420 = fdiv double 0xBFE45F306DC9C883, %a
  br label %__nv_y1.exit416

.lr.ph:                                           ; preds = %2418
  %2421 = fdiv double 2.000000e+00, %a
  %2422 = add nsw i32 %n, -1
  %2423 = add nsw i32 %n, -2
  %xtraiter = and i32 %2422, 3
  %2424 = icmp ult i32 %2423, 3
  br i1 %2424, label %._crit_edge.unr-lcssa, label %.lr.ph.new

.lr.ph.new:                                       ; preds = %.lr.ph
  %unroll_iter = and i32 %2422, -4
  br label %2425

2425:                                             ; preds = %2425, %.lr.ph.new
  %i.071 = phi i32 [ 1, %.lr.ph.new ], [ %2445, %2425 ]
  %yim1.070 = phi double [ %r.i95.5, %.lr.ph.new ], [ %2439, %2425 ]
  %yi.069 = phi double [ %r.i.7, %.lr.ph.new ], [ %2444, %2425 ]
  %niter = phi i32 [ 0, %.lr.ph.new ], [ %niter.next.3, %2425 ]
  %2426 = sitofp i32 %i.071 to double
  %2427 = fmul double %2421, %2426
  %2428 = fneg double %yim1.070
  %2429 = call double @llvm.fma.f64(double %2427, double %yi.069, double %2428)
  %2430 = add nuw nsw i32 %i.071, 1
  %2431 = sitofp i32 %2430 to double
  %2432 = fmul double %2421, %2431
  %2433 = fneg double %yi.069
  %2434 = call double @llvm.fma.f64(double %2432, double %2429, double %2433)
  %2435 = add nuw nsw i32 %i.071, 2
  %2436 = sitofp i32 %2435 to double
  %2437 = fmul double %2421, %2436
  %2438 = fneg double %2429
  %2439 = call double @llvm.fma.f64(double %2437, double %2434, double %2438)
  %2440 = add nuw nsw i32 %i.071, 3
  %2441 = sitofp i32 %2440 to double
  %2442 = fmul double %2421, %2441
  %2443 = fneg double %2434
  %2444 = call double @llvm.fma.f64(double %2442, double %2439, double %2443)
  %2445 = add nuw nsw i32 %i.071, 4
  %niter.next.3 = add i32 %niter, 4
  %niter.ncmp.3.not = icmp eq i32 %niter.next.3, %unroll_iter
  br i1 %niter.ncmp.3.not, label %._crit_edge.unr-lcssa, label %2425

._crit_edge.unr-lcssa:                            ; preds = %2425, %.lr.ph
  %.lcssa.ph = phi double [ undef, %.lr.ph ], [ %2444, %2425 ]
  %i.071.unr = phi i32 [ 1, %.lr.ph ], [ %2445, %2425 ]
  %yim1.070.unr = phi double [ %r.i95.5, %.lr.ph ], [ %2439, %2425 ]
  %yi.069.unr = phi double [ %r.i.7, %.lr.ph ], [ %2444, %2425 ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %._crit_edge, label %.epil.preheader

.epil.preheader:                                  ; preds = %._crit_edge.unr-lcssa, %.epil.preheader
  %i.071.epil = phi i32 [ %2450, %.epil.preheader ], [ %i.071.unr, %._crit_edge.unr-lcssa ]
  %yim1.070.epil = phi double [ %yi.069.epil, %.epil.preheader ], [ %yim1.070.unr, %._crit_edge.unr-lcssa ]
  %yi.069.epil = phi double [ %2449, %.epil.preheader ], [ %yi.069.unr, %._crit_edge.unr-lcssa ]
  %epil.iter = phi i32 [ %epil.iter.next, %.epil.preheader ], [ 0, %._crit_edge.unr-lcssa ]
  %2446 = sitofp i32 %i.071.epil to double
  %2447 = fmul double %2421, %2446
  %2448 = fneg double %yim1.070.epil
  %2449 = call double @llvm.fma.f64(double %2447, double %yi.069.epil, double %2448)
  %2450 = add nuw nsw i32 %i.071.epil, 1
  %epil.iter.next = add i32 %epil.iter, 1
  %epil.iter.cmp.not = icmp eq i32 %epil.iter.next, %xtraiter
  br i1 %epil.iter.cmp.not, label %._crit_edge, label %.epil.preheader, !llvm.loop !88

._crit_edge:                                      ; preds = %.epil.preheader, %._crit_edge.unr-lcssa
  %yip1.0.lcssa = phi double [ %.lcssa.ph, %._crit_edge.unr-lcssa ], [ %2449, %.epil.preheader ]
  %2451 = call double @llvm.fabs.f64(double %yip1.0.lcssa)
  %2452 = fcmp ugt double %2451, 0x7FF0000000000000
  br i1 %2452, label %2453, label %__nv_y1.exit416

2453:                                             ; preds = %._crit_edge
  br label %__nv_y1.exit416

__nv_y1.exit416:                                  ; preds = %2415, %._crit_edge, %2413, %2414, %2453, %2419, %__nv_y0.exit266
  %.0 = phi double [ %r.i239.5, %__nv_y0.exit266 ], [ %2420, %2419 ], [ 0xFFF0000000000000, %2453 ], [ %r.i.6, %2414 ], [ %r.i389.5, %2413 ], [ %yip1.0.lcssa, %._crit_edge ], [ 0xFFF8000000000000, %2415 ]
  ret double %.0
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_jn(i32 %n, double %a) local_unnamed_addr #111 {
  %q.i.i.i.i137 = alloca i32, align 4
  %q.i.i.i.i.i153 = alloca i32, align 4
  %q.i.i.i167 = alloca i32, align 4
  %q.i.i.i.i75 = alloca i32, align 4
  %q.i.i.i.i.i91 = alloca i32, align 4
  %q.i.i.i105 = alloca i32, align 4
  %q.i.i.i.i14 = alloca i32, align 4
  %q.i.i.i.i.i30 = alloca i32, align 4
  %q.i.i.i44 = alloca i32, align 4
  %q.i.i.i.i = alloca i32, align 4
  %q.i.i.i.i.i = alloca i32, align 4
  %q.i.i.i = alloca i32, align 4
  switch i32 %n, label %513 [
    i32 0, label %1
    i32 1, label %257
  ]

1:                                                ; preds = %0
  %2 = call double @llvm.fabs.f64(double %a)
  %3 = fcmp ugt double %2, 0x400FB319F277BBE5
  br i1 %3, label %24, label %4

4:                                                ; preds = %1
  %5 = fadd double %2, 0xC0033D152E971B40
  %6 = fadd double %5, 0x3CA0F539D7DA258E
  %7 = call double @llvm.fma.f64(double %6, double noundef 0xBCC0D18564C48C61, double noundef 0xBCFCF8F9A8C294BC)
  %8 = call double @llvm.fma.f64(double %7, double %6, double noundef 0x3D3FAB983CAE498B)
  %9 = call double @llvm.fma.f64(double %8, double %6, double noundef 0x3D7CD7C018579B88)
  %10 = call double @llvm.fma.f64(double %9, double %6, double noundef 0xBDBBDD2342D64FDD)
  %11 = call double @llvm.fma.f64(double %10, double %6, double noundef 0xBDF5C2D9416B1E2B)
  %12 = call double @llvm.fma.f64(double %11, double %6, double noundef 0x3E32951D73174DD5)
  %13 = call double @llvm.fma.f64(double %12, double %6, double noundef 0x3E67FF99802CAEB5)
  %14 = call double @llvm.fma.f64(double %13, double %6, double noundef 0xBEA1CCE305C4C9F7)
  %15 = call double @llvm.fma.f64(double %14, double %6, double noundef 0xBED232C77E29E1BB)
  %16 = call double @llvm.fma.f64(double %15, double %6, double noundef 0x3F06ED3B9F0EF757)
  %17 = call double @llvm.fma.f64(double %16, double %6, double noundef 0x3F315382BA096A62)
  %18 = call double @llvm.fma.f64(double %17, double %6, double noundef 0xBF61F992590D1AE4)
  %19 = call double @llvm.fma.f64(double %18, double %6, double noundef 0xBF81BB1CBE1A465F)
  %20 = call double @llvm.fma.f64(double %19, double %6, double noundef 0x3FACFAE864368D84)
  %21 = call double @llvm.fma.f64(double %20, double %6, double noundef 0x3FBBA1DEEA0294A3)
  %22 = call double @llvm.fma.f64(double %21, double %6, double noundef 0xBFE09CDB36551280)
  %23 = fmul double %6, %22
  br label %__nv_j0.exit

24:                                               ; preds = %1
  %25 = fcmp ugt double %2, 0x401C58FD1A62F5EC
  br i1 %25, label %46, label %26

26:                                               ; preds = %24
  %27 = fadd double %2, 0xC016148F5B2C2E45
  %28 = fadd double %27, 0xBC975054CD60A517
  %29 = call double @llvm.fma.f64(double %28, double noundef 0x3CBCB0A8F126B343, double noundef 0x3CF83FD1F333EB61)
  %30 = call double @llvm.fma.f64(double %29, double %28, double noundef 0xBD4100E33E3FB413)
  %31 = call double @llvm.fma.f64(double %30, double %28, double noundef 0xBD7846076D004627)
  %32 = call double @llvm.fma.f64(double %31, double %28, double noundef 0x3DBE2F1D4F90720D)
  %33 = call double @llvm.fma.f64(double %32, double %28, double noundef 0x3DF1D03B1E4A119B)
  %34 = call double @llvm.fma.f64(double %33, double %28, double noundef 0xBE341D72B1B3BCE9)
  %35 = call double @llvm.fma.f64(double %34, double %28, double noundef 0xBE62DA37CE2A9EF8)
  %36 = call double @llvm.fma.f64(double %35, double %28, double noundef 0x3EA32E6D9974F763)
  %37 = call double @llvm.fma.f64(double %36, double %28, double noundef 0x3ECAD77D744A1879)
  %38 = call double @llvm.fma.f64(double %37, double %28, double noundef 0xBF0863F481A37337)
  %39 = call double @llvm.fma.f64(double %38, double %28, double noundef 0xBF26F641F418F0F4)
  %40 = call double @llvm.fma.f64(double %39, double %28, double noundef 0x3F627E31FE9A969E)
  %41 = call double @llvm.fma.f64(double %40, double %28, double noundef 0x3F72F7FFE9025628)
  %42 = call double @llvm.fma.f64(double %41, double %28, double noundef 0xBFAB2150CB41E8BF)
  %43 = call double @llvm.fma.f64(double %42, double %28, double noundef 0xBF9F8F72E7A848DE)
  %44 = call double @llvm.fma.f64(double %43, double %28, double noundef 0x3FD5C6E60A097823)
  %45 = fmul double %28, %44
  br label %__nv_j0.exit

46:                                               ; preds = %24
  %47 = fcmp ugt double %2, 0x402471FCB6A7A8C0
  br i1 %47, label %68, label %48

48:                                               ; preds = %46
  %49 = fadd double %2, 0xC0214EB56CCCDECA
  %50 = fadd double %49, 0x3CB51970714C7C25
  %51 = call double @llvm.fma.f64(double %50, double noundef 0xBCBDB7FFCF659E24, double noundef 0xBCF4B3A71AAAC629)
  %52 = call double @llvm.fma.f64(double %51, double %50, double noundef 0x3D417EC150ECDCE7)
  %53 = call double @llvm.fma.f64(double %52, double %50, double noundef 0x3D7438F5EA1D10B2)
  %54 = call double @llvm.fma.f64(double %53, double %50, double noundef 0xBDBEDAE7EC2C9E87)
  %55 = call double @llvm.fma.f64(double %54, double %50, double noundef 0xBDECADD2C4B91F58)
  %56 = call double @llvm.fma.f64(double %55, double %50, double noundef 0x3E34582C8EE12204)
  %57 = call double @llvm.fma.f64(double %56, double %50, double noundef 0x3E5CEDA451DD20F8)
  %58 = call double @llvm.fma.f64(double %57, double %50, double noundef 0xBEA30E8CC3165E2F)
  %59 = call double @llvm.fma.f64(double %58, double %50, double noundef 0xBEC3324842BB1A2E)
  %60 = call double @llvm.fma.f64(double %59, double %50, double noundef 0x3F07800BC54FBDDB)
  %61 = call double @llvm.fma.f64(double %60, double %50, double noundef 0x3F1D79605276949A)
  %62 = call double @llvm.fma.f64(double %61, double %50, double noundef 0xBF60E0D60385A629)
  %63 = call double @llvm.fma.f64(double %62, double %50, double noundef 0xBF648E63600D82F3)
  %64 = call double @llvm.fma.f64(double %63, double %50, double noundef 0x3FA68B984EC6493A)
  %65 = call double @llvm.fma.f64(double %64, double %50, double noundef 0x3F900F7FCF183E0B)
  %66 = call double @llvm.fma.f64(double %65, double %50, double noundef 0xBFD15F7977A772D4)
  %67 = fmul double %50, %66
  br label %__nv_j0.exit

68:                                               ; preds = %46
  %69 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %70 = icmp eq i32 %69, 200
  br i1 %70, label %77, label %71

71:                                               ; preds = %68
  %72 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %73 = icmp eq i32 %72, 350
  br i1 %73, label %77, label %74

74:                                               ; preds = %71
  %75 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %76 = icmp eq i32 %75, 370
  br i1 %76, label %77, label %79

77:                                               ; preds = %74, %71, %68
  %78 = fcmp oeq double %2, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i

79:                                               ; preds = %74
  %80 = call i32 @llvm.nvvm.d2i.lo(double %2) #139
  %81 = call i32 @llvm.nvvm.d2i.hi(double %2) #139
  %82 = and i32 %81, 2147483647
  %83 = icmp eq i32 %82, 2146435072
  %84 = icmp eq i32 %80, 0
  %85 = select i1 %83, i1 %84, i1 false
  br label %__nv_isinfd.exit.i

__nv_isinfd.exit.i:                               ; preds = %79, %77
  %.07.in = phi i1 [ %78, %77 ], [ %85, %79 ]
  br i1 %.07.in, label %__nv_j0.exit, label %86

86:                                               ; preds = %__nv_isinfd.exit.i
  %87 = call double @llvm.nvvm.rcp.approx.ftz.d(double %2) #139
  %88 = fneg double %2
  %89 = call double @llvm.fma.f64(double %88, double %87, double noundef 1.000000e+00)
  %90 = call double @llvm.fma.f64(double %89, double %89, double %89)
  %91 = call double @llvm.fma.f64(double %90, double %87, double %87)
  %92 = fmul double %91, %91
  %93 = call double @llvm.fma.f64(double %92, double noundef 0xC0D115CB8C11A9DC, double noundef 0x409927467A655012)
  %94 = call double @llvm.fma.f64(double %93, double %92, double noundef 0xC05751787E247BD4)
  %95 = call double @llvm.fma.f64(double %94, double %92, double noundef 0x401704C4E5FC36B2)
  %96 = call double @llvm.fma.f64(double %95, double %92, double noundef 0xBFE15B747A2FD531)
  %97 = call double @llvm.fma.f64(double %96, double %92, double noundef 0x3FBA7FEACF6CB79B)
  %98 = call double @llvm.fma.f64(double %97, double %92, double noundef 0xBFAFFFFFEDDCF548)
  %99 = call double @llvm.fma.f64(double %98, double %92, double noundef 0x3FEFFFFFFFFFC9E5)
  %100 = call double @llvm.fma.f64(double %92, double noundef 0xC14602FE1C34685E, double noundef 0x410ECD4523B12B84)
  %101 = call double @llvm.fma.f64(double %100, double %92, double noundef 0xC0C7A2FC1972F05A)
  %102 = call double @llvm.fma.f64(double %101, double %92, double noundef 0x407EBA131F7E5BEB)
  %103 = call double @llvm.fma.f64(double %102, double %92, double noundef 0xC0373B92E6E7CC7D)
  %104 = call double @llvm.fma.f64(double %103, double %92, double noundef 0x3FFA31BEE63A2F08)
  %105 = call double @llvm.fma.f64(double %104, double %92, double noundef 0xBFCAD320104D5D05)
  %106 = call double @llvm.fma.f64(double %105, double %92, double noundef 0x3FB0AAAA9C76D07E)
  %107 = call double @llvm.fma.f64(double %106, double %92, double noundef 0xBFBFFFFFFFFDACEC)
  %108 = call double @llvm.fma.f64(double %107, double %91, double %2)
  %109 = call double @llvm.nvvm.rsqrt.approx.d(double %2) #139
  %110 = fmul double %109, 0x3FE9884533D43651
  %111 = fmul double %110, %99
  %112 = fmul double %108, 0x3FE45F306DC9C883
  %113 = call i32 @llvm.nvvm.d2i.rn(double %112) #139
  store i32 %113, i32* %q.i.i.i, align 4, !noalias !68
  %114 = sitofp i32 %113 to double
  %115 = fneg double %114
  %116 = call double @llvm.fma.f64(double %115, double noundef 0x3FF921FB54442D18, double %108)
  %117 = call double @llvm.fma.f64(double %115, double noundef 0x3C91A62633145C00, double %116)
  %118 = call double @llvm.fma.f64(double %115, double noundef 0x397B839A252049C0, double %117)
  %119 = call double @llvm.fabs.f64(double %108)
  %120 = fcmp ult double %119, 0x41E0000000000000
  br i1 %120, label %__internal_trig_reduction_kerneld.exit.i.i, label %121

121:                                              ; preds = %86
  %122 = call fastcc double @__internal_trig_reduction_slowpathd(double %108, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i) #139
  %.pre50 = load i32, i32* %q.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i

__internal_trig_reduction_kerneld.exit.i.i:       ; preds = %121, %86
  %123 = phi i32 [ %.pre50, %121 ], [ %113, %86 ]
  %t.i.i.i.0 = phi double [ %122, %121 ], [ %118, %86 ]
  %124 = and i32 %123, 3
  %125 = uitofp i32 %124 to double
  %126 = fadd double %t.i.i.i.0, 0xBFE921FB54442D18
  %127 = fmul double %125, 0x3FF921FB54442D18
  %128 = fadd double %126, %127
  %129 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %130 = icmp eq i32 %129, 350
  br i1 %130, label %134, label %131

131:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i
  %132 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %133 = icmp eq i32 %132, 370
  br i1 %133, label %134, label %193

134:                                              ; preds = %131, %__internal_trig_reduction_kerneld.exit.i.i
  %135 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %136 = icmp eq i32 %135, 200
  br i1 %136, label %143, label %137

137:                                              ; preds = %134
  %138 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %139 = icmp eq i32 %138, 350
  br i1 %139, label %143, label %140

140:                                              ; preds = %137
  %141 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %142 = icmp eq i32 %141, 370
  br i1 %142, label %143, label %146

143:                                              ; preds = %140, %137, %134
  %144 = call double @llvm.fabs.f64(double %128)
  %145 = fcmp oeq double %144, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i

146:                                              ; preds = %140
  %147 = call i32 @llvm.nvvm.d2i.lo(double %128) #139
  %148 = call i32 @llvm.nvvm.d2i.hi(double %128) #139
  %149 = and i32 %148, 2147483647
  %150 = icmp eq i32 %149, 2146435072
  %151 = icmp eq i32 %147, 0
  %152 = select i1 %150, i1 %151, i1 false
  br label %__nv_isinfd.exit.i.i.i.i

__nv_isinfd.exit.i.i.i.i:                         ; preds = %146, %143
  %.04.in = phi i1 [ %145, %143 ], [ %152, %146 ]
  br i1 %.04.in, label %153, label %155

153:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %154 = fmul double %128, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

155:                                              ; preds = %__nv_isinfd.exit.i.i.i.i
  %156 = fmul double %128, 0x3FE45F306DC9C883
  %157 = call i32 @llvm.nvvm.d2i.rn(double %156) #139
  store i32 %157, i32* %q.i.i.i.i.i, align 4, !noalias !68
  %158 = sitofp i32 %157 to double
  %159 = fneg double %158
  %160 = call double @llvm.fma.f64(double %159, double noundef 0x3FF921FB54442D18, double %128)
  %161 = call double @llvm.fma.f64(double %159, double noundef 0x3C91A62633145C00, double %160)
  %162 = call double @llvm.fma.f64(double %159, double noundef 0x397B839A252049C0, double %161)
  %163 = call double @llvm.fabs.f64(double %128)
  %164 = fcmp ult double %163, 0x41E0000000000000
  br i1 %164, label %__internal_trig_reduction_kerneld.exit.i.i.i.i, label %165

165:                                              ; preds = %155
  %166 = call fastcc double @__internal_trig_reduction_slowpathd(double %128, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i) #139
  %.pre52 = load i32, i32* %q.i.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i.i:   ; preds = %155, %165, %153
  %i.i.i.i.i.0 = phi i32 [ 0, %153 ], [ %.pre52, %165 ], [ %157, %155 ]
  %.03 = phi double [ %154, %153 ], [ %166, %165 ], [ %162, %155 ]
  %167 = fmul double %.03, %.03
  %168 = call double @llvm.fma.f64(double %167, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %169 = call double @llvm.fma.f64(double %168, double %167, double noundef 0xBE927E4F8E06E6D9)
  %170 = call double @llvm.fma.f64(double %169, double %167, double noundef 0x3EFA01A019DDBCE9)
  %171 = call double @llvm.fma.f64(double %170, double %167, double noundef 0xBF56C16C16C15D47)
  %172 = call double @llvm.fma.f64(double %171, double %167, double noundef 0x3FA5555555555551)
  %173 = call double @llvm.fma.f64(double %172, double %167, double noundef -5.000000e-01)
  %174 = call double @llvm.fma.f64(double %173, double %167, double noundef 1.000000e+00)
  %175 = call double @llvm.fma.f64(double %167, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %176 = call double @llvm.fma.f64(double %175, double %167, double noundef 0x3EC71DE369ACE392)
  %177 = call double @llvm.fma.f64(double %176, double %167, double noundef 0xBF2A01A019DB62A1)
  %178 = call double @llvm.fma.f64(double %177, double %167, double noundef 0x3F81111111110818)
  %179 = call double @llvm.fma.f64(double %178, double %167, double noundef 0xBFC5555555555554)
  %180 = call double @llvm.fma.f64(double %179, double %167, double noundef 0.000000e+00)
  %181 = call double @llvm.fma.f64(double %180, double %.03, double %.03)
  %182 = call i32 @llvm.nvvm.d2i.hi(double %181) #139
  %183 = call i32 @llvm.nvvm.d2i.lo(double %181) #139
  %184 = xor i32 %182, -2147483648
  %185 = call double @llvm.nvvm.lohi.i2d(i32 %183, i32 %184) #139
  %186 = and i32 %i.i.i.i.i.0, 1
  %.not21 = icmp eq i32 %186, 0
  %spec.select1 = select i1 %.not21, double %174, double %185
  %187 = and i32 %i.i.i.i.i.0, 2
  %.not22 = icmp eq i32 %187, 0
  br i1 %.not22, label %__internal_cos_offset.exit.i, label %188

188:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i
  %189 = call i32 @llvm.nvvm.d2i.hi(double %spec.select1) #139
  %190 = call i32 @llvm.nvvm.d2i.lo(double %spec.select1) #139
  %191 = xor i32 %189, -2147483648
  %192 = call double @llvm.nvvm.lohi.i2d(i32 %190, i32 %191) #139
  br label %__internal_cos_offset.exit.i

193:                                              ; preds = %131
  %194 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %195 = icmp eq i32 %194, 200
  br i1 %195, label %202, label %196

196:                                              ; preds = %193
  %197 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %198 = icmp eq i32 %197, 350
  br i1 %198, label %202, label %199

199:                                              ; preds = %196
  %200 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %201 = icmp eq i32 %200, 370
  br i1 %201, label %202, label %205

202:                                              ; preds = %199, %196, %193
  %203 = call double @llvm.fabs.f64(double %128)
  %204 = fcmp oeq double %203, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i

205:                                              ; preds = %199
  %206 = call i32 @llvm.nvvm.d2i.lo(double %128) #139
  %207 = call i32 @llvm.nvvm.d2i.hi(double %128) #139
  %208 = and i32 %207, 2147483647
  %209 = icmp eq i32 %208, 2146435072
  %210 = icmp eq i32 %206, 0
  %211 = select i1 %209, i1 %210, i1 false
  br label %__nv_isinfd.exit.i.i.i

__nv_isinfd.exit.i.i.i:                           ; preds = %205, %202
  %.05.in = phi i1 [ %204, %202 ], [ %211, %205 ]
  br i1 %.05.in, label %212, label %214

212:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %213 = fmul double %128, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i

214:                                              ; preds = %__nv_isinfd.exit.i.i.i
  %215 = fmul double %128, 0x3FE45F306DC9C883
  %216 = call i32 @llvm.nvvm.d2i.rn(double %215) #139
  store i32 %216, i32* %q.i.i.i.i, align 4, !noalias !68
  %217 = sitofp i32 %216 to double
  %218 = fneg double %217
  %219 = call double @llvm.fma.f64(double %218, double noundef 0x3FF921FB54442D18, double %128)
  %220 = call double @llvm.fma.f64(double %218, double noundef 0x3C91A62633145C00, double %219)
  %221 = call double @llvm.fma.f64(double %218, double noundef 0x397B839A252049C0, double %220)
  %222 = call double @llvm.fabs.f64(double %128)
  %223 = fcmp ult double %222, 0x41E0000000000000
  br i1 %223, label %__internal_trig_reduction_kerneld.exit.i.i.i, label %224

224:                                              ; preds = %214
  %225 = call fastcc double @__internal_trig_reduction_slowpathd(double %128, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i) #139
  %.pre51 = load i32, i32* %q.i.i.i.i, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i

__internal_trig_reduction_kerneld.exit.i.i.i:     ; preds = %224, %214
  %226 = phi i32 [ %.pre51, %224 ], [ %216, %214 ]
  %t.i1.i.i.i.0 = phi double [ %225, %224 ], [ %221, %214 ]
  %phi.bo17 = add i32 %226, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i

__internal_sin_cos_kerneld.exit.i.i.i:            ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i, %212
  %i.i.i.i.0 = phi i32 [ 1, %212 ], [ %phi.bo17, %__internal_trig_reduction_kerneld.exit.i.i.i ]
  %z.i.i.i.0 = phi double [ %213, %212 ], [ %t.i1.i.i.i.0, %__internal_trig_reduction_kerneld.exit.i.i.i ]
  %227 = shl i32 %i.i.i.i.0, 3
  %228 = and i32 %227, 8
  %229 = zext i32 %228 to i64
  %230 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %229
  %231 = fmul double %z.i.i.i.0, %z.i.i.i.0
  %232 = and i32 %i.i.i.i.0, 1
  %.not18 = icmp eq i32 %232, 0
  %233 = select i1 %.not18, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %234 = getelementptr inbounds double, double addrspace(1)* %230, i64 1
  %235 = load double, double addrspace(1)* %234, align 8
  %236 = call double @llvm.fma.f64(double noundef %233, double %231, double %235)
  %237 = getelementptr inbounds double, double addrspace(1)* %230, i64 2
  %238 = load double, double addrspace(1)* %237, align 8
  %239 = call double @llvm.fma.f64(double %236, double %231, double %238)
  %240 = getelementptr inbounds double, double addrspace(1)* %230, i64 3
  %241 = load double, double addrspace(1)* %240, align 8
  %242 = call double @llvm.fma.f64(double %239, double %231, double %241)
  %243 = getelementptr inbounds double, double addrspace(1)* %230, i64 4
  %244 = load double, double addrspace(1)* %243, align 8
  %245 = call double @llvm.fma.f64(double %242, double %231, double %244)
  %246 = getelementptr inbounds double, double addrspace(1)* %230, i64 5
  %247 = load double, double addrspace(1)* %246, align 8
  %248 = call double @llvm.fma.f64(double %245, double %231, double %247)
  %249 = getelementptr inbounds double, double addrspace(1)* %230, i64 6
  %250 = load double, double addrspace(1)* %249, align 8
  %251 = call double @llvm.fma.f64(double %248, double %231, double %250)
  %252 = call double @llvm.fma.f64(double %251, double %z.i.i.i.0, double %z.i.i.i.0)
  %253 = call double @llvm.fma.f64(double %251, double %231, double noundef 1.000000e+00)
  %spec.select2 = select i1 %.not18, double %252, double %253
  %254 = and i32 %i.i.i.i.0, 2
  %.not20 = icmp eq i32 %254, 0
  %255 = call double @llvm.fma.f64(double %spec.select2, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not20, double %spec.select2, double %255
  br label %__internal_cos_offset.exit.i

__internal_cos_offset.exit.i:                     ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i, %188, %__internal_sin_cos_kerneld.exit.i.i.i
  %.02 = phi double [ %.1, %__internal_sin_cos_kerneld.exit.i.i.i ], [ %192, %188 ], [ %spec.select1, %__internal_trig_reduction_kerneld.exit.i.i.i.i ]
  %256 = fmul double %111, %.02
  br label %__nv_j0.exit

257:                                              ; preds = %0
  %258 = call double @llvm.fabs.f64(double %a)
  %259 = fcmp ugt double %258, 0x400353AABAD7B784
  br i1 %259, label %277, label %260

260:                                              ; preds = %257
  %261 = call double @llvm.fma.f64(double %258, double noundef 0x3D020E4ADCDE2AD3, double noundef 0xBD4DD167A0DC3F55)
  %262 = call double @llvm.fma.f64(double %261, double %258, double noundef 0x3D5503F5A491E487)
  %263 = call double @llvm.fma.f64(double %262, double %258, double noundef 0x3DC1F29940C2403A)
  %264 = call double @llvm.fma.f64(double %263, double %258, double noundef 0x3D84CF9302EACDEF)
  %265 = call double @llvm.fma.f64(double %264, double %258, double noundef 0xBE384A53DBBCA436)
  %266 = call double @llvm.fma.f64(double %265, double %258, double noundef 0x3D9779BEE4F63BCC)
  %267 = call double @llvm.fma.f64(double %266, double %258, double noundef 0x3EA6C160E414F3F0)
  %268 = call double @llvm.fma.f64(double %267, double %258, double noundef 0x3D8F3D2F12430699)
  %269 = call double @llvm.fma.f64(double %268, double %258, double noundef 0xBF0C71C72C0CED04)
  %270 = call double @llvm.fma.f64(double %269, double %258, double noundef 0x3D659BCA506F1128)
  %271 = call double @llvm.fma.f64(double %270, double %258, double noundef 0x3F65555555506982)
  %272 = call double @llvm.fma.f64(double %271, double %258, double noundef 0x3D15BA0B425F1BFB)
  %273 = call double @llvm.fma.f64(double %272, double %258, double noundef 0xBFB0000000000065)
  %274 = call double @llvm.fma.f64(double %273, double %258, double noundef 0x3C8729A7253FB679)
  %275 = call double @llvm.fma.f64(double %274, double %258, double noundef 5.000000e-01)
  %276 = fmul double %258, %275
  br label %__nv_j1.exit

277:                                              ; preds = %257
  %278 = fcmp ugt double %258, 0x4015B1D0574614EA
  br i1 %278, label %298, label %279

279:                                              ; preds = %277
  %280 = fadd double %258, 0xC00EA75575AF6F09
  %281 = fadd double %280, 0x3CA60155A9D1B256
  %282 = call double @llvm.fma.f64(double %281, double noundef 0xBCF8D3CDBB60175E, double noundef 0x3D41011A1DF02DAD)
  %283 = call double @llvm.fma.f64(double %282, double %281, double noundef 0x3D76013AC1E5E222)
  %284 = call double @llvm.fma.f64(double %283, double %281, double noundef 0xBDBEC315D96D5F03)
  %285 = call double @llvm.fma.f64(double %284, double %281, double noundef 0xBDF03BE1B4B57207)
  %286 = call double @llvm.fma.f64(double %285, double %281, double noundef 0x3E345695F8B660F7)
  %287 = call double @llvm.fma.f64(double %286, double %281, double noundef 0x3E617069FCFCFFF4)
  %288 = call double @llvm.fma.f64(double %287, double %281, double noundef 0xBEA33825C36745EB)
  %289 = call double @llvm.fma.f64(double %288, double %281, double noundef 0xBEC9799D4F90931B)
  %290 = call double @llvm.fma.f64(double %289, double %281, double noundef 0x3F083A06E2F7DF13)
  %291 = call double @llvm.fma.f64(double %290, double %281, double noundef 0x3F26E4C2D53A7CF6)
  %292 = call double @llvm.fma.f64(double %291, double %281, double noundef 0xBF624B3409957B1C)
  %293 = call double @llvm.fma.f64(double %292, double %281, double noundef 0xBF7537544C3325DF)
  %294 = call double @llvm.fma.f64(double %293, double %281, double noundef 0x3FAB589D1DA138E2)
  %295 = call double @llvm.fma.f64(double %294, double %281, double noundef 0x3FAAE8A39F51AD13)
  %296 = call double @llvm.fma.f64(double %295, double %281, double noundef 0xBFD9C6CF582CBF7F)
  %297 = fmul double %281, %296
  br label %__nv_j1.exit

298:                                              ; preds = %277
  %299 = fcmp ugt double %258, 0x40213065E54C1AA9
  br i1 %299, label %319, label %300

300:                                              ; preds = %298
  %301 = fadd double %258, 0xC01C0FF5F3B47250
  %302 = fadd double %301, 0x3C9B226D9D243827
  %303 = call double @llvm.fma.f64(double %302, double noundef 0x3CF3EB867515FAD6, double noundef 0xBD40E8363DB649A9)
  %304 = call double @llvm.fma.f64(double %303, double %302, double noundef 0xBD73B7DD4A6608FB)
  %305 = call double @llvm.fma.f64(double %304, double %302, double noundef 0x3DBEC5E01482C750)
  %306 = call double @llvm.fma.f64(double %305, double %302, double noundef 0x3DEC62BB9E882103)
  %307 = call double @llvm.fma.f64(double %306, double %302, double noundef 0xBE34462EED732A23)
  %308 = call double @llvm.fma.f64(double %307, double %302, double noundef 0xBE5D48DCAD7DC59B)
  %309 = call double @llvm.fma.f64(double %308, double %302, double noundef 0x3EA3026DF29167E9)
  %310 = call double @llvm.fma.f64(double %309, double %302, double noundef 0x3EC4255B0119666C)
  %311 = call double @llvm.fma.f64(double %310, double %302, double noundef 0xBF0796A751B32693)
  %312 = call double @llvm.fma.f64(double %311, double %302, double noundef 0xBF207358BBDBA284)
  %313 = call double @llvm.fma.f64(double %312, double %302, double noundef 0x3F613FBC7D6927B1)
  %314 = call double @llvm.fma.f64(double %313, double %302, double noundef 0x3F69A4B292E3DD75)
  %315 = call double @llvm.fma.f64(double %314, double %302, double noundef 0xBFA80C83BDEEE4FB)
  %316 = call double @llvm.fma.f64(double %315, double %302, double noundef 0xBF95E70DC60362BF)
  %317 = call double @llvm.fma.f64(double %316, double %302, double noundef 0x3FD33518B3874E8A)
  %318 = fmul double %302, %317
  br label %__nv_j1.exit

319:                                              ; preds = %298
  %320 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %321 = icmp eq i32 %320, 200
  br i1 %321, label %328, label %322

322:                                              ; preds = %319
  %323 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %324 = icmp eq i32 %323, 350
  br i1 %324, label %328, label %325

325:                                              ; preds = %322
  %326 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %327 = icmp eq i32 %326, 370
  br i1 %327, label %328, label %330

328:                                              ; preds = %325, %322, %319
  %329 = fcmp oeq double %258, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i56

330:                                              ; preds = %325
  %331 = call i32 @llvm.nvvm.d2i.lo(double %258) #139
  %332 = call i32 @llvm.nvvm.d2i.hi(double %258) #139
  %333 = and i32 %332, 2147483647
  %334 = icmp eq i32 %333, 2146435072
  %335 = icmp eq i32 %331, 0
  %336 = select i1 %334, i1 %335, i1 false
  br label %__nv_isinfd.exit.i56

__nv_isinfd.exit.i56:                             ; preds = %330, %328
  %.014.in = phi i1 [ %329, %328 ], [ %336, %330 ]
  br i1 %.014.in, label %__nv_j1.exit, label %337

337:                                              ; preds = %__nv_isinfd.exit.i56
  %338 = call double @llvm.nvvm.rcp.approx.ftz.d(double %258) #139
  %339 = fneg double %258
  %340 = call double @llvm.fma.f64(double %339, double %338, double noundef 1.000000e+00)
  %341 = call double @llvm.fma.f64(double %340, double %340, double %340)
  %342 = call double @llvm.fma.f64(double %341, double %338, double %338)
  %343 = fmul double %342, %342
  %344 = call double @llvm.fma.f64(double %343, double noundef 0x40CD02EA3F2F6751, double noundef 0xC099C06322A3F8BE)
  %345 = call double @llvm.fma.f64(double %344, double %343, double noundef 0x405B89354DA77324)
  %346 = call double @llvm.fma.f64(double %345, double %343, double noundef 0xC01E352294653188)
  %347 = call double @llvm.fma.f64(double %346, double %343, double noundef 0x3FE9BC7DB16BD7A7)
  %348 = call double @llvm.fma.f64(double %347, double %343, double noundef 0xBFC8BFE1C3A4F741)
  %349 = call double @llvm.fma.f64(double %348, double %343, double noundef 0x3FC7FFFFF0D00BE2)
  %350 = call double @llvm.fma.f64(double %349, double %343, double noundef 0x3FF00000000068CC)
  %351 = call double @llvm.fma.f64(double %343, double noundef 0xC18DA26B212FDC9A, double noundef 0x415A30AC6857BEE0)
  %352 = call double @llvm.fma.f64(double %351, double %343, double noundef 0xC11764222AD7C910)
  %353 = call double @llvm.fma.f64(double %352, double %343, double noundef 0x40CEB02E0C306857)
  %354 = call double @llvm.fma.f64(double %353, double %343, double noundef 0xC08351859FA2B23B)
  %355 = call double @llvm.fma.f64(double %354, double %343, double noundef 0x403E65A07AF51F42)
  %356 = call double @llvm.fma.f64(double %355, double %343, double noundef 0xC002F2B817F77A57)
  %357 = call double @llvm.fma.f64(double %356, double %343, double noundef 0x3FD7BCC34DA069FD)
  %358 = call double @llvm.fma.f64(double %357, double %343, double noundef 0xBFC4FFFFF8A44463)
  %359 = call double @llvm.fma.f64(double %358, double %343, double noundef 0x3FD7FFFFFFFF5CD7)
  %360 = call double @llvm.fma.f64(double %359, double %342, double %258)
  %361 = call double @llvm.nvvm.rsqrt.approx.d(double %258) #139
  %362 = fmul double %361, 0x3FE9884533D43651
  %363 = fmul double %362, %350
  %364 = fmul double %360, 0x3FE45F306DC9C883
  %365 = call i32 @llvm.nvvm.d2i.rn(double %364) #139
  store i32 %365, i32* %q.i.i.i44, align 4, !noalias !68
  %366 = sitofp i32 %365 to double
  %367 = fneg double %366
  %368 = call double @llvm.fma.f64(double %367, double noundef 0x3FF921FB54442D18, double %360)
  %369 = call double @llvm.fma.f64(double %367, double noundef 0x3C91A62633145C00, double %368)
  %370 = call double @llvm.fma.f64(double %367, double noundef 0x397B839A252049C0, double %369)
  %371 = call double @llvm.fabs.f64(double %360)
  %372 = fcmp ult double %371, 0x41E0000000000000
  br i1 %372, label %__internal_trig_reduction_kerneld.exit.i.i57, label %373

373:                                              ; preds = %337
  %374 = call fastcc double @__internal_trig_reduction_slowpathd(double %360, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i44) #139
  %.pre = load i32, i32* %q.i.i.i44, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i57

__internal_trig_reduction_kerneld.exit.i.i57:     ; preds = %373, %337
  %375 = phi i32 [ %.pre, %373 ], [ %365, %337 ]
  %t.i.i.i43.0 = phi double [ %374, %373 ], [ %370, %337 ]
  %376 = and i32 %375, 3
  %377 = uitofp i32 %376 to double
  %378 = fadd double %t.i.i.i43.0, 0xC002D97C7F3321D2
  %379 = fmul double %377, 0x3FF921FB54442D18
  %380 = fadd double %378, %379
  %381 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %382 = icmp eq i32 %381, 350
  br i1 %382, label %386, label %383

383:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i57
  %384 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %385 = icmp eq i32 %384, 370
  br i1 %385, label %386, label %445

386:                                              ; preds = %383, %__internal_trig_reduction_kerneld.exit.i.i57
  %387 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %388 = icmp eq i32 %387, 200
  br i1 %388, label %395, label %389

389:                                              ; preds = %386
  %390 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %391 = icmp eq i32 %390, 350
  br i1 %391, label %395, label %392

392:                                              ; preds = %389
  %393 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %394 = icmp eq i32 %393, 370
  br i1 %394, label %395, label %398

395:                                              ; preds = %392, %389, %386
  %396 = call double @llvm.fabs.f64(double %380)
  %397 = fcmp oeq double %396, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i58

398:                                              ; preds = %392
  %399 = call i32 @llvm.nvvm.d2i.lo(double %380) #139
  %400 = call i32 @llvm.nvvm.d2i.hi(double %380) #139
  %401 = and i32 %400, 2147483647
  %402 = icmp eq i32 %401, 2146435072
  %403 = icmp eq i32 %399, 0
  %404 = select i1 %402, i1 %403, i1 false
  br label %__nv_isinfd.exit.i.i.i.i58

__nv_isinfd.exit.i.i.i.i58:                       ; preds = %398, %395
  %.010.in = phi i1 [ %397, %395 ], [ %404, %398 ]
  br i1 %.010.in, label %405, label %407

405:                                              ; preds = %__nv_isinfd.exit.i.i.i.i58
  %406 = fmul double %380, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i59

407:                                              ; preds = %__nv_isinfd.exit.i.i.i.i58
  %408 = fmul double %380, 0x3FE45F306DC9C883
  %409 = call i32 @llvm.nvvm.d2i.rn(double %408) #139
  store i32 %409, i32* %q.i.i.i.i.i30, align 4, !noalias !68
  %410 = sitofp i32 %409 to double
  %411 = fneg double %410
  %412 = call double @llvm.fma.f64(double %411, double noundef 0x3FF921FB54442D18, double %380)
  %413 = call double @llvm.fma.f64(double %411, double noundef 0x3C91A62633145C00, double %412)
  %414 = call double @llvm.fma.f64(double %411, double noundef 0x397B839A252049C0, double %413)
  %415 = call double @llvm.fabs.f64(double %380)
  %416 = fcmp ult double %415, 0x41E0000000000000
  br i1 %416, label %__internal_trig_reduction_kerneld.exit.i.i.i.i59, label %417

417:                                              ; preds = %407
  %418 = call fastcc double @__internal_trig_reduction_slowpathd(double %380, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i30) #139
  %.pre49 = load i32, i32* %q.i.i.i.i.i30, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i59

__internal_trig_reduction_kerneld.exit.i.i.i.i59: ; preds = %407, %417, %405
  %i.i.i.i.i37.0 = phi i32 [ 0, %405 ], [ %.pre49, %417 ], [ %409, %407 ]
  %.09 = phi double [ %406, %405 ], [ %418, %417 ], [ %414, %407 ]
  %419 = fmul double %.09, %.09
  %420 = call double @llvm.fma.f64(double %419, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %421 = call double @llvm.fma.f64(double %420, double %419, double noundef 0xBE927E4F8E06E6D9)
  %422 = call double @llvm.fma.f64(double %421, double %419, double noundef 0x3EFA01A019DDBCE9)
  %423 = call double @llvm.fma.f64(double %422, double %419, double noundef 0xBF56C16C16C15D47)
  %424 = call double @llvm.fma.f64(double %423, double %419, double noundef 0x3FA5555555555551)
  %425 = call double @llvm.fma.f64(double %424, double %419, double noundef -5.000000e-01)
  %426 = call double @llvm.fma.f64(double %425, double %419, double noundef 1.000000e+00)
  %427 = call double @llvm.fma.f64(double %419, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %428 = call double @llvm.fma.f64(double %427, double %419, double noundef 0x3EC71DE369ACE392)
  %429 = call double @llvm.fma.f64(double %428, double %419, double noundef 0xBF2A01A019DB62A1)
  %430 = call double @llvm.fma.f64(double %429, double %419, double noundef 0x3F81111111110818)
  %431 = call double @llvm.fma.f64(double %430, double %419, double noundef 0xBFC5555555555554)
  %432 = call double @llvm.fma.f64(double %431, double %419, double noundef 0.000000e+00)
  %433 = call double @llvm.fma.f64(double %432, double %.09, double %.09)
  %434 = call i32 @llvm.nvvm.d2i.hi(double %433) #139
  %435 = call i32 @llvm.nvvm.d2i.lo(double %433) #139
  %436 = xor i32 %434, -2147483648
  %437 = call double @llvm.nvvm.lohi.i2d(i32 %435, i32 %436) #139
  %438 = and i32 %i.i.i.i.i37.0, 1
  %.not15 = icmp eq i32 %438, 0
  %spec.select4 = select i1 %.not15, double %426, double %437
  %439 = and i32 %i.i.i.i.i37.0, 2
  %.not16 = icmp eq i32 %439, 0
  br i1 %.not16, label %__internal_cos_offset.exit.i64, label %440

440:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i59
  %441 = call i32 @llvm.nvvm.d2i.hi(double %spec.select4) #139
  %442 = call i32 @llvm.nvvm.d2i.lo(double %spec.select4) #139
  %443 = xor i32 %441, -2147483648
  %444 = call double @llvm.nvvm.lohi.i2d(i32 %442, i32 %443) #139
  br label %__internal_cos_offset.exit.i64

445:                                              ; preds = %383
  %446 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %447 = icmp eq i32 %446, 200
  br i1 %447, label %454, label %448

448:                                              ; preds = %445
  %449 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %450 = icmp eq i32 %449, 350
  br i1 %450, label %454, label %451

451:                                              ; preds = %448
  %452 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %453 = icmp eq i32 %452, 370
  br i1 %453, label %454, label %457

454:                                              ; preds = %451, %448, %445
  %455 = call double @llvm.fabs.f64(double %380)
  %456 = fcmp oeq double %455, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i61

457:                                              ; preds = %451
  %458 = call i32 @llvm.nvvm.d2i.lo(double %380) #139
  %459 = call i32 @llvm.nvvm.d2i.hi(double %380) #139
  %460 = and i32 %459, 2147483647
  %461 = icmp eq i32 %460, 2146435072
  %462 = icmp eq i32 %458, 0
  %463 = select i1 %461, i1 %462, i1 false
  br label %__nv_isinfd.exit.i.i.i61

__nv_isinfd.exit.i.i.i61:                         ; preds = %457, %454
  %.011.in = phi i1 [ %456, %454 ], [ %463, %457 ]
  br i1 %.011.in, label %464, label %466

464:                                              ; preds = %__nv_isinfd.exit.i.i.i61
  %465 = fmul double %380, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i63

466:                                              ; preds = %__nv_isinfd.exit.i.i.i61
  %467 = fmul double %380, 0x3FE45F306DC9C883
  %468 = call i32 @llvm.nvvm.d2i.rn(double %467) #139
  store i32 %468, i32* %q.i.i.i.i14, align 4, !noalias !68
  %469 = sitofp i32 %468 to double
  %470 = fneg double %469
  %471 = call double @llvm.fma.f64(double %470, double noundef 0x3FF921FB54442D18, double %380)
  %472 = call double @llvm.fma.f64(double %470, double noundef 0x3C91A62633145C00, double %471)
  %473 = call double @llvm.fma.f64(double %470, double noundef 0x397B839A252049C0, double %472)
  %474 = call double @llvm.fabs.f64(double %380)
  %475 = fcmp ult double %474, 0x41E0000000000000
  br i1 %475, label %__internal_trig_reduction_kerneld.exit.i.i.i62, label %476

476:                                              ; preds = %466
  %477 = call fastcc double @__internal_trig_reduction_slowpathd(double %380, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i14) #139
  %.pre48 = load i32, i32* %q.i.i.i.i14, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i62

__internal_trig_reduction_kerneld.exit.i.i.i62:   ; preds = %476, %466
  %478 = phi i32 [ %.pre48, %476 ], [ %468, %466 ]
  %t.i1.i.i.i13.0 = phi double [ %477, %476 ], [ %473, %466 ]
  %phi.bo = add i32 %478, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i63

__internal_sin_cos_kerneld.exit.i.i.i63:          ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i62, %464
  %i.i.i.i41.0 = phi i32 [ 1, %464 ], [ %phi.bo, %__internal_trig_reduction_kerneld.exit.i.i.i62 ]
  %z.i.i.i40.0 = phi double [ %465, %464 ], [ %t.i1.i.i.i13.0, %__internal_trig_reduction_kerneld.exit.i.i.i62 ]
  %479 = shl i32 %i.i.i.i41.0, 3
  %480 = and i32 %479, 8
  %481 = zext i32 %480 to i64
  %482 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %481
  %483 = fmul double %z.i.i.i40.0, %z.i.i.i40.0
  %484 = and i32 %i.i.i.i41.0, 1
  %.not = icmp eq i32 %484, 0
  %485 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %486 = getelementptr inbounds double, double addrspace(1)* %482, i64 1
  %487 = load double, double addrspace(1)* %486, align 8
  %488 = call double @llvm.fma.f64(double noundef %485, double %483, double %487)
  %489 = getelementptr inbounds double, double addrspace(1)* %482, i64 2
  %490 = load double, double addrspace(1)* %489, align 8
  %491 = call double @llvm.fma.f64(double %488, double %483, double %490)
  %492 = getelementptr inbounds double, double addrspace(1)* %482, i64 3
  %493 = load double, double addrspace(1)* %492, align 8
  %494 = call double @llvm.fma.f64(double %491, double %483, double %493)
  %495 = getelementptr inbounds double, double addrspace(1)* %482, i64 4
  %496 = load double, double addrspace(1)* %495, align 8
  %497 = call double @llvm.fma.f64(double %494, double %483, double %496)
  %498 = getelementptr inbounds double, double addrspace(1)* %482, i64 5
  %499 = load double, double addrspace(1)* %498, align 8
  %500 = call double @llvm.fma.f64(double %497, double %483, double %499)
  %501 = getelementptr inbounds double, double addrspace(1)* %482, i64 6
  %502 = load double, double addrspace(1)* %501, align 8
  %503 = call double @llvm.fma.f64(double %500, double %483, double %502)
  %504 = call double @llvm.fma.f64(double %503, double %z.i.i.i40.0, double %z.i.i.i40.0)
  %505 = call double @llvm.fma.f64(double %503, double %483, double noundef 1.000000e+00)
  %spec.select5 = select i1 %.not, double %504, double %505
  %506 = and i32 %i.i.i.i41.0, 2
  %.not14 = icmp eq i32 %506, 0
  %507 = call double @llvm.fma.f64(double %spec.select5, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.113 = select i1 %.not14, double %spec.select5, double %507
  br label %__internal_cos_offset.exit.i64

__internal_cos_offset.exit.i64:                   ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i59, %440, %__internal_sin_cos_kerneld.exit.i.i.i63
  %.08 = phi double [ %.113, %__internal_sin_cos_kerneld.exit.i.i.i63 ], [ %444, %440 ], [ %spec.select4, %__internal_trig_reduction_kerneld.exit.i.i.i.i59 ]
  %508 = fmul double %363, %.08
  br label %__nv_j1.exit

__nv_j1.exit:                                     ; preds = %279, %__internal_cos_offset.exit.i64, %__nv_isinfd.exit.i56, %300, %260
  %r.i50.3 = phi double [ %276, %260 ], [ %297, %279 ], [ %318, %300 ], [ %508, %__internal_cos_offset.exit.i64 ], [ 0.000000e+00, %__nv_isinfd.exit.i56 ]
  %509 = fcmp olt double %a, 0.000000e+00
  %510 = fneg double %r.i50.3
  %r.i50.4 = select i1 %509, double %510, double %r.i50.3
  %511 = fcmp olt double %258, 1.000000e-30
  %512 = fmul double %a, 5.000000e-01
  %r.i50.5 = select i1 %511, double %512, double %r.i50.4
  br label %__nv_j0.exit

513:                                              ; preds = %0
  %514 = icmp slt i32 %n, 0
  br i1 %514, label %__nv_j0.exit, label %515

515:                                              ; preds = %513
  %516 = call double @llvm.fabs.f64(double %a)
  %517 = sitofp i32 %n to double
  %518 = fadd double %517, -1.000000e+00
  %519 = fcmp ogt double %516, %518
  br i1 %519, label %520, label %1054

520:                                              ; preds = %515
  %521 = fdiv double 2.000000e+00, %a
  %522 = fcmp ugt double %516, 0x400353AABAD7B784
  br i1 %522, label %540, label %523

523:                                              ; preds = %520
  %524 = call double @llvm.fma.f64(double %516, double noundef 0x3D020E4ADCDE2AD3, double noundef 0xBD4DD167A0DC3F55)
  %525 = call double @llvm.fma.f64(double %524, double %516, double noundef 0x3D5503F5A491E487)
  %526 = call double @llvm.fma.f64(double %525, double %516, double noundef 0x3DC1F29940C2403A)
  %527 = call double @llvm.fma.f64(double %526, double %516, double noundef 0x3D84CF9302EACDEF)
  %528 = call double @llvm.fma.f64(double %527, double %516, double noundef 0xBE384A53DBBCA436)
  %529 = call double @llvm.fma.f64(double %528, double %516, double noundef 0x3D9779BEE4F63BCC)
  %530 = call double @llvm.fma.f64(double %529, double %516, double noundef 0x3EA6C160E414F3F0)
  %531 = call double @llvm.fma.f64(double %530, double %516, double noundef 0x3D8F3D2F12430699)
  %532 = call double @llvm.fma.f64(double %531, double %516, double noundef 0xBF0C71C72C0CED04)
  %533 = call double @llvm.fma.f64(double %532, double %516, double noundef 0x3D659BCA506F1128)
  %534 = call double @llvm.fma.f64(double %533, double %516, double noundef 0x3F65555555506982)
  %535 = call double @llvm.fma.f64(double %534, double %516, double noundef 0x3D15BA0B425F1BFB)
  %536 = call double @llvm.fma.f64(double %535, double %516, double noundef 0xBFB0000000000065)
  %537 = call double @llvm.fma.f64(double %536, double %516, double noundef 0x3C8729A7253FB679)
  %538 = call double @llvm.fma.f64(double %537, double %516, double noundef 5.000000e-01)
  %539 = fmul double %516, %538
  br label %__nv_j1.exit126

540:                                              ; preds = %520
  %541 = fcmp ugt double %516, 0x4015B1D0574614EA
  br i1 %541, label %561, label %542

542:                                              ; preds = %540
  %543 = fadd double %516, 0xC00EA75575AF6F09
  %544 = fadd double %543, 0x3CA60155A9D1B256
  %545 = call double @llvm.fma.f64(double %544, double noundef 0xBCF8D3CDBB60175E, double noundef 0x3D41011A1DF02DAD)
  %546 = call double @llvm.fma.f64(double %545, double %544, double noundef 0x3D76013AC1E5E222)
  %547 = call double @llvm.fma.f64(double %546, double %544, double noundef 0xBDBEC315D96D5F03)
  %548 = call double @llvm.fma.f64(double %547, double %544, double noundef 0xBDF03BE1B4B57207)
  %549 = call double @llvm.fma.f64(double %548, double %544, double noundef 0x3E345695F8B660F7)
  %550 = call double @llvm.fma.f64(double %549, double %544, double noundef 0x3E617069FCFCFFF4)
  %551 = call double @llvm.fma.f64(double %550, double %544, double noundef 0xBEA33825C36745EB)
  %552 = call double @llvm.fma.f64(double %551, double %544, double noundef 0xBEC9799D4F90931B)
  %553 = call double @llvm.fma.f64(double %552, double %544, double noundef 0x3F083A06E2F7DF13)
  %554 = call double @llvm.fma.f64(double %553, double %544, double noundef 0x3F26E4C2D53A7CF6)
  %555 = call double @llvm.fma.f64(double %554, double %544, double noundef 0xBF624B3409957B1C)
  %556 = call double @llvm.fma.f64(double %555, double %544, double noundef 0xBF7537544C3325DF)
  %557 = call double @llvm.fma.f64(double %556, double %544, double noundef 0x3FAB589D1DA138E2)
  %558 = call double @llvm.fma.f64(double %557, double %544, double noundef 0x3FAAE8A39F51AD13)
  %559 = call double @llvm.fma.f64(double %558, double %544, double noundef 0xBFD9C6CF582CBF7F)
  %560 = fmul double %544, %559
  br label %__nv_j1.exit126

561:                                              ; preds = %540
  %562 = fcmp ugt double %516, 0x40213065E54C1AA9
  br i1 %562, label %582, label %563

563:                                              ; preds = %561
  %564 = fadd double %516, 0xC01C0FF5F3B47250
  %565 = fadd double %564, 0x3C9B226D9D243827
  %566 = call double @llvm.fma.f64(double %565, double noundef 0x3CF3EB867515FAD6, double noundef 0xBD40E8363DB649A9)
  %567 = call double @llvm.fma.f64(double %566, double %565, double noundef 0xBD73B7DD4A6608FB)
  %568 = call double @llvm.fma.f64(double %567, double %565, double noundef 0x3DBEC5E01482C750)
  %569 = call double @llvm.fma.f64(double %568, double %565, double noundef 0x3DEC62BB9E882103)
  %570 = call double @llvm.fma.f64(double %569, double %565, double noundef 0xBE34462EED732A23)
  %571 = call double @llvm.fma.f64(double %570, double %565, double noundef 0xBE5D48DCAD7DC59B)
  %572 = call double @llvm.fma.f64(double %571, double %565, double noundef 0x3EA3026DF29167E9)
  %573 = call double @llvm.fma.f64(double %572, double %565, double noundef 0x3EC4255B0119666C)
  %574 = call double @llvm.fma.f64(double %573, double %565, double noundef 0xBF0796A751B32693)
  %575 = call double @llvm.fma.f64(double %574, double %565, double noundef 0xBF207358BBDBA284)
  %576 = call double @llvm.fma.f64(double %575, double %565, double noundef 0x3F613FBC7D6927B1)
  %577 = call double @llvm.fma.f64(double %576, double %565, double noundef 0x3F69A4B292E3DD75)
  %578 = call double @llvm.fma.f64(double %577, double %565, double noundef 0xBFA80C83BDEEE4FB)
  %579 = call double @llvm.fma.f64(double %578, double %565, double noundef 0xBF95E70DC60362BF)
  %580 = call double @llvm.fma.f64(double %579, double %565, double noundef 0x3FD33518B3874E8A)
  %581 = fmul double %565, %580
  br label %__nv_j1.exit126

582:                                              ; preds = %561
  %583 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %584 = icmp eq i32 %583, 200
  br i1 %584, label %591, label %585

585:                                              ; preds = %582
  %586 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %587 = icmp eq i32 %586, 350
  br i1 %587, label %591, label %588

588:                                              ; preds = %585
  %589 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %590 = icmp eq i32 %589, 370
  br i1 %590, label %591, label %593

591:                                              ; preds = %588, %585, %582
  %592 = fcmp oeq double %516, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i117

593:                                              ; preds = %588
  %594 = call i32 @llvm.nvvm.d2i.lo(double %516) #139
  %595 = call i32 @llvm.nvvm.d2i.hi(double %516) #139
  %596 = and i32 %595, 2147483647
  %597 = icmp eq i32 %596, 2146435072
  %598 = icmp eq i32 %594, 0
  %599 = select i1 %597, i1 %598, i1 false
  br label %__nv_isinfd.exit.i117

__nv_isinfd.exit.i117:                            ; preds = %593, %591
  %.021.in = phi i1 [ %592, %591 ], [ %599, %593 ]
  br i1 %.021.in, label %__nv_j1.exit126, label %600

600:                                              ; preds = %__nv_isinfd.exit.i117
  %601 = call double @llvm.nvvm.rcp.approx.ftz.d(double %516) #139
  %602 = fneg double %516
  %603 = call double @llvm.fma.f64(double %602, double %601, double noundef 1.000000e+00)
  %604 = call double @llvm.fma.f64(double %603, double %603, double %603)
  %605 = call double @llvm.fma.f64(double %604, double %601, double %601)
  %606 = fmul double %605, %605
  %607 = call double @llvm.fma.f64(double %606, double noundef 0x40CD02EA3F2F6751, double noundef 0xC099C06322A3F8BE)
  %608 = call double @llvm.fma.f64(double %607, double %606, double noundef 0x405B89354DA77324)
  %609 = call double @llvm.fma.f64(double %608, double %606, double noundef 0xC01E352294653188)
  %610 = call double @llvm.fma.f64(double %609, double %606, double noundef 0x3FE9BC7DB16BD7A7)
  %611 = call double @llvm.fma.f64(double %610, double %606, double noundef 0xBFC8BFE1C3A4F741)
  %612 = call double @llvm.fma.f64(double %611, double %606, double noundef 0x3FC7FFFFF0D00BE2)
  %613 = call double @llvm.fma.f64(double %612, double %606, double noundef 0x3FF00000000068CC)
  %614 = call double @llvm.fma.f64(double %606, double noundef 0xC18DA26B212FDC9A, double noundef 0x415A30AC6857BEE0)
  %615 = call double @llvm.fma.f64(double %614, double %606, double noundef 0xC11764222AD7C910)
  %616 = call double @llvm.fma.f64(double %615, double %606, double noundef 0x40CEB02E0C306857)
  %617 = call double @llvm.fma.f64(double %616, double %606, double noundef 0xC08351859FA2B23B)
  %618 = call double @llvm.fma.f64(double %617, double %606, double noundef 0x403E65A07AF51F42)
  %619 = call double @llvm.fma.f64(double %618, double %606, double noundef 0xC002F2B817F77A57)
  %620 = call double @llvm.fma.f64(double %619, double %606, double noundef 0x3FD7BCC34DA069FD)
  %621 = call double @llvm.fma.f64(double %620, double %606, double noundef 0xBFC4FFFFF8A44463)
  %622 = call double @llvm.fma.f64(double %621, double %606, double noundef 0x3FD7FFFFFFFF5CD7)
  %623 = call double @llvm.fma.f64(double %622, double %605, double %516)
  %624 = call double @llvm.nvvm.rsqrt.approx.d(double %516) #139
  %625 = fmul double %624, 0x3FE9884533D43651
  %626 = fmul double %625, %613
  %627 = fmul double %623, 0x3FE45F306DC9C883
  %628 = call i32 @llvm.nvvm.d2i.rn(double %627) #139
  store i32 %628, i32* %q.i.i.i105, align 4, !noalias !68
  %629 = sitofp i32 %628 to double
  %630 = fneg double %629
  %631 = call double @llvm.fma.f64(double %630, double noundef 0x3FF921FB54442D18, double %623)
  %632 = call double @llvm.fma.f64(double %630, double noundef 0x3C91A62633145C00, double %631)
  %633 = call double @llvm.fma.f64(double %630, double noundef 0x397B839A252049C0, double %632)
  %634 = call double @llvm.fabs.f64(double %623)
  %635 = fcmp ult double %634, 0x41E0000000000000
  br i1 %635, label %__internal_trig_reduction_kerneld.exit.i.i118, label %636

636:                                              ; preds = %600
  %637 = call fastcc double @__internal_trig_reduction_slowpathd(double %623, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i105) #139
  %.pre53 = load i32, i32* %q.i.i.i105, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i118

__internal_trig_reduction_kerneld.exit.i.i118:    ; preds = %636, %600
  %638 = phi i32 [ %.pre53, %636 ], [ %628, %600 ]
  %t.i.i.i104.0 = phi double [ %637, %636 ], [ %633, %600 ]
  %639 = and i32 %638, 3
  %640 = uitofp i32 %639 to double
  %641 = fadd double %t.i.i.i104.0, 0xC002D97C7F3321D2
  %642 = fmul double %640, 0x3FF921FB54442D18
  %643 = fadd double %641, %642
  %644 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %645 = icmp eq i32 %644, 350
  br i1 %645, label %649, label %646

646:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i118
  %647 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %648 = icmp eq i32 %647, 370
  br i1 %648, label %649, label %708

649:                                              ; preds = %646, %__internal_trig_reduction_kerneld.exit.i.i118
  %650 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %651 = icmp eq i32 %650, 200
  br i1 %651, label %658, label %652

652:                                              ; preds = %649
  %653 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %654 = icmp eq i32 %653, 350
  br i1 %654, label %658, label %655

655:                                              ; preds = %652
  %656 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %657 = icmp eq i32 %656, 370
  br i1 %657, label %658, label %661

658:                                              ; preds = %655, %652, %649
  %659 = call double @llvm.fabs.f64(double %643)
  %660 = fcmp oeq double %659, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i119

661:                                              ; preds = %655
  %662 = call i32 @llvm.nvvm.d2i.lo(double %643) #139
  %663 = call i32 @llvm.nvvm.d2i.hi(double %643) #139
  %664 = and i32 %663, 2147483647
  %665 = icmp eq i32 %664, 2146435072
  %666 = icmp eq i32 %662, 0
  %667 = select i1 %665, i1 %666, i1 false
  br label %__nv_isinfd.exit.i.i.i.i119

__nv_isinfd.exit.i.i.i.i119:                      ; preds = %661, %658
  %.017.in = phi i1 [ %660, %658 ], [ %667, %661 ]
  br i1 %.017.in, label %668, label %670

668:                                              ; preds = %__nv_isinfd.exit.i.i.i.i119
  %669 = fmul double %643, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i120

670:                                              ; preds = %__nv_isinfd.exit.i.i.i.i119
  %671 = fmul double %643, 0x3FE45F306DC9C883
  %672 = call i32 @llvm.nvvm.d2i.rn(double %671) #139
  store i32 %672, i32* %q.i.i.i.i.i91, align 4, !noalias !68
  %673 = sitofp i32 %672 to double
  %674 = fneg double %673
  %675 = call double @llvm.fma.f64(double %674, double noundef 0x3FF921FB54442D18, double %643)
  %676 = call double @llvm.fma.f64(double %674, double noundef 0x3C91A62633145C00, double %675)
  %677 = call double @llvm.fma.f64(double %674, double noundef 0x397B839A252049C0, double %676)
  %678 = call double @llvm.fabs.f64(double %643)
  %679 = fcmp ult double %678, 0x41E0000000000000
  br i1 %679, label %__internal_trig_reduction_kerneld.exit.i.i.i.i120, label %680

680:                                              ; preds = %670
  %681 = call fastcc double @__internal_trig_reduction_slowpathd(double %643, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i91) #139
  %.pre55 = load i32, i32* %q.i.i.i.i.i91, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i120

__internal_trig_reduction_kerneld.exit.i.i.i.i120: ; preds = %670, %680, %668
  %i.i.i.i.i98.0 = phi i32 [ 0, %668 ], [ %.pre55, %680 ], [ %672, %670 ]
  %.016 = phi double [ %669, %668 ], [ %681, %680 ], [ %677, %670 ]
  %682 = fmul double %.016, %.016
  %683 = call double @llvm.fma.f64(double %682, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %684 = call double @llvm.fma.f64(double %683, double %682, double noundef 0xBE927E4F8E06E6D9)
  %685 = call double @llvm.fma.f64(double %684, double %682, double noundef 0x3EFA01A019DDBCE9)
  %686 = call double @llvm.fma.f64(double %685, double %682, double noundef 0xBF56C16C16C15D47)
  %687 = call double @llvm.fma.f64(double %686, double %682, double noundef 0x3FA5555555555551)
  %688 = call double @llvm.fma.f64(double %687, double %682, double noundef -5.000000e-01)
  %689 = call double @llvm.fma.f64(double %688, double %682, double noundef 1.000000e+00)
  %690 = call double @llvm.fma.f64(double %682, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %691 = call double @llvm.fma.f64(double %690, double %682, double noundef 0x3EC71DE369ACE392)
  %692 = call double @llvm.fma.f64(double %691, double %682, double noundef 0xBF2A01A019DB62A1)
  %693 = call double @llvm.fma.f64(double %692, double %682, double noundef 0x3F81111111110818)
  %694 = call double @llvm.fma.f64(double %693, double %682, double noundef 0xBFC5555555555554)
  %695 = call double @llvm.fma.f64(double %694, double %682, double noundef 0.000000e+00)
  %696 = call double @llvm.fma.f64(double %695, double %.016, double %.016)
  %697 = call i32 @llvm.nvvm.d2i.hi(double %696) #139
  %698 = call i32 @llvm.nvvm.d2i.lo(double %696) #139
  %699 = xor i32 %697, -2147483648
  %700 = call double @llvm.nvvm.lohi.i2d(i32 %698, i32 %699) #139
  %701 = and i32 %i.i.i.i.i98.0, 1
  %.not34 = icmp eq i32 %701, 0
  %spec.select7 = select i1 %.not34, double %689, double %700
  %702 = and i32 %i.i.i.i.i98.0, 2
  %.not35 = icmp eq i32 %702, 0
  br i1 %.not35, label %__internal_cos_offset.exit.i125, label %703

703:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i120
  %704 = call i32 @llvm.nvvm.d2i.hi(double %spec.select7) #139
  %705 = call i32 @llvm.nvvm.d2i.lo(double %spec.select7) #139
  %706 = xor i32 %704, -2147483648
  %707 = call double @llvm.nvvm.lohi.i2d(i32 %705, i32 %706) #139
  br label %__internal_cos_offset.exit.i125

708:                                              ; preds = %646
  %709 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %710 = icmp eq i32 %709, 200
  br i1 %710, label %717, label %711

711:                                              ; preds = %708
  %712 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %713 = icmp eq i32 %712, 350
  br i1 %713, label %717, label %714

714:                                              ; preds = %711
  %715 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %716 = icmp eq i32 %715, 370
  br i1 %716, label %717, label %720

717:                                              ; preds = %714, %711, %708
  %718 = call double @llvm.fabs.f64(double %643)
  %719 = fcmp oeq double %718, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i122

720:                                              ; preds = %714
  %721 = call i32 @llvm.nvvm.d2i.lo(double %643) #139
  %722 = call i32 @llvm.nvvm.d2i.hi(double %643) #139
  %723 = and i32 %722, 2147483647
  %724 = icmp eq i32 %723, 2146435072
  %725 = icmp eq i32 %721, 0
  %726 = select i1 %724, i1 %725, i1 false
  br label %__nv_isinfd.exit.i.i.i122

__nv_isinfd.exit.i.i.i122:                        ; preds = %720, %717
  %.018.in = phi i1 [ %719, %717 ], [ %726, %720 ]
  br i1 %.018.in, label %727, label %729

727:                                              ; preds = %__nv_isinfd.exit.i.i.i122
  %728 = fmul double %643, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i124

729:                                              ; preds = %__nv_isinfd.exit.i.i.i122
  %730 = fmul double %643, 0x3FE45F306DC9C883
  %731 = call i32 @llvm.nvvm.d2i.rn(double %730) #139
  store i32 %731, i32* %q.i.i.i.i75, align 4, !noalias !68
  %732 = sitofp i32 %731 to double
  %733 = fneg double %732
  %734 = call double @llvm.fma.f64(double %733, double noundef 0x3FF921FB54442D18, double %643)
  %735 = call double @llvm.fma.f64(double %733, double noundef 0x3C91A62633145C00, double %734)
  %736 = call double @llvm.fma.f64(double %733, double noundef 0x397B839A252049C0, double %735)
  %737 = call double @llvm.fabs.f64(double %643)
  %738 = fcmp ult double %737, 0x41E0000000000000
  br i1 %738, label %__internal_sin_cos_kerneld.exit.i.i.i124, label %739

739:                                              ; preds = %729
  %740 = call fastcc double @__internal_trig_reduction_slowpathd(double %643, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i75) #139
  %.pre54 = load i32, i32* %q.i.i.i.i75, align 4
  br label %__internal_sin_cos_kerneld.exit.i.i.i124

__internal_sin_cos_kerneld.exit.i.i.i124:         ; preds = %729, %739, %727
  %i.i.i.i102.0 = phi i32 [ 0, %727 ], [ %.pre54, %739 ], [ %731, %729 ]
  %z.i.i.i101.0 = phi double [ %728, %727 ], [ %740, %739 ], [ %736, %729 ]
  %741 = add nsw i32 %i.i.i.i102.0, 1
  %742 = shl i32 %741, 3
  %743 = and i32 %742, 8
  %744 = zext i32 %743 to i64
  %745 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %744
  %746 = fmul double %z.i.i.i101.0, %z.i.i.i101.0
  %747 = and i32 %741, 1
  %.not25 = icmp eq i32 %747, 0
  %748 = select i1 %.not25, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %749 = getelementptr inbounds double, double addrspace(1)* %745, i64 1
  %750 = load double, double addrspace(1)* %749, align 8
  %751 = call double @llvm.fma.f64(double noundef %748, double %746, double %750)
  %752 = getelementptr inbounds double, double addrspace(1)* %745, i64 2
  %753 = load double, double addrspace(1)* %752, align 8
  %754 = call double @llvm.fma.f64(double %751, double %746, double %753)
  %755 = getelementptr inbounds double, double addrspace(1)* %745, i64 3
  %756 = load double, double addrspace(1)* %755, align 8
  %757 = call double @llvm.fma.f64(double %754, double %746, double %756)
  %758 = getelementptr inbounds double, double addrspace(1)* %745, i64 4
  %759 = load double, double addrspace(1)* %758, align 8
  %760 = call double @llvm.fma.f64(double %757, double %746, double %759)
  %761 = getelementptr inbounds double, double addrspace(1)* %745, i64 5
  %762 = load double, double addrspace(1)* %761, align 8
  %763 = call double @llvm.fma.f64(double %760, double %746, double %762)
  %764 = getelementptr inbounds double, double addrspace(1)* %745, i64 6
  %765 = load double, double addrspace(1)* %764, align 8
  %766 = call double @llvm.fma.f64(double %763, double %746, double %765)
  %767 = call double @llvm.fma.f64(double %766, double %z.i.i.i101.0, double %z.i.i.i101.0)
  %768 = call double @llvm.fma.f64(double %766, double %746, double noundef 1.000000e+00)
  %spec.select8 = select i1 %.not25, double %767, double %768
  %769 = and i32 %741, 2
  %.not27 = icmp eq i32 %769, 0
  %770 = call double @llvm.fma.f64(double %spec.select8, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.120 = select i1 %.not27, double %spec.select8, double %770
  br label %__internal_cos_offset.exit.i125

__internal_cos_offset.exit.i125:                  ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i120, %703, %__internal_sin_cos_kerneld.exit.i.i.i124
  %.015 = phi double [ %.120, %__internal_sin_cos_kerneld.exit.i.i.i124 ], [ %707, %703 ], [ %spec.select7, %__internal_trig_reduction_kerneld.exit.i.i.i.i120 ]
  %771 = fmul double %626, %.015
  br label %__nv_j1.exit126

__nv_j1.exit126:                                  ; preds = %542, %__internal_cos_offset.exit.i125, %__nv_isinfd.exit.i117, %563, %523
  %r.i111.3 = phi double [ %539, %523 ], [ %560, %542 ], [ %581, %563 ], [ %771, %__internal_cos_offset.exit.i125 ], [ 0.000000e+00, %__nv_isinfd.exit.i117 ]
  %772 = fcmp olt double %a, 0.000000e+00
  %773 = fneg double %r.i111.3
  %r.i111.4 = select i1 %772, double %773, double %r.i111.3
  %774 = fcmp olt double %516, 1.000000e-30
  %775 = fmul double %a, 5.000000e-01
  %r.i111.5 = select i1 %774, double %775, double %r.i111.4
  %776 = fcmp ugt double %516, 0x400FB319F277BBE5
  br i1 %776, label %797, label %777

777:                                              ; preds = %__nv_j1.exit126
  %778 = fadd double %516, 0xC0033D152E971B40
  %779 = fadd double %778, 0x3CA0F539D7DA258E
  %780 = call double @llvm.fma.f64(double %779, double noundef 0xBCC0D18564C48C61, double noundef 0xBCFCF8F9A8C294BC)
  %781 = call double @llvm.fma.f64(double %780, double %779, double noundef 0x3D3FAB983CAE498B)
  %782 = call double @llvm.fma.f64(double %781, double %779, double noundef 0x3D7CD7C018579B88)
  %783 = call double @llvm.fma.f64(double %782, double %779, double noundef 0xBDBBDD2342D64FDD)
  %784 = call double @llvm.fma.f64(double %783, double %779, double noundef 0xBDF5C2D9416B1E2B)
  %785 = call double @llvm.fma.f64(double %784, double %779, double noundef 0x3E32951D73174DD5)
  %786 = call double @llvm.fma.f64(double %785, double %779, double noundef 0x3E67FF99802CAEB5)
  %787 = call double @llvm.fma.f64(double %786, double %779, double noundef 0xBEA1CCE305C4C9F7)
  %788 = call double @llvm.fma.f64(double %787, double %779, double noundef 0xBED232C77E29E1BB)
  %789 = call double @llvm.fma.f64(double %788, double %779, double noundef 0x3F06ED3B9F0EF757)
  %790 = call double @llvm.fma.f64(double %789, double %779, double noundef 0x3F315382BA096A62)
  %791 = call double @llvm.fma.f64(double %790, double %779, double noundef 0xBF61F992590D1AE4)
  %792 = call double @llvm.fma.f64(double %791, double %779, double noundef 0xBF81BB1CBE1A465F)
  %793 = call double @llvm.fma.f64(double %792, double %779, double noundef 0x3FACFAE864368D84)
  %794 = call double @llvm.fma.f64(double %793, double %779, double noundef 0x3FBBA1DEEA0294A3)
  %795 = call double @llvm.fma.f64(double %794, double %779, double noundef 0xBFE09CDB36551280)
  %796 = fmul double %779, %795
  br label %.lr.ph

797:                                              ; preds = %__nv_j1.exit126
  %798 = fcmp ugt double %516, 0x401C58FD1A62F5EC
  br i1 %798, label %819, label %799

799:                                              ; preds = %797
  %800 = fadd double %516, 0xC016148F5B2C2E45
  %801 = fadd double %800, 0xBC975054CD60A517
  %802 = call double @llvm.fma.f64(double %801, double noundef 0x3CBCB0A8F126B343, double noundef 0x3CF83FD1F333EB61)
  %803 = call double @llvm.fma.f64(double %802, double %801, double noundef 0xBD4100E33E3FB413)
  %804 = call double @llvm.fma.f64(double %803, double %801, double noundef 0xBD7846076D004627)
  %805 = call double @llvm.fma.f64(double %804, double %801, double noundef 0x3DBE2F1D4F90720D)
  %806 = call double @llvm.fma.f64(double %805, double %801, double noundef 0x3DF1D03B1E4A119B)
  %807 = call double @llvm.fma.f64(double %806, double %801, double noundef 0xBE341D72B1B3BCE9)
  %808 = call double @llvm.fma.f64(double %807, double %801, double noundef 0xBE62DA37CE2A9EF8)
  %809 = call double @llvm.fma.f64(double %808, double %801, double noundef 0x3EA32E6D9974F763)
  %810 = call double @llvm.fma.f64(double %809, double %801, double noundef 0x3ECAD77D744A1879)
  %811 = call double @llvm.fma.f64(double %810, double %801, double noundef 0xBF0863F481A37337)
  %812 = call double @llvm.fma.f64(double %811, double %801, double noundef 0xBF26F641F418F0F4)
  %813 = call double @llvm.fma.f64(double %812, double %801, double noundef 0x3F627E31FE9A969E)
  %814 = call double @llvm.fma.f64(double %813, double %801, double noundef 0x3F72F7FFE9025628)
  %815 = call double @llvm.fma.f64(double %814, double %801, double noundef 0xBFAB2150CB41E8BF)
  %816 = call double @llvm.fma.f64(double %815, double %801, double noundef 0xBF9F8F72E7A848DE)
  %817 = call double @llvm.fma.f64(double %816, double %801, double noundef 0x3FD5C6E60A097823)
  %818 = fmul double %801, %817
  br label %.lr.ph

819:                                              ; preds = %797
  %820 = fcmp ugt double %516, 0x402471FCB6A7A8C0
  br i1 %820, label %841, label %821

821:                                              ; preds = %819
  %822 = fadd double %516, 0xC0214EB56CCCDECA
  %823 = fadd double %822, 0x3CB51970714C7C25
  %824 = call double @llvm.fma.f64(double %823, double noundef 0xBCBDB7FFCF659E24, double noundef 0xBCF4B3A71AAAC629)
  %825 = call double @llvm.fma.f64(double %824, double %823, double noundef 0x3D417EC150ECDCE7)
  %826 = call double @llvm.fma.f64(double %825, double %823, double noundef 0x3D7438F5EA1D10B2)
  %827 = call double @llvm.fma.f64(double %826, double %823, double noundef 0xBDBEDAE7EC2C9E87)
  %828 = call double @llvm.fma.f64(double %827, double %823, double noundef 0xBDECADD2C4B91F58)
  %829 = call double @llvm.fma.f64(double %828, double %823, double noundef 0x3E34582C8EE12204)
  %830 = call double @llvm.fma.f64(double %829, double %823, double noundef 0x3E5CEDA451DD20F8)
  %831 = call double @llvm.fma.f64(double %830, double %823, double noundef 0xBEA30E8CC3165E2F)
  %832 = call double @llvm.fma.f64(double %831, double %823, double noundef 0xBEC3324842BB1A2E)
  %833 = call double @llvm.fma.f64(double %832, double %823, double noundef 0x3F07800BC54FBDDB)
  %834 = call double @llvm.fma.f64(double %833, double %823, double noundef 0x3F1D79605276949A)
  %835 = call double @llvm.fma.f64(double %834, double %823, double noundef 0xBF60E0D60385A629)
  %836 = call double @llvm.fma.f64(double %835, double %823, double noundef 0xBF648E63600D82F3)
  %837 = call double @llvm.fma.f64(double %836, double %823, double noundef 0x3FA68B984EC6493A)
  %838 = call double @llvm.fma.f64(double %837, double %823, double noundef 0x3F900F7FCF183E0B)
  %839 = call double @llvm.fma.f64(double %838, double %823, double noundef 0xBFD15F7977A772D4)
  %840 = fmul double %823, %839
  br label %.lr.ph

841:                                              ; preds = %819
  %842 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %843 = icmp eq i32 %842, 200
  br i1 %843, label %850, label %844

844:                                              ; preds = %841
  %845 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %846 = icmp eq i32 %845, 350
  br i1 %846, label %850, label %847

847:                                              ; preds = %844
  %848 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %849 = icmp eq i32 %848, 370
  br i1 %849, label %850, label %852

850:                                              ; preds = %847, %844, %841
  %851 = fcmp oeq double %516, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i179

852:                                              ; preds = %847
  %853 = call i32 @llvm.nvvm.d2i.lo(double %516) #139
  %854 = call i32 @llvm.nvvm.d2i.hi(double %516) #139
  %855 = and i32 %854, 2147483647
  %856 = icmp eq i32 %855, 2146435072
  %857 = icmp eq i32 %853, 0
  %858 = select i1 %856, i1 %857, i1 false
  br label %__nv_isinfd.exit.i179

__nv_isinfd.exit.i179:                            ; preds = %852, %850
  %.028.in = phi i1 [ %851, %850 ], [ %858, %852 ]
  br i1 %.028.in, label %.lr.ph, label %859

859:                                              ; preds = %__nv_isinfd.exit.i179
  %860 = call double @llvm.nvvm.rcp.approx.ftz.d(double %516) #139
  %861 = fneg double %516
  %862 = call double @llvm.fma.f64(double %861, double %860, double noundef 1.000000e+00)
  %863 = call double @llvm.fma.f64(double %862, double %862, double %862)
  %864 = call double @llvm.fma.f64(double %863, double %860, double %860)
  %865 = fmul double %864, %864
  %866 = call double @llvm.fma.f64(double %865, double noundef 0xC0D115CB8C11A9DC, double noundef 0x409927467A655012)
  %867 = call double @llvm.fma.f64(double %866, double %865, double noundef 0xC05751787E247BD4)
  %868 = call double @llvm.fma.f64(double %867, double %865, double noundef 0x401704C4E5FC36B2)
  %869 = call double @llvm.fma.f64(double %868, double %865, double noundef 0xBFE15B747A2FD531)
  %870 = call double @llvm.fma.f64(double %869, double %865, double noundef 0x3FBA7FEACF6CB79B)
  %871 = call double @llvm.fma.f64(double %870, double %865, double noundef 0xBFAFFFFFEDDCF548)
  %872 = call double @llvm.fma.f64(double %871, double %865, double noundef 0x3FEFFFFFFFFFC9E5)
  %873 = call double @llvm.fma.f64(double %865, double noundef 0xC14602FE1C34685E, double noundef 0x410ECD4523B12B84)
  %874 = call double @llvm.fma.f64(double %873, double %865, double noundef 0xC0C7A2FC1972F05A)
  %875 = call double @llvm.fma.f64(double %874, double %865, double noundef 0x407EBA131F7E5BEB)
  %876 = call double @llvm.fma.f64(double %875, double %865, double noundef 0xC0373B92E6E7CC7D)
  %877 = call double @llvm.fma.f64(double %876, double %865, double noundef 0x3FFA31BEE63A2F08)
  %878 = call double @llvm.fma.f64(double %877, double %865, double noundef 0xBFCAD320104D5D05)
  %879 = call double @llvm.fma.f64(double %878, double %865, double noundef 0x3FB0AAAA9C76D07E)
  %880 = call double @llvm.fma.f64(double %879, double %865, double noundef 0xBFBFFFFFFFFDACEC)
  %881 = call double @llvm.fma.f64(double %880, double %864, double %516)
  %882 = call double @llvm.nvvm.rsqrt.approx.d(double %516) #139
  %883 = fmul double %882, 0x3FE9884533D43651
  %884 = fmul double %883, %872
  %885 = fmul double %881, 0x3FE45F306DC9C883
  %886 = call i32 @llvm.nvvm.d2i.rn(double %885) #139
  store i32 %886, i32* %q.i.i.i167, align 4, !noalias !68
  %887 = sitofp i32 %886 to double
  %888 = fneg double %887
  %889 = call double @llvm.fma.f64(double %888, double noundef 0x3FF921FB54442D18, double %881)
  %890 = call double @llvm.fma.f64(double %888, double noundef 0x3C91A62633145C00, double %889)
  %891 = call double @llvm.fma.f64(double %888, double noundef 0x397B839A252049C0, double %890)
  %892 = call double @llvm.fabs.f64(double %881)
  %893 = fcmp ult double %892, 0x41E0000000000000
  br i1 %893, label %__internal_trig_reduction_kerneld.exit.i.i180, label %894

894:                                              ; preds = %859
  %895 = call fastcc double @__internal_trig_reduction_slowpathd(double %881, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i167) #139
  %.pre56 = load i32, i32* %q.i.i.i167, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i180

__internal_trig_reduction_kerneld.exit.i.i180:    ; preds = %894, %859
  %896 = phi i32 [ %.pre56, %894 ], [ %886, %859 ]
  %t.i.i.i166.0 = phi double [ %895, %894 ], [ %891, %859 ]
  %897 = and i32 %896, 3
  %898 = uitofp i32 %897 to double
  %899 = fadd double %t.i.i.i166.0, 0xBFE921FB54442D18
  %900 = fmul double %898, 0x3FF921FB54442D18
  %901 = fadd double %899, %900
  %902 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %903 = icmp eq i32 %902, 350
  br i1 %903, label %907, label %904

904:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i180
  %905 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %906 = icmp eq i32 %905, 370
  br i1 %906, label %907, label %966

907:                                              ; preds = %904, %__internal_trig_reduction_kerneld.exit.i.i180
  %908 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %909 = icmp eq i32 %908, 200
  br i1 %909, label %916, label %910

910:                                              ; preds = %907
  %911 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %912 = icmp eq i32 %911, 350
  br i1 %912, label %916, label %913

913:                                              ; preds = %910
  %914 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %915 = icmp eq i32 %914, 370
  br i1 %915, label %916, label %919

916:                                              ; preds = %913, %910, %907
  %917 = call double @llvm.fabs.f64(double %901)
  %918 = fcmp oeq double %917, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i.i181

919:                                              ; preds = %913
  %920 = call i32 @llvm.nvvm.d2i.lo(double %901) #139
  %921 = call i32 @llvm.nvvm.d2i.hi(double %901) #139
  %922 = and i32 %921, 2147483647
  %923 = icmp eq i32 %922, 2146435072
  %924 = icmp eq i32 %920, 0
  %925 = select i1 %923, i1 %924, i1 false
  br label %__nv_isinfd.exit.i.i.i.i181

__nv_isinfd.exit.i.i.i.i181:                      ; preds = %919, %916
  %.024.in = phi i1 [ %918, %916 ], [ %925, %919 ]
  br i1 %.024.in, label %926, label %928

926:                                              ; preds = %__nv_isinfd.exit.i.i.i.i181
  %927 = fmul double %901, 0.000000e+00
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i182

928:                                              ; preds = %__nv_isinfd.exit.i.i.i.i181
  %929 = fmul double %901, 0x3FE45F306DC9C883
  %930 = call i32 @llvm.nvvm.d2i.rn(double %929) #139
  store i32 %930, i32* %q.i.i.i.i.i153, align 4, !noalias !68
  %931 = sitofp i32 %930 to double
  %932 = fneg double %931
  %933 = call double @llvm.fma.f64(double %932, double noundef 0x3FF921FB54442D18, double %901)
  %934 = call double @llvm.fma.f64(double %932, double noundef 0x3C91A62633145C00, double %933)
  %935 = call double @llvm.fma.f64(double %932, double noundef 0x397B839A252049C0, double %934)
  %936 = call double @llvm.fabs.f64(double %901)
  %937 = fcmp ult double %936, 0x41E0000000000000
  br i1 %937, label %__internal_trig_reduction_kerneld.exit.i.i.i.i182, label %938

938:                                              ; preds = %928
  %939 = call fastcc double @__internal_trig_reduction_slowpathd(double %901, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i.i153) #139
  %.pre58 = load i32, i32* %q.i.i.i.i.i153, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i.i182

__internal_trig_reduction_kerneld.exit.i.i.i.i182: ; preds = %928, %938, %926
  %i.i.i.i.i160.0 = phi i32 [ 0, %926 ], [ %.pre58, %938 ], [ %930, %928 ]
  %.023 = phi double [ %927, %926 ], [ %939, %938 ], [ %935, %928 ]
  %940 = fmul double %.023, %.023
  %941 = call double @llvm.fma.f64(double %940, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %942 = call double @llvm.fma.f64(double %941, double %940, double noundef 0xBE927E4F8E06E6D9)
  %943 = call double @llvm.fma.f64(double %942, double %940, double noundef 0x3EFA01A019DDBCE9)
  %944 = call double @llvm.fma.f64(double %943, double %940, double noundef 0xBF56C16C16C15D47)
  %945 = call double @llvm.fma.f64(double %944, double %940, double noundef 0x3FA5555555555551)
  %946 = call double @llvm.fma.f64(double %945, double %940, double noundef -5.000000e-01)
  %947 = call double @llvm.fma.f64(double %946, double %940, double noundef 1.000000e+00)
  %948 = call double @llvm.fma.f64(double %940, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %949 = call double @llvm.fma.f64(double %948, double %940, double noundef 0x3EC71DE369ACE392)
  %950 = call double @llvm.fma.f64(double %949, double %940, double noundef 0xBF2A01A019DB62A1)
  %951 = call double @llvm.fma.f64(double %950, double %940, double noundef 0x3F81111111110818)
  %952 = call double @llvm.fma.f64(double %951, double %940, double noundef 0xBFC5555555555554)
  %953 = call double @llvm.fma.f64(double %952, double %940, double noundef 0.000000e+00)
  %954 = call double @llvm.fma.f64(double %953, double %.023, double %.023)
  %955 = call i32 @llvm.nvvm.d2i.hi(double %954) #139
  %956 = call i32 @llvm.nvvm.d2i.lo(double %954) #139
  %957 = xor i32 %955, -2147483648
  %958 = call double @llvm.nvvm.lohi.i2d(i32 %956, i32 %957) #139
  %959 = and i32 %i.i.i.i.i160.0, 1
  %.not32 = icmp eq i32 %959, 0
  %spec.select10 = select i1 %.not32, double %947, double %958
  %960 = and i32 %i.i.i.i.i160.0, 2
  %.not33 = icmp eq i32 %960, 0
  br i1 %.not33, label %__internal_cos_offset.exit.i187, label %961

961:                                              ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i182
  %962 = call i32 @llvm.nvvm.d2i.hi(double %spec.select10) #139
  %963 = call i32 @llvm.nvvm.d2i.lo(double %spec.select10) #139
  %964 = xor i32 %962, -2147483648
  %965 = call double @llvm.nvvm.lohi.i2d(i32 %963, i32 %964) #139
  br label %__internal_cos_offset.exit.i187

966:                                              ; preds = %904
  %967 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %968 = icmp eq i32 %967, 200
  br i1 %968, label %975, label %969

969:                                              ; preds = %966
  %970 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %971 = icmp eq i32 %970, 350
  br i1 %971, label %975, label %972

972:                                              ; preds = %969
  %973 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %974 = icmp eq i32 %973, 370
  br i1 %974, label %975, label %978

975:                                              ; preds = %972, %969, %966
  %976 = call double @llvm.fabs.f64(double %901)
  %977 = fcmp oeq double %976, 0x7FF0000000000000
  br label %__nv_isinfd.exit.i.i.i184

978:                                              ; preds = %972
  %979 = call i32 @llvm.nvvm.d2i.lo(double %901) #139
  %980 = call i32 @llvm.nvvm.d2i.hi(double %901) #139
  %981 = and i32 %980, 2147483647
  %982 = icmp eq i32 %981, 2146435072
  %983 = icmp eq i32 %979, 0
  %984 = select i1 %982, i1 %983, i1 false
  br label %__nv_isinfd.exit.i.i.i184

__nv_isinfd.exit.i.i.i184:                        ; preds = %978, %975
  %.025.in = phi i1 [ %977, %975 ], [ %984, %978 ]
  br i1 %.025.in, label %985, label %987

985:                                              ; preds = %__nv_isinfd.exit.i.i.i184
  %986 = fmul double %901, 0.000000e+00
  br label %__internal_sin_cos_kerneld.exit.i.i.i186

987:                                              ; preds = %__nv_isinfd.exit.i.i.i184
  %988 = fmul double %901, 0x3FE45F306DC9C883
  %989 = call i32 @llvm.nvvm.d2i.rn(double %988) #139
  store i32 %989, i32* %q.i.i.i.i137, align 4, !noalias !68
  %990 = sitofp i32 %989 to double
  %991 = fneg double %990
  %992 = call double @llvm.fma.f64(double %991, double noundef 0x3FF921FB54442D18, double %901)
  %993 = call double @llvm.fma.f64(double %991, double noundef 0x3C91A62633145C00, double %992)
  %994 = call double @llvm.fma.f64(double %991, double noundef 0x397B839A252049C0, double %993)
  %995 = call double @llvm.fabs.f64(double %901)
  %996 = fcmp ult double %995, 0x41E0000000000000
  br i1 %996, label %__internal_trig_reduction_kerneld.exit.i.i.i185, label %997

997:                                              ; preds = %987
  %998 = call fastcc double @__internal_trig_reduction_slowpathd(double %901, i32* noalias nocapture nofree noundef nonnull writeonly align 4 dereferenceable(4) %q.i.i.i.i137) #139
  %.pre57 = load i32, i32* %q.i.i.i.i137, align 4
  br label %__internal_trig_reduction_kerneld.exit.i.i.i185

__internal_trig_reduction_kerneld.exit.i.i.i185:  ; preds = %997, %987
  %999 = phi i32 [ %.pre57, %997 ], [ %989, %987 ]
  %t.i1.i.i.i136.0 = phi double [ %998, %997 ], [ %994, %987 ]
  %phi.bo28 = add i32 %999, 1
  br label %__internal_sin_cos_kerneld.exit.i.i.i186

__internal_sin_cos_kerneld.exit.i.i.i186:         ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i185, %985
  %i.i.i.i164.0 = phi i32 [ 1, %985 ], [ %phi.bo28, %__internal_trig_reduction_kerneld.exit.i.i.i185 ]
  %z.i.i.i163.0 = phi double [ %986, %985 ], [ %t.i1.i.i.i136.0, %__internal_trig_reduction_kerneld.exit.i.i.i185 ]
  %1000 = shl i32 %i.i.i.i164.0, 3
  %1001 = and i32 %1000, 8
  %1002 = zext i32 %1001 to i64
  %1003 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %1002
  %1004 = fmul double %z.i.i.i163.0, %z.i.i.i163.0
  %1005 = and i32 %i.i.i.i164.0, 1
  %.not29 = icmp eq i32 %1005, 0
  %1006 = select i1 %.not29, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %1007 = getelementptr inbounds double, double addrspace(1)* %1003, i64 1
  %1008 = load double, double addrspace(1)* %1007, align 8
  %1009 = call double @llvm.fma.f64(double noundef %1006, double %1004, double %1008)
  %1010 = getelementptr inbounds double, double addrspace(1)* %1003, i64 2
  %1011 = load double, double addrspace(1)* %1010, align 8
  %1012 = call double @llvm.fma.f64(double %1009, double %1004, double %1011)
  %1013 = getelementptr inbounds double, double addrspace(1)* %1003, i64 3
  %1014 = load double, double addrspace(1)* %1013, align 8
  %1015 = call double @llvm.fma.f64(double %1012, double %1004, double %1014)
  %1016 = getelementptr inbounds double, double addrspace(1)* %1003, i64 4
  %1017 = load double, double addrspace(1)* %1016, align 8
  %1018 = call double @llvm.fma.f64(double %1015, double %1004, double %1017)
  %1019 = getelementptr inbounds double, double addrspace(1)* %1003, i64 5
  %1020 = load double, double addrspace(1)* %1019, align 8
  %1021 = call double @llvm.fma.f64(double %1018, double %1004, double %1020)
  %1022 = getelementptr inbounds double, double addrspace(1)* %1003, i64 6
  %1023 = load double, double addrspace(1)* %1022, align 8
  %1024 = call double @llvm.fma.f64(double %1021, double %1004, double %1023)
  %1025 = call double @llvm.fma.f64(double %1024, double %z.i.i.i163.0, double %z.i.i.i163.0)
  %1026 = call double @llvm.fma.f64(double %1024, double %1004, double noundef 1.000000e+00)
  %spec.select11 = select i1 %.not29, double %1025, double %1026
  %1027 = and i32 %i.i.i.i164.0, 2
  %.not31 = icmp eq i32 %1027, 0
  %1028 = call double @llvm.fma.f64(double %spec.select11, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.127 = select i1 %.not31, double %spec.select11, double %1028
  br label %__internal_cos_offset.exit.i187

__internal_cos_offset.exit.i187:                  ; preds = %__internal_trig_reduction_kerneld.exit.i.i.i.i182, %961, %__internal_sin_cos_kerneld.exit.i.i.i186
  %.022 = phi double [ %.127, %__internal_sin_cos_kerneld.exit.i.i.i186 ], [ %965, %961 ], [ %spec.select10, %__internal_trig_reduction_kerneld.exit.i.i.i.i182 ]
  %1029 = fmul double %884, %.022
  br label %.lr.ph

.lr.ph:                                           ; preds = %777, %821, %__nv_isinfd.exit.i179, %__internal_cos_offset.exit.i187, %799
  %r.i173.3 = phi double [ %796, %777 ], [ %818, %799 ], [ %840, %821 ], [ %1029, %__internal_cos_offset.exit.i187 ], [ 0.000000e+00, %__nv_isinfd.exit.i179 ]
  %1030 = add nsw i32 %n, -1
  %1031 = add nsw i32 %n, -2
  %xtraiter = and i32 %1030, 3
  %1032 = icmp ult i32 %1031, 3
  br i1 %1032, label %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa, label %.lr.ph.new

.lr.ph.new:                                       ; preds = %.lr.ph
  %unroll_iter = and i32 %1030, -4
  br label %1033

1033:                                             ; preds = %1033, %.lr.ph.new
  %jim1.038 = phi double [ %r.i173.3, %.lr.ph.new ], [ %1047, %1033 ]
  %ji.037 = phi double [ %r.i111.5, %.lr.ph.new ], [ %1052, %1033 ]
  %i.036 = phi i32 [ 1, %.lr.ph.new ], [ %1053, %1033 ]
  %niter = phi i32 [ 0, %.lr.ph.new ], [ %niter.next.3, %1033 ]
  %1034 = sitofp i32 %i.036 to double
  %1035 = fmul double %521, %1034
  %1036 = fneg double %jim1.038
  %1037 = call double @llvm.fma.f64(double %1035, double %ji.037, double %1036)
  %1038 = add nuw nsw i32 %i.036, 1
  %1039 = sitofp i32 %1038 to double
  %1040 = fmul double %521, %1039
  %1041 = fneg double %ji.037
  %1042 = call double @llvm.fma.f64(double %1040, double %1037, double %1041)
  %1043 = add nuw nsw i32 %i.036, 2
  %1044 = sitofp i32 %1043 to double
  %1045 = fmul double %521, %1044
  %1046 = fneg double %1037
  %1047 = call double @llvm.fma.f64(double %1045, double %1042, double %1046)
  %1048 = add nuw nsw i32 %i.036, 3
  %1049 = sitofp i32 %1048 to double
  %1050 = fmul double %521, %1049
  %1051 = fneg double %1042
  %1052 = call double @llvm.fma.f64(double %1050, double %1047, double %1051)
  %1053 = add nuw nsw i32 %i.036, 4
  %niter.next.3 = add i32 %niter, 4
  %niter.ncmp.3.not = icmp eq i32 %niter.next.3, %unroll_iter
  br i1 %niter.ncmp.3.not, label %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa, label %1033

1054:                                             ; preds = %515
  %1055 = fdiv double 2.000000e+00, %a
  %1056 = mul nsw i32 %n, 60
  %1057 = sitofp i32 %1056 to double
  %1058 = call double @llvm.sqrt.f64(double %1057)
  %1059 = fptosi double %1058 to i32
  %1060 = add nsw i32 %1059, %n
  %1061 = and i32 %1060, -2
  %1062 = icmp sgt i32 %1061, 0
  br i1 %1062, label %__nv_fabsf.exit, label %__nv_j0.exit

__nv_fabsf.exit:                                  ; preds = %1054, %1078
  %ji1.043 = phi double [ %jim12.0, %1078 ], [ 1.000000e+00, %1054 ]
  %i.142 = phi i32 [ %1079, %1078 ], [ %1061, %1054 ]
  %sum.041 = phi double [ %sum.2, %1078 ], [ 0.000000e+00, %1054 ]
  %lambda.040 = phi double [ %spec.select12, %1078 ], [ 0.000000e+00, %1054 ]
  %jip1.139 = phi double [ %jip1.2, %1078 ], [ 0.000000e+00, %1054 ]
  %1063 = sitofp i32 %i.142 to double
  %1064 = fmul double %1055, %1063
  %1065 = fneg double %jip1.139
  %1066 = call double @llvm.fma.f64(double %1064, double %ji1.043, double %1065)
  %1067 = fptrunc double %1066 to float
  %1068 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not23 = icmp eq i32 %1068, 0
  %1069 = call float @llvm.nvvm.fabs.ftz.f(float %1067) #139
  %1070 = call float @llvm.fabs.f32(float %1067)
  %.0 = select i1 %.not23, float %1070, float %1069
  %1071 = fpext float %.0 to double
  %1072 = fcmp ogt double %1071, 1.000000e+15
  br i1 %1072, label %1073, label %1078

1073:                                             ; preds = %__nv_fabsf.exit
  %1074 = fmul double %1066, 1.000000e-15
  %1075 = fmul double %ji1.043, 1.000000e-15
  %1076 = fmul double %lambda.040, 1.000000e-15
  %1077 = fmul double %sum.041, 1.000000e-15
  br label %1078

1078:                                             ; preds = %1073, %__nv_fabsf.exit
  %jip1.2 = phi double [ %1075, %1073 ], [ %ji1.043, %__nv_fabsf.exit ]
  %lambda.1 = phi double [ %1076, %1073 ], [ %lambda.040, %__nv_fabsf.exit ]
  %sum.1 = phi double [ %1077, %1073 ], [ %sum.041, %__nv_fabsf.exit ]
  %jim12.0 = phi double [ %1074, %1073 ], [ %1066, %__nv_fabsf.exit ]
  %1079 = add nsw i32 %i.142, -1
  %1080 = icmp eq i32 %1079, %n
  %spec.select12 = select i1 %1080, double %jim12.0, double %lambda.1
  %1081 = and i32 %i.142, 1
  %.not24 = icmp eq i32 %1081, 0
  %1082 = fmul double %jim12.0, 2.000000e+00
  %1083 = select i1 %.not24, double -0.000000e+00, double %1082
  %sum.2 = fadd double %sum.1, %1083
  %1084 = icmp ugt i32 %i.142, 1
  br i1 %1084, label %__nv_fabsf.exit, label %._crit_edge

._crit_edge:                                      ; preds = %1078
  %1085 = fsub double %sum.2, %jim12.0
  %1086 = fdiv double %spec.select12, %1085
  br label %__nv_j0.exit

.__nv_j0.exit.loopexit_crit_edge.unr-lcssa:       ; preds = %1033, %.lr.ph
  %.lcssa.ph = phi double [ undef, %.lr.ph ], [ %1052, %1033 ]
  %jim1.038.unr = phi double [ %r.i173.3, %.lr.ph ], [ %1047, %1033 ]
  %ji.037.unr = phi double [ %r.i111.5, %.lr.ph ], [ %1052, %1033 ]
  %i.036.unr = phi i32 [ 1, %.lr.ph ], [ %1053, %1033 ]
  %lcmp.mod.not = icmp eq i32 %xtraiter, 0
  br i1 %lcmp.mod.not, label %__nv_j0.exit, label %.epil.preheader

.epil.preheader:                                  ; preds = %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa, %.epil.preheader
  %jim1.038.epil = phi double [ %ji.037.epil, %.epil.preheader ], [ %jim1.038.unr, %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa ]
  %ji.037.epil = phi double [ %1090, %.epil.preheader ], [ %ji.037.unr, %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa ]
  %i.036.epil = phi i32 [ %1091, %.epil.preheader ], [ %i.036.unr, %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa ]
  %epil.iter = phi i32 [ %epil.iter.next, %.epil.preheader ], [ 0, %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa ]
  %1087 = sitofp i32 %i.036.epil to double
  %1088 = fmul double %521, %1087
  %1089 = fneg double %jim1.038.epil
  %1090 = call double @llvm.fma.f64(double %1088, double %ji.037.epil, double %1089)
  %1091 = add nuw nsw i32 %i.036.epil, 1
  %epil.iter.next = add i32 %epil.iter, 1
  %epil.iter.cmp.not = icmp eq i32 %epil.iter.next, %xtraiter
  br i1 %epil.iter.cmp.not, label %__nv_j0.exit, label %.epil.preheader, !llvm.loop !89

__nv_j0.exit:                                     ; preds = %.epil.preheader, %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa, %1054, %._crit_edge, %513, %4, %48, %__nv_isinfd.exit.i, %__internal_cos_offset.exit.i, %26, %__nv_j1.exit
  %.01 = phi double [ %r.i50.5, %__nv_j1.exit ], [ %23, %4 ], [ %45, %26 ], [ %67, %48 ], [ %256, %__internal_cos_offset.exit.i ], [ 0.000000e+00, %__nv_isinfd.exit.i ], [ 0xFFF8000000000000, %513 ], [ %1086, %._crit_edge ], [ -0.000000e+00, %1054 ], [ %.lcssa.ph, %.__nv_j0.exit.loopexit_crit_edge.unr-lcssa ], [ %1090, %.epil.preheader ]
  ret double %.01
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_erf(double %a) local_unnamed_addr #112 {
__internal_mexpm1.exit:
  %0 = call double @llvm.fabs.f64(double %a) #140
  %1 = call double @llvm.fma.f64(double %0, double noundef 0xBCF0679AFBA6F279, double noundef 0x3D47088FDB46FA5F) #140
  %2 = call double @llvm.fma.f64(double %1, double %0, double noundef 0xBD8DF9F9B976A9B2) #140
  %3 = call double @llvm.fma.f64(double %2, double %0, double noundef 0x3DC7F1F5590CC332) #140
  %4 = call double @llvm.fma.f64(double %3, double %0, double noundef 0xBDFA28A3CD2D56C4) #140
  %5 = call double @llvm.fma.f64(double %4, double %0, double noundef 0x3E2485EE67835925) #140
  %6 = call double @llvm.fma.f64(double %5, double %0, double noundef 0xBE476DB45919F583) #140
  %7 = call double @llvm.fma.f64(double %6, double %0, double noundef 0x3E62D698D98C8D71) #140
  %8 = call double @llvm.fma.f64(double %7, double %0, double noundef 0xBE720A2C7155D5C6) #140
  %9 = call double @llvm.fma.f64(double %8, double %0, double noundef 0xBE41D29B37CA1397) #140
  %10 = call double @llvm.fma.f64(double %9, double %0, double noundef 0x3EA2EF6CC0F67A49) #140
  %11 = call double @llvm.fma.f64(double %10, double %0, double noundef 0xBEC102B892333B6F) #140
  %12 = call double @llvm.fma.f64(double %11, double %0, double noundef 0x3ECA30375BA9A84E) #140
  %13 = call double @llvm.fma.f64(double %12, double %0, double noundef 0x3ECAAD18DEDEA43E) #140
  %14 = call double @llvm.fma.f64(double %13, double %0, double noundef 0xBEFF05355BC5B225) #140
  %15 = call double @llvm.fma.f64(double %14, double %0, double noundef 0x3F10E37A3108BC8B) #140
  %16 = call double @llvm.fma.f64(double %15, double %0, double noundef 0x3EFB292D828E5CB2) #140
  %17 = call double @llvm.fma.f64(double %16, double %0, double noundef 0xBF4356626EBF9BFA) #140
  %18 = call double @llvm.fma.f64(double %17, double %0, double noundef 0x3F5BCA68F73D6AFC) #140
  %19 = call double @llvm.fma.f64(double %18, double %0, double noundef 0xBF2B6B69EBBC280B) #140
  %20 = call double @llvm.fma.f64(double %19, double %0, double noundef 0xBF9396685912A453) #140
  %21 = call double @llvm.fma.f64(double %20, double %0, double noundef 0x3FBA4F4E2A1ABEF8) #140
  %22 = call double @llvm.fma.f64(double %21, double %0, double noundef 0x3FE45F306DC9C8BB) #140
  %23 = call double @llvm.fma.f64(double %22, double %0, double noundef 0x3FC06EBA8214DB69) #140
  %24 = call double @llvm.fma.f64(double %23, double %0, double %0) #140
  %25 = fsub double %0, %24
  %26 = call double @llvm.fma.f64(double %23, double %0, double %25) #140
  %27 = fneg double %24
  %28 = fneg double %26
  %29 = fptrunc double %27 to float
  %30 = fmul float %29, 0x3FF7154760000000
  %31 = call float @llvm.round.f32(float %30) #140
  %32 = fpext float %31 to double
  %33 = fneg double %32
  %34 = call double @llvm.fma.f64(double %33, double noundef 0x3FE62E42FEFA39EF, double %27) #140
  %35 = call double @llvm.fma.f64(double %34, double noundef 0x3E5AE904A4741B81, double noundef 0x3E928A27F89B6999) #140
  %36 = call double @llvm.fma.f64(double %35, double %34, double noundef 0x3EC71DE715FF7E07) #140
  %37 = call double @llvm.fma.f64(double %36, double %34, double noundef 0x3EFA019A6B0AC45A) #140
  %38 = call double @llvm.fma.f64(double %37, double %34, double noundef 0x3F2A01A017EED94F) #140
  %39 = call double @llvm.fma.f64(double %38, double %34, double noundef 0x3F56C16C17F2A71B) #140
  %40 = call double @llvm.fma.f64(double %39, double %34, double noundef 0x3F811111111173C4) #140
  %41 = call double @llvm.fma.f64(double %40, double %34, double noundef 0x3FA555555555211A) #140
  %42 = call double @llvm.fma.f64(double %41, double %34, double noundef 0x3FC5555555555540) #140
  %43 = call double @llvm.fma.f64(double %42, double %34, double noundef 0x3FE0000000000005) #140
  %44 = fmul double %34, %43
  %45 = call double @llvm.fma.f64(double %44, double %34, double %28) #140
  %46 = fadd double %34, %45
  %47 = call float @llvm.nvvm.ex2.approx.ftz.f(float %31) #142
  %48 = fpext float %47 to double
  %49 = fsub double 1.000000e+00, %48
  %50 = fneg double %46
  %51 = call double @llvm.fma.f64(double %50, double %48, double %49) #140
  %52 = fcmp oge double %0, 0x4017AFB48DC96626
  %poly.0 = select i1 %52, double 1.000000e+00, double %51
  %53 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %54 = call i32 @llvm.nvvm.d2i.lo(double %poly.0) #142
  %55 = call i32 @llvm.nvvm.d2i.hi(double %poly.0) #142
  %56 = and i32 %53, -2147483648
  %57 = or i32 %55, %56
  %58 = call double @llvm.nvvm.lohi.i2d(i32 %54, i32 %57) #142
  ret double %58
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_erfinv(double %a) local_unnamed_addr #113 {
__internal_fast_icmp_abs_lt.exit:
  %0 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %1 = bitcast i32 %0 to float
  %2 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %2, 0
  %3 = call float @llvm.nvvm.fabs.ftz.f(float %1) #139
  %4 = call float @llvm.fabs.f32(float %1)
  %.0 = select i1 %.not, float %4, float %3
  %5 = fcmp olt float %.0, 1.875000e+00
  br i1 %5, label %6, label %115

6:                                                ; preds = %__internal_fast_icmp_abs_lt.exit
  %7 = fneg double %a
  %8 = call double @llvm.fma.f64(double %a, double %7, double noundef 1.000000e+00)
  %9 = call i32 @llvm.nvvm.d2i.hi(double %8) #139
  %10 = call i32 @llvm.nvvm.d2i.lo(double %8) #139
  %11 = lshr i32 %9, 20
  %12 = and i32 %11, 2046
  %13 = add nuw i32 %12, 2147482626
  %14 = call double @llvm.nvvm.lohi.i2d(i32 %13, i32 noundef 1127219200) #139
  %15 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %16 = fsub double %14, %15
  %17 = and i32 %9, -2145386497
  %18 = or i32 %17, 1071644672
  %19 = call double @llvm.nvvm.lohi.i2d(i32 %10, i32 %18) #139
  %20 = fadd double %19, -1.000000e+00
  %21 = fadd double %19, 1.000000e+00
  %22 = call double @llvm.nvvm.rcp.approx.ftz.d(double %21) #139
  %23 = fneg double %21
  %24 = call double @llvm.fma.f64(double %23, double %22, double noundef 1.000000e+00)
  %25 = call double @llvm.fma.f64(double %24, double %24, double %24)
  %26 = call double @llvm.fma.f64(double %25, double %22, double %22)
  %27 = fmul double %20, %26
  %28 = call double @llvm.fma.f64(double %27, double noundef -2.000000e+00, double %20)
  %29 = fneg double %27
  %30 = call double @llvm.fma.f64(double %29, double %20, double %28)
  %31 = call double @llvm.fma.f64(double %30, double %26, double %27)
  %32 = fmul double %31, %31
  %33 = call double @llvm.fma.f64(double %32, double noundef 0x3FB5C5C218C775C9, double noundef 0x3FA55CF59CDC5D89)
  %34 = call double @llvm.fma.f64(double %33, double %32, double noundef 0x3FAEFD18CF6EBB9C)
  %35 = call double @llvm.fma.f64(double %34, double %32, double noundef 0x3FB10682EDCB8D1B)
  %36 = call double @llvm.fma.f64(double %35, double %32, double noundef 0x3FB3B1DD3AC7FC96)
  %37 = call double @llvm.fma.f64(double %36, double %32, double noundef 0x3FB745CB459B54A6)
  %38 = call double @llvm.fma.f64(double %37, double %32, double noundef 0x3FBC71C741A0669F)
  %39 = call double @llvm.fma.f64(double %38, double %32, double noundef 0x3FC249249209112E)
  %40 = call double @llvm.fma.f64(double %39, double %32, double noundef 0x3FC99999999A06C1)
  %41 = call double @llvm.fma.f64(double %40, double %32, double noundef 0x3FD5555555555535)
  %42 = fmul double %32, %41
  %43 = call double @llvm.fma.f64(double %42, double %31, double %31)
  %44 = fadd double %43, %43
  %45 = call double @llvm.fma.f64(double %16, double noundef 0x3FE62E42FEFA39EF, double %44)
  %46 = call i32 @llvm.nvvm.d2i.hi(double %45) #139
  %47 = icmp ult i32 %46, -1072103424
  br i1 %47, label %48, label %72

48:                                               ; preds = %6
  %49 = fsub double -3.125000e+00, %45
  %50 = call double @llvm.fma.f64(double %49, double noundef 0xBBB135D2E746E627, double noundef 0xBC08DDF93324D327)
  %51 = call double @llvm.fma.f64(double %50, double %49, double noundef 0x3C37B83EEF0B7C9F)
  %52 = call double @llvm.fma.f64(double %51, double %49, double noundef 0x3C69BA72CD589B91)
  %53 = call double @llvm.fma.f64(double %52, double %49, double noundef 0xBCA33689090A6B96)
  %54 = call double @llvm.fma.f64(double %53, double %49, double noundef 0x3C782E11898132E0)
  %55 = call double @llvm.fma.f64(double %54, double %49, double noundef 0x3CFDE4ACFD9E26BA)
  %56 = call double @llvm.fma.f64(double %55, double %49, double noundef 0xBD26D33EED66C487)
  %57 = call double @llvm.fma.f64(double %56, double %49, double noundef 0xBD36F2167040D8E2)
  %58 = call double @llvm.fma.f64(double %57, double %49, double noundef 0x3D872A22C2D77E20)
  %59 = call double @llvm.fma.f64(double %58, double %49, double noundef 0xBDAC8859C4E5C0AF)
  %60 = call double @llvm.fma.f64(double %59, double %49, double noundef 0xBDCDC583D118A561)
  %61 = call double @llvm.fma.f64(double %60, double %49, double noundef 0x3E120F47CCF46B3C)
  %62 = call double @llvm.fma.f64(double %61, double %49, double noundef 0xBE31A9E38DC84D60)
  %63 = call double @llvm.fma.f64(double %62, double %49, double noundef 0xBE5F36CD6D3D46A9)
  %64 = call double @llvm.fma.f64(double %63, double %49, double noundef 0x3E9C6B4F5D03B787)
  %65 = call double @llvm.fma.f64(double %64, double %49, double noundef 0xBEB6E8A5434AE8A2)
  %66 = call double @llvm.fma.f64(double %65, double %49, double noundef 0xBEED1D1F7B8736F6)
  %67 = call double @llvm.fma.f64(double %66, double %49, double noundef 0x3F2879C2A212F024)
  %68 = call double @llvm.fma.f64(double %67, double %49, double noundef 0xBF4845769484FCA8)
  %69 = call double @llvm.fma.f64(double %68, double %49, double noundef 0xBF78B6C33114F909)
  %70 = call double @llvm.fma.f64(double %69, double %49, double noundef 0x3FCEBD80D9B13E28)
  %71 = call double @llvm.fma.f64(double %70, double %49, double noundef 0x3FFA755E7C99AE86)
  br label %120

72:                                               ; preds = %6
  %73 = fneg double %45
  %74 = call double @llvm.sqrt.f64(double %73)
  %75 = call i32 @llvm.nvvm.d2i.hi(double %74) #139
  %76 = icmp slt i32 %75, 1074790400
  br i1 %76, label %77, label %97

77:                                               ; preds = %72
  %78 = fadd double %74, -3.250000e+00
  %79 = call double @llvm.fma.f64(double %78, double noundef 0x3E23040F87DBD932, double noundef 0x3E785CBE52878635)
  %80 = call double @llvm.fma.f64(double %79, double %78, double noundef 0xBE92777453DD3955)
  %81 = call double @llvm.fma.f64(double %80, double %78, double noundef 0x3E5395ABCD554C6C)
  %82 = call double @llvm.fma.f64(double %81, double %78, double noundef 0x3EB936388A3790AD)
  %83 = call double @llvm.fma.f64(double %82, double %78, double noundef 0xBED0D5DB812B5083)
  %84 = call double @llvm.fma.f64(double %83, double %78, double noundef 0x3EC8860CD5D652F6)
  %85 = call double @llvm.fma.f64(double %84, double %78, double noundef 0x3EEA29A0CACDFB23)
  %86 = call double @llvm.fma.f64(double %85, double %78, double noundef 0xBF08CEF1F80281F2)
  %87 = call double @llvm.fma.f64(double %86, double %78, double noundef 0x3F11E684D0B9188A)
  %88 = call double @llvm.fma.f64(double %87, double %78, double noundef 0x3EF932CD54C8A222)
  %89 = call double @llvm.fma.f64(double %88, double %78, double noundef 0xBF37448A89EF8AA3)
  %90 = call double @llvm.fma.f64(double %89, double %78, double noundef 0x3F4F3CC55AD40C25)
  %91 = call double @llvm.fma.f64(double %90, double %78, double noundef 0xBF5BA924132F38B1)
  %92 = call double @llvm.fma.f64(double %91, double %78, double noundef 0x3F6468EECA533CF8)
  %93 = call double @llvm.fma.f64(double %92, double %78, double noundef 0xBF6EBADABB891BBD)
  %94 = call double @llvm.fma.f64(double %93, double %78, double noundef 0x3F75FFCFE5B76AFC)
  %95 = call double @llvm.fma.f64(double %94, double %78, double noundef 0x3FF0158A6D641D39)
  %96 = call double @llvm.fma.f64(double %95, double %78, double noundef 0x4008ABCC380D5A48)
  br label %120

97:                                               ; preds = %72
  %98 = fadd double %74, -5.000000e+00
  %99 = call double @llvm.fma.f64(double %98, double noundef 0xBDBDCEC3A7785389, double noundef 0xBDF18FEEC0E38727)
  %100 = call double @llvm.fma.f64(double %99, double %98, double noundef 0x3E19E6BF2DDA45E3)
  %101 = call double @llvm.fma.f64(double %100, double %98, double noundef 0xBE30468FB24E2F5F)
  %102 = call double @llvm.fma.f64(double %101, double %98, double noundef 0x3E405AC6A8FBA182)
  %103 = call double @llvm.fma.f64(double %102, double %98, double noundef 0xBE50102E495FB9C0)
  %104 = call double @llvm.fma.f64(double %103, double %98, double noundef 0x3E5F4C20E1334AF8)
  %105 = call double @llvm.fma.f64(double %104, double %98, double noundef 0xBE722D220FDF9C3E)
  %106 = call double @llvm.fma.f64(double %105, double %98, double noundef 0x3E8EBC8BB824CB54)
  %107 = call double @llvm.fma.f64(double %106, double %98, double noundef 0xBEB0A8D40EA372CC)
  %108 = call double @llvm.fma.f64(double %107, double %98, double noundef 0x3ED2FBD29D093D2B)
  %109 = call double @llvm.fma.f64(double %108, double %98, double noundef 0xBEF4A3497E1E0FAC)
  %110 = call double @llvm.fma.f64(double %109, double %98, double noundef 0x3F13EBF4EB00938F)
  %111 = call double @llvm.fma.f64(double %110, double %98, double noundef 0xBF2C2F36A8FC5D53)
  %112 = call double @llvm.fma.f64(double %111, double %98, double noundef 0xBF222EA5DF04047C)
  %113 = call double @llvm.fma.f64(double %112, double %98, double noundef 0x3FF02A30D1FBA0DC)
  %114 = call double @llvm.fma.f64(double %113, double %98, double noundef 0x4013664DDD1AD7FB)
  br label %120

115:                                              ; preds = %__internal_fast_icmp_abs_lt.exit
  %116 = call double @llvm.fabs.f64(double %a)
  %117 = fcmp ugt double %116, 0x7FF0000000000000
  br i1 %117, label %120, label %118

118:                                              ; preds = %115
  %119 = fcmp oeq double %116, 1.000000e+00
  %r.2 = select i1 %119, double 0x7FF0000000000000, double 0xFFF8000000000000
  br label %120

120:                                              ; preds = %118, %115, %48, %97, %77
  %r.4 = phi double [ %71, %48 ], [ %96, %77 ], [ %114, %97 ], [ %r.2, %118 ], [ %a, %115 ]
  %121 = fmul double %r.4, %a
  ret double %121
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_erfcinv(double %a) local_unnamed_addr #114 {
  %1 = fsub double 2.000000e+00, %a
  %2 = fcmp oge double %a, 0x3F4FA4D2AD8F904D
  %3 = fcmp ole double %a, 0x3FFFFC0B65AA4E0E
  %or.cond = and i1 %2, %3
  br i1 %or.cond, label %4, label %68

4:                                                ; preds = %0
  %5 = fmul double %1, %a
  %6 = call i32 @llvm.nvvm.d2i.hi(double %5) #142
  %7 = call i32 @llvm.nvvm.d2i.lo(double %5) #142
  %8 = lshr i32 %6, 20
  %9 = and i32 %8, 2046
  %10 = add nuw i32 %9, 2147482626
  %11 = call double @llvm.nvvm.lohi.i2d(i32 %10, i32 noundef 1127219200) #142
  %12 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %13 = fsub double %11, %12
  %14 = and i32 %6, -2145386497
  %15 = or i32 %14, 1071644672
  %16 = call double @llvm.nvvm.lohi.i2d(i32 %7, i32 %15) #142
  %17 = fadd double %16, -1.000000e+00
  %18 = fadd double %16, 1.000000e+00
  %19 = call double @llvm.nvvm.rcp.approx.ftz.d(double %18) #142
  %20 = fneg double %18
  %21 = call double @llvm.fma.f64(double %20, double %19, double noundef 1.000000e+00) #140
  %22 = call double @llvm.fma.f64(double %21, double %21, double %21) #140
  %23 = call double @llvm.fma.f64(double %22, double %19, double %19) #140
  %24 = fmul double %17, %23
  %25 = call double @llvm.fma.f64(double %24, double noundef -2.000000e+00, double %17) #140
  %26 = fneg double %24
  %27 = call double @llvm.fma.f64(double %26, double %17, double %25) #140
  %28 = call double @llvm.fma.f64(double %27, double %23, double %24) #140
  %29 = fmul double %28, %28
  %30 = call double @llvm.fma.f64(double %29, double noundef 0x3FB5C5C218C775C9, double noundef 0x3FA55CF59CDC5D89) #140
  %31 = call double @llvm.fma.f64(double %30, double %29, double noundef 0x3FAEFD18CF6EBB9C) #140
  %32 = call double @llvm.fma.f64(double %31, double %29, double noundef 0x3FB10682EDCB8D1B) #140
  %33 = call double @llvm.fma.f64(double %32, double %29, double noundef 0x3FB3B1DD3AC7FC96) #140
  %34 = call double @llvm.fma.f64(double %33, double %29, double noundef 0x3FB745CB459B54A6) #140
  %35 = call double @llvm.fma.f64(double %34, double %29, double noundef 0x3FBC71C741A0669F) #140
  %36 = call double @llvm.fma.f64(double %35, double %29, double noundef 0x3FC249249209112E) #140
  %37 = call double @llvm.fma.f64(double %36, double %29, double noundef 0x3FC99999999A06C1) #140
  %38 = call double @llvm.fma.f64(double %37, double %29, double noundef 0x3FD5555555555535) #140
  %39 = fmul double %29, %38
  %40 = call double @llvm.fma.f64(double %39, double %28, double %28) #140
  %41 = fadd double %40, %40
  %42 = call double @llvm.fma.f64(double %13, double noundef 0x3FE62E42FEFA39EF, double %41) #140
  %43 = fsub double -3.125000e+00, %42
  %44 = call double @llvm.fma.f64(double %43, double noundef 0xBBB135D2E746E627, double noundef 0xBC08DDF93324D327) #140
  %45 = call double @llvm.fma.f64(double %44, double %43, double noundef 0x3C37B83EEF0B7C9F) #140
  %46 = call double @llvm.fma.f64(double %45, double %43, double noundef 0x3C69BA72CD589B91) #140
  %47 = call double @llvm.fma.f64(double %46, double %43, double noundef 0xBCA33689090A6B96) #140
  %48 = call double @llvm.fma.f64(double %47, double %43, double noundef 0x3C782E11898132E0) #140
  %49 = call double @llvm.fma.f64(double %48, double %43, double noundef 0x3CFDE4ACFD9E26BA) #140
  %50 = call double @llvm.fma.f64(double %49, double %43, double noundef 0xBD26D33EED66C487) #140
  %51 = call double @llvm.fma.f64(double %50, double %43, double noundef 0xBD36F2167040D8E2) #140
  %52 = call double @llvm.fma.f64(double %51, double %43, double noundef 0x3D872A22C2D77E20) #140
  %53 = call double @llvm.fma.f64(double %52, double %43, double noundef 0xBDAC8859C4E5C0AF) #140
  %54 = call double @llvm.fma.f64(double %53, double %43, double noundef 0xBDCDC583D118A561) #140
  %55 = call double @llvm.fma.f64(double %54, double %43, double noundef 0x3E120F47CCF46B3C) #140
  %56 = call double @llvm.fma.f64(double %55, double %43, double noundef 0xBE31A9E38DC84D60) #140
  %57 = call double @llvm.fma.f64(double %56, double %43, double noundef 0xBE5F36CD6D3D46A9) #140
  %58 = call double @llvm.fma.f64(double %57, double %43, double noundef 0x3E9C6B4F5D03B787) #140
  %59 = call double @llvm.fma.f64(double %58, double %43, double noundef 0xBEB6E8A5434AE8A2) #140
  %60 = call double @llvm.fma.f64(double %59, double %43, double noundef 0xBEED1D1F7B8736F6) #140
  %61 = call double @llvm.fma.f64(double %60, double %43, double noundef 0x3F2879C2A212F024) #140
  %62 = call double @llvm.fma.f64(double %61, double %43, double noundef 0xBF4845769484FCA8) #140
  %63 = call double @llvm.fma.f64(double %62, double %43, double noundef 0xBF78B6C33114F909) #140
  %64 = call double @llvm.fma.f64(double %63, double %43, double noundef 0x3FCEBD80D9B13E28) #140
  %65 = call double @llvm.fma.f64(double %64, double %43, double noundef 0x3FFA755E7C99AE86) #140
  %66 = fneg double %a
  %67 = call double @llvm.fma.f64(double %65, double %66, double %65) #140
  br label %215

68:                                               ; preds = %0
  %69 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %70 = icmp sgt i32 %69, 1072693247
  %spec.select = select i1 %70, double %1, double %a
  %71 = call i32 @llvm.nvvm.d2i.hi(double %spec.select) #142
  %72 = bitcast i32 %71 to float
  %73 = fcmp ult float %72, 0x3D657FE5E0000000
  br i1 %73, label %136, label %74

74:                                               ; preds = %68
  %75 = call i32 @llvm.nvvm.d2i.lo(double %spec.select) #142
  %76 = lshr i32 %71, 20
  %77 = and i32 %76, 2046
  %78 = add nuw i32 %77, 2147482626
  %79 = call double @llvm.nvvm.lohi.i2d(i32 %78, i32 noundef 1127219200) #142
  %80 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %81 = fsub double %79, %80
  %82 = and i32 %71, -2145386497
  %83 = or i32 %82, 1071644672
  %84 = call double @llvm.nvvm.lohi.i2d(i32 %75, i32 %83) #142
  %85 = fadd double %84, -1.000000e+00
  %86 = fadd double %84, 1.000000e+00
  %87 = call double @llvm.nvvm.rcp.approx.ftz.d(double %86) #142
  %88 = fneg double %86
  %89 = call double @llvm.fma.f64(double %88, double %87, double noundef 1.000000e+00) #140
  %90 = call double @llvm.fma.f64(double %89, double %89, double %89) #140
  %91 = call double @llvm.fma.f64(double %90, double %87, double %87) #140
  %92 = fmul double %85, %91
  %93 = call double @llvm.fma.f64(double %92, double noundef -2.000000e+00, double %85) #140
  %94 = fneg double %92
  %95 = call double @llvm.fma.f64(double %94, double %85, double %93) #140
  %96 = call double @llvm.fma.f64(double %95, double %91, double %92) #140
  %97 = fmul double %96, %96
  %98 = call double @llvm.fma.f64(double %97, double noundef 0x3FB5C5C218C775C9, double noundef 0x3FA55CF59CDC5D89) #140
  %99 = call double @llvm.fma.f64(double %98, double %97, double noundef 0x3FAEFD18CF6EBB9C) #140
  %100 = call double @llvm.fma.f64(double %99, double %97, double noundef 0x3FB10682EDCB8D1B) #140
  %101 = call double @llvm.fma.f64(double %100, double %97, double noundef 0x3FB3B1DD3AC7FC96) #140
  %102 = call double @llvm.fma.f64(double %101, double %97, double noundef 0x3FB745CB459B54A6) #140
  %103 = call double @llvm.fma.f64(double %102, double %97, double noundef 0x3FBC71C741A0669F) #140
  %104 = call double @llvm.fma.f64(double %103, double %97, double noundef 0x3FC249249209112E) #140
  %105 = call double @llvm.fma.f64(double %104, double %97, double noundef 0x3FC99999999A06C1) #140
  %106 = call double @llvm.fma.f64(double %105, double %97, double noundef 0x3FD5555555555535) #140
  %107 = fmul double %97, %106
  %108 = call double @llvm.fma.f64(double %107, double %96, double %96) #140
  %109 = fadd double %108, %108
  %110 = call double @llvm.fma.f64(double %81, double noundef 0x3FE62E42FEFA39EF, double %109) #140
  %111 = fneg double %110
  %112 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %111) #141, !srcloc !84
  %113 = fneg double %112
  %114 = fmul double %112, %113
  %115 = call double @llvm.fma.f64(double %111, double %114, double noundef 1.000000e+00)
  %116 = call double @llvm.fma.f64(double %115, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %117 = fmul double %112, %115
  %118 = call double @llvm.fma.f64(double %116, double %117, double %112)
  %119 = call double @llvm.fma.f64(double %118, double noundef 0x3FEBE9222591AFAB, double noundef 0x4000A0E7333839AA)
  %120 = call double @llvm.fma.f64(double %119, double %118, double noundef 0x4008768CF7E57D5C)
  %121 = call double @llvm.fma.f64(double %120, double %118, double noundef 0x400B77E7E28DA583)
  %122 = call double @llvm.fma.f64(double %121, double %118, double noundef 0x3FF34F26A4F99CF9)
  %123 = call double @llvm.fma.f64(double %122, double %118, double noundef 0x3FC1F674ADB019ED)
  %124 = call double @llvm.fma.f64(double %123, double %118, double noundef 0x3F75DDAE9506431D)
  %125 = call double @llvm.fma.f64(double %124, double %118, double noundef 0x3F0ADA49AA32489C)
  %126 = fadd double %118, 0x4001E90FF51C2197
  %127 = call double @llvm.fma.f64(double %126, double %118, double noundef 0x40111EA3A7CF3820)
  %128 = call double @llvm.fma.f64(double %127, double %118, double noundef 0x4011A0E4A4749594)
  %129 = call double @llvm.fma.f64(double %128, double %118, double noundef 0x400D4E977D38C14D)
  %130 = call double @llvm.fma.f64(double %129, double %118, double noundef 0x3FF37FD567EC0D5F)
  %131 = call double @llvm.fma.f64(double %130, double %118, double noundef 0x3FC1FB9D7F676033)
  %132 = call double @llvm.fma.f64(double %131, double %118, double noundef 0x3F75DDCDF98946E4)
  %133 = call double @llvm.fma.f64(double %132, double %118, double noundef 0x3F0ADA42D79D8DBB)
  %134 = fmul double %118, %133
  %135 = fdiv double %125, %134
  br label %213

136:                                              ; preds = %68
  %137 = call i32 @llvm.nvvm.d2i.lo(double %spec.select) #139
  %138 = icmp slt i32 %71, 1048576
  br i1 %138, label %139, label %143

139:                                              ; preds = %136
  %140 = fmul double %spec.select, 0x4350000000000000
  %141 = call i32 @llvm.nvvm.d2i.hi(double %140) #139
  %142 = call i32 @llvm.nvvm.d2i.lo(double %140) #139
  br label %143

143:                                              ; preds = %139, %136
  %.0 = phi double [ %140, %139 ], [ %spec.select, %136 ]
  %ihi.i.0 = phi i32 [ %141, %139 ], [ %71, %136 ]
  %ilo.i.0 = phi i32 [ %142, %139 ], [ %137, %136 ]
  %e.i.0 = phi i32 [ -1077, %139 ], [ -1023, %136 ]
  %144 = add i32 %ihi.i.0, -1
  %145 = icmp ult i32 %144, 2146435071
  br i1 %145, label %146, label %195

146:                                              ; preds = %143
  %147 = lshr i32 %ihi.i.0, 20
  %148 = add nsw i32 %e.i.0, %147
  %149 = and i32 %ihi.i.0, -2146435073
  %150 = or i32 %149, 1072693248
  %151 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %150) #139
  %152 = icmp sgt i32 %150, 1073127582
  br i1 %152, label %153, label %159

153:                                              ; preds = %146
  %154 = call i32 @llvm.nvvm.d2i.lo(double %151) #139
  %155 = call i32 @llvm.nvvm.d2i.hi(double %151) #139
  %156 = add i32 %155, -1048576
  %157 = call double @llvm.nvvm.lohi.i2d(i32 %154, i32 %156) #139
  %158 = add nsw i32 %148, 1
  br label %159

159:                                              ; preds = %153, %146
  %m.i.0 = phi double [ %157, %153 ], [ %151, %146 ]
  %e.i.1 = phi i32 [ %158, %153 ], [ %148, %146 ]
  %160 = fadd double %m.i.0, -1.000000e+00
  %161 = fadd double %m.i.0, 1.000000e+00
  %162 = call double @llvm.nvvm.rcp.approx.ftz.d(double %161) #139
  %163 = fneg double %161
  %164 = call double @llvm.fma.f64(double %163, double %162, double noundef 1.000000e+00)
  %165 = call double @llvm.fma.f64(double %164, double %164, double %164)
  %166 = call double @llvm.fma.f64(double %165, double %162, double %162)
  %167 = fmul double %160, %166
  %168 = fadd double %167, %167
  %169 = fmul double %168, %168
  %170 = call double @llvm.fma.f64(double %169, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %171 = call double @llvm.fma.f64(double %170, double %169, double noundef 0x3EF3B2669F02676F)
  %172 = call double @llvm.fma.f64(double %171, double %169, double noundef 0x3F1745CBA9AB0956)
  %173 = call double @llvm.fma.f64(double %172, double %169, double noundef 0x3F3C71C72D1B5154)
  %174 = call double @llvm.fma.f64(double %173, double %169, double noundef 0x3F624924923BE72D)
  %175 = call double @llvm.fma.f64(double %174, double %169, double noundef 0x3F8999999999A3C4)
  %176 = call double @llvm.fma.f64(double %175, double %169, double noundef 0x3FB5555555555554)
  %177 = fsub double %160, %168
  %178 = fmul double %177, 2.000000e+00
  %179 = fneg double %168
  %180 = call double @llvm.fma.f64(double %179, double %160, double %178)
  %181 = fmul double %166, %180
  %182 = fmul double %169, %176
  %183 = call double @llvm.fma.f64(double %182, double %168, double %181)
  %184 = xor i32 %e.i.1, -2147483648
  %185 = call double @llvm.nvvm.lohi.i2d(i32 %184, i32 noundef 1127219200) #139
  %186 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %187 = fsub double %185, %186
  %188 = call double @llvm.fma.f64(double %187, double noundef 0x3FE62E42FEFA39EF, double %168)
  %189 = fneg double %187
  %190 = call double @llvm.fma.f64(double %189, double noundef 0x3FE62E42FEFA39EF, double %188)
  %191 = fsub double %190, %168
  %192 = fsub double %183, %191
  %193 = call double @llvm.fma.f64(double %187, double noundef 0x3C7ABC9E3B39803F, double %192)
  %194 = fadd double %188, %193
  br label %__nv_log.exit

195:                                              ; preds = %143
  %196 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %197 = call i32 @llvm.nvvm.d2i.hi(double %.0) #139
  %198 = bitcast i32 %197 to float
  %199 = fcmp oeq float %198, 0.000000e+00
  %q.i.0 = select i1 %199, double 0xFFF0000000000000, double %196
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %195, %159
  %q.i.1 = phi double [ %194, %159 ], [ %q.i.0, %195 ]
  %200 = fneg double %q.i.1
  %201 = call double @llvm.nvvm.rsqrt.approx.d(double %200) #139
  %202 = call double @llvm.fma.f64(double %201, double noundef 0x3FE8E2101C71B0BF, double noundef 0x3FFA2013964E259C)
  %203 = call double @llvm.fma.f64(double %202, double %201, double noundef 0x3FDABFE90921BE68)
  %204 = call double @llvm.fma.f64(double %203, double %201, double noundef 0x3F97E41314DE00D4)
  %205 = call double @llvm.fma.f64(double %204, double %201, double noundef 0x3F311BD487102E94)
  %206 = fadd double %201, 0x3FF59895C30BAA54
  %207 = call double @llvm.fma.f64(double %206, double %201, double noundef 0x3FFAE8E5956A143F)
  %208 = call double @llvm.fma.f64(double %207, double %201, double noundef 0x3FDACCE85FF7383D)
  %209 = call double @llvm.fma.f64(double %208, double %201, double noundef 0x3F97E43B6CAC34FE)
  %210 = call double @llvm.fma.f64(double %209, double %201, double noundef 0x3F311BD08289EB12)
  %211 = fmul double %201, %210
  %212 = fdiv double %205, %211
  br label %213

213:                                              ; preds = %__nv_log.exit, %74
  %t.0 = phi double [ %135, %74 ], [ %212, %__nv_log.exit ]
  %214 = fneg double %t.0
  %spec.select2 = select i1 %70, double %214, double %t.0
  br label %215

215:                                              ; preds = %213, %4
  %t.2 = phi double [ %67, %4 ], [ %spec.select2, %213 ]
  ret double %t.2
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_normcdfinv(double %a) local_unnamed_addr #115 {
  %1 = fadd double %a, %a
  %2 = fsub double 2.000000e+00, %1
  %3 = fcmp oge double %1, 0x3F4FA4D2AD8F904D
  %4 = fcmp ole double %1, 0x3FFFFC0B65AA4E0E
  %or.cond = and i1 %3, %4
  br i1 %or.cond, label %5, label %69

5:                                                ; preds = %0
  %6 = fmul double %1, %2
  %7 = call i32 @llvm.nvvm.d2i.hi(double %6) #142
  %8 = call i32 @llvm.nvvm.d2i.lo(double %6) #142
  %9 = lshr i32 %7, 20
  %10 = and i32 %9, 2046
  %11 = add nuw i32 %10, 2147482626
  %12 = call double @llvm.nvvm.lohi.i2d(i32 %11, i32 noundef 1127219200) #142
  %13 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %14 = fsub double %12, %13
  %15 = and i32 %7, -2145386497
  %16 = or i32 %15, 1071644672
  %17 = call double @llvm.nvvm.lohi.i2d(i32 %8, i32 %16) #142
  %18 = fadd double %17, -1.000000e+00
  %19 = fadd double %17, 1.000000e+00
  %20 = call double @llvm.nvvm.rcp.approx.ftz.d(double %19) #142
  %21 = fneg double %19
  %22 = call double @llvm.fma.f64(double %21, double %20, double noundef 1.000000e+00) #140
  %23 = call double @llvm.fma.f64(double %22, double %22, double %22) #140
  %24 = call double @llvm.fma.f64(double %23, double %20, double %20) #140
  %25 = fmul double %18, %24
  %26 = call double @llvm.fma.f64(double %25, double noundef -2.000000e+00, double %18) #140
  %27 = fneg double %25
  %28 = call double @llvm.fma.f64(double %27, double %18, double %26) #140
  %29 = call double @llvm.fma.f64(double %28, double %24, double %25) #140
  %30 = fmul double %29, %29
  %31 = call double @llvm.fma.f64(double %30, double noundef 0x3FB5C5C218C775C9, double noundef 0x3FA55CF59CDC5D89) #140
  %32 = call double @llvm.fma.f64(double %31, double %30, double noundef 0x3FAEFD18CF6EBB9C) #140
  %33 = call double @llvm.fma.f64(double %32, double %30, double noundef 0x3FB10682EDCB8D1B) #140
  %34 = call double @llvm.fma.f64(double %33, double %30, double noundef 0x3FB3B1DD3AC7FC96) #140
  %35 = call double @llvm.fma.f64(double %34, double %30, double noundef 0x3FB745CB459B54A6) #140
  %36 = call double @llvm.fma.f64(double %35, double %30, double noundef 0x3FBC71C741A0669F) #140
  %37 = call double @llvm.fma.f64(double %36, double %30, double noundef 0x3FC249249209112E) #140
  %38 = call double @llvm.fma.f64(double %37, double %30, double noundef 0x3FC99999999A06C1) #140
  %39 = call double @llvm.fma.f64(double %38, double %30, double noundef 0x3FD5555555555535) #140
  %40 = fmul double %30, %39
  %41 = call double @llvm.fma.f64(double %40, double %29, double %29) #140
  %42 = fadd double %41, %41
  %43 = call double @llvm.fma.f64(double %14, double noundef 0x3FE62E42FEFA39EF, double %42) #140
  %44 = fsub double -3.125000e+00, %43
  %45 = call double @llvm.fma.f64(double %44, double noundef 0xBBB135D2E746E627, double noundef 0xBC08DDF93324D327) #140
  %46 = call double @llvm.fma.f64(double %45, double %44, double noundef 0x3C37B83EEF0B7C9F) #140
  %47 = call double @llvm.fma.f64(double %46, double %44, double noundef 0x3C69BA72CD589B91) #140
  %48 = call double @llvm.fma.f64(double %47, double %44, double noundef 0xBCA33689090A6B96) #140
  %49 = call double @llvm.fma.f64(double %48, double %44, double noundef 0x3C782E11898132E0) #140
  %50 = call double @llvm.fma.f64(double %49, double %44, double noundef 0x3CFDE4ACFD9E26BA) #140
  %51 = call double @llvm.fma.f64(double %50, double %44, double noundef 0xBD26D33EED66C487) #140
  %52 = call double @llvm.fma.f64(double %51, double %44, double noundef 0xBD36F2167040D8E2) #140
  %53 = call double @llvm.fma.f64(double %52, double %44, double noundef 0x3D872A22C2D77E20) #140
  %54 = call double @llvm.fma.f64(double %53, double %44, double noundef 0xBDAC8859C4E5C0AF) #140
  %55 = call double @llvm.fma.f64(double %54, double %44, double noundef 0xBDCDC583D118A561) #140
  %56 = call double @llvm.fma.f64(double %55, double %44, double noundef 0x3E120F47CCF46B3C) #140
  %57 = call double @llvm.fma.f64(double %56, double %44, double noundef 0xBE31A9E38DC84D60) #140
  %58 = call double @llvm.fma.f64(double %57, double %44, double noundef 0xBE5F36CD6D3D46A9) #140
  %59 = call double @llvm.fma.f64(double %58, double %44, double noundef 0x3E9C6B4F5D03B787) #140
  %60 = call double @llvm.fma.f64(double %59, double %44, double noundef 0xBEB6E8A5434AE8A2) #140
  %61 = call double @llvm.fma.f64(double %60, double %44, double noundef 0xBEED1D1F7B8736F6) #140
  %62 = call double @llvm.fma.f64(double %61, double %44, double noundef 0x3F2879C2A212F024) #140
  %63 = call double @llvm.fma.f64(double %62, double %44, double noundef 0xBF4845769484FCA8) #140
  %64 = call double @llvm.fma.f64(double %63, double %44, double noundef 0xBF78B6C33114F909) #140
  %65 = call double @llvm.fma.f64(double %64, double %44, double noundef 0x3FCEBD80D9B13E28) #140
  %66 = call double @llvm.fma.f64(double %65, double %44, double noundef 0x3FFA755E7C99AE86) #140
  %67 = fneg double %1
  %68 = call double @llvm.fma.f64(double %66, double %67, double %66) #140
  br label %__nv_erfcinv.exit

69:                                               ; preds = %0
  %70 = call i32 @llvm.nvvm.d2i.hi(double %1) #142
  %71 = icmp sgt i32 %70, 1072693247
  %spec.select = select i1 %71, double %2, double %1
  %72 = call i32 @llvm.nvvm.d2i.hi(double %spec.select) #142
  %73 = bitcast i32 %72 to float
  %74 = fcmp ult float %73, 0x3D657FE5E0000000
  br i1 %74, label %137, label %75

75:                                               ; preds = %69
  %76 = call i32 @llvm.nvvm.d2i.lo(double %spec.select) #142
  %77 = lshr i32 %72, 20
  %78 = and i32 %77, 2046
  %79 = add nuw i32 %78, 2147482626
  %80 = call double @llvm.nvvm.lohi.i2d(i32 %79, i32 noundef 1127219200) #142
  %81 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %82 = fsub double %80, %81
  %83 = and i32 %72, -2145386497
  %84 = or i32 %83, 1071644672
  %85 = call double @llvm.nvvm.lohi.i2d(i32 %76, i32 %84) #142
  %86 = fadd double %85, -1.000000e+00
  %87 = fadd double %85, 1.000000e+00
  %88 = call double @llvm.nvvm.rcp.approx.ftz.d(double %87) #142
  %89 = fneg double %87
  %90 = call double @llvm.fma.f64(double %89, double %88, double noundef 1.000000e+00) #140
  %91 = call double @llvm.fma.f64(double %90, double %90, double %90) #140
  %92 = call double @llvm.fma.f64(double %91, double %88, double %88) #140
  %93 = fmul double %86, %92
  %94 = call double @llvm.fma.f64(double %93, double noundef -2.000000e+00, double %86) #140
  %95 = fneg double %93
  %96 = call double @llvm.fma.f64(double %95, double %86, double %94) #140
  %97 = call double @llvm.fma.f64(double %96, double %92, double %93) #140
  %98 = fmul double %97, %97
  %99 = call double @llvm.fma.f64(double %98, double noundef 0x3FB5C5C218C775C9, double noundef 0x3FA55CF59CDC5D89) #140
  %100 = call double @llvm.fma.f64(double %99, double %98, double noundef 0x3FAEFD18CF6EBB9C) #140
  %101 = call double @llvm.fma.f64(double %100, double %98, double noundef 0x3FB10682EDCB8D1B) #140
  %102 = call double @llvm.fma.f64(double %101, double %98, double noundef 0x3FB3B1DD3AC7FC96) #140
  %103 = call double @llvm.fma.f64(double %102, double %98, double noundef 0x3FB745CB459B54A6) #140
  %104 = call double @llvm.fma.f64(double %103, double %98, double noundef 0x3FBC71C741A0669F) #140
  %105 = call double @llvm.fma.f64(double %104, double %98, double noundef 0x3FC249249209112E) #140
  %106 = call double @llvm.fma.f64(double %105, double %98, double noundef 0x3FC99999999A06C1) #140
  %107 = call double @llvm.fma.f64(double %106, double %98, double noundef 0x3FD5555555555535) #140
  %108 = fmul double %98, %107
  %109 = call double @llvm.fma.f64(double %108, double %97, double %97) #140
  %110 = fadd double %109, %109
  %111 = call double @llvm.fma.f64(double %82, double noundef 0x3FE62E42FEFA39EF, double %110) #140
  %112 = fneg double %111
  %113 = call double asm "rsqrt.approx.ftz.f64 $0, $1;", "=d,d"(double %112) #141, !srcloc !84
  %114 = fneg double %113
  %115 = fmul double %113, %114
  %116 = call double @llvm.fma.f64(double %112, double %115, double noundef 1.000000e+00)
  %117 = call double @llvm.fma.f64(double %116, double noundef 3.750000e-01, double noundef 5.000000e-01)
  %118 = fmul double %113, %116
  %119 = call double @llvm.fma.f64(double %117, double %118, double %113)
  %120 = call double @llvm.fma.f64(double %119, double noundef 0x3FEBE9222591AFAB, double noundef 0x4000A0E7333839AA)
  %121 = call double @llvm.fma.f64(double %120, double %119, double noundef 0x4008768CF7E57D5C)
  %122 = call double @llvm.fma.f64(double %121, double %119, double noundef 0x400B77E7E28DA583)
  %123 = call double @llvm.fma.f64(double %122, double %119, double noundef 0x3FF34F26A4F99CF9)
  %124 = call double @llvm.fma.f64(double %123, double %119, double noundef 0x3FC1F674ADB019ED)
  %125 = call double @llvm.fma.f64(double %124, double %119, double noundef 0x3F75DDAE9506431D)
  %126 = call double @llvm.fma.f64(double %125, double %119, double noundef 0x3F0ADA49AA32489C)
  %127 = fadd double %119, 0x4001E90FF51C2197
  %128 = call double @llvm.fma.f64(double %127, double %119, double noundef 0x40111EA3A7CF3820)
  %129 = call double @llvm.fma.f64(double %128, double %119, double noundef 0x4011A0E4A4749594)
  %130 = call double @llvm.fma.f64(double %129, double %119, double noundef 0x400D4E977D38C14D)
  %131 = call double @llvm.fma.f64(double %130, double %119, double noundef 0x3FF37FD567EC0D5F)
  %132 = call double @llvm.fma.f64(double %131, double %119, double noundef 0x3FC1FB9D7F676033)
  %133 = call double @llvm.fma.f64(double %132, double %119, double noundef 0x3F75DDCDF98946E4)
  %134 = call double @llvm.fma.f64(double %133, double %119, double noundef 0x3F0ADA42D79D8DBB)
  %135 = fmul double %119, %134
  %136 = fdiv double %126, %135
  br label %214

137:                                              ; preds = %69
  %138 = call i32 @llvm.nvvm.d2i.lo(double %spec.select) #139
  %139 = icmp slt i32 %72, 1048576
  br i1 %139, label %140, label %144

140:                                              ; preds = %137
  %141 = fmul double %spec.select, 0x4350000000000000
  %142 = call i32 @llvm.nvvm.d2i.hi(double %141) #139
  %143 = call i32 @llvm.nvvm.d2i.lo(double %141) #139
  br label %144

144:                                              ; preds = %140, %137
  %.0 = phi double [ %141, %140 ], [ %spec.select, %137 ]
  %ihi.i.i.0 = phi i32 [ %142, %140 ], [ %72, %137 ]
  %ilo.i.i.0 = phi i32 [ %143, %140 ], [ %138, %137 ]
  %e.i.i.0 = phi i32 [ -1077, %140 ], [ -1023, %137 ]
  %145 = add i32 %ihi.i.i.0, -1
  %146 = icmp ult i32 %145, 2146435071
  br i1 %146, label %147, label %196

147:                                              ; preds = %144
  %148 = lshr i32 %ihi.i.i.0, 20
  %149 = add nsw i32 %e.i.i.0, %148
  %150 = and i32 %ihi.i.i.0, -2146435073
  %151 = or i32 %150, 1072693248
  %152 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i.0, i32 %151) #139
  %153 = icmp sgt i32 %151, 1073127582
  br i1 %153, label %154, label %160

154:                                              ; preds = %147
  %155 = call i32 @llvm.nvvm.d2i.lo(double %152) #139
  %156 = call i32 @llvm.nvvm.d2i.hi(double %152) #139
  %157 = add i32 %156, -1048576
  %158 = call double @llvm.nvvm.lohi.i2d(i32 %155, i32 %157) #139
  %159 = add nsw i32 %149, 1
  br label %160

160:                                              ; preds = %154, %147
  %m.i.i.0 = phi double [ %158, %154 ], [ %152, %147 ]
  %e.i.i.1 = phi i32 [ %159, %154 ], [ %149, %147 ]
  %161 = fadd double %m.i.i.0, -1.000000e+00
  %162 = fadd double %m.i.i.0, 1.000000e+00
  %163 = call double @llvm.nvvm.rcp.approx.ftz.d(double %162) #139
  %164 = fneg double %162
  %165 = call double @llvm.fma.f64(double %164, double %163, double noundef 1.000000e+00)
  %166 = call double @llvm.fma.f64(double %165, double %165, double %165)
  %167 = call double @llvm.fma.f64(double %166, double %163, double %163)
  %168 = fmul double %161, %167
  %169 = fadd double %168, %168
  %170 = fmul double %169, %169
  %171 = call double @llvm.fma.f64(double %170, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %172 = call double @llvm.fma.f64(double %171, double %170, double noundef 0x3EF3B2669F02676F)
  %173 = call double @llvm.fma.f64(double %172, double %170, double noundef 0x3F1745CBA9AB0956)
  %174 = call double @llvm.fma.f64(double %173, double %170, double noundef 0x3F3C71C72D1B5154)
  %175 = call double @llvm.fma.f64(double %174, double %170, double noundef 0x3F624924923BE72D)
  %176 = call double @llvm.fma.f64(double %175, double %170, double noundef 0x3F8999999999A3C4)
  %177 = call double @llvm.fma.f64(double %176, double %170, double noundef 0x3FB5555555555554)
  %178 = fsub double %161, %169
  %179 = fmul double %178, 2.000000e+00
  %180 = fneg double %169
  %181 = call double @llvm.fma.f64(double %180, double %161, double %179)
  %182 = fmul double %167, %181
  %183 = fmul double %170, %177
  %184 = call double @llvm.fma.f64(double %183, double %169, double %182)
  %185 = xor i32 %e.i.i.1, -2147483648
  %186 = call double @llvm.nvvm.lohi.i2d(i32 %185, i32 noundef 1127219200) #139
  %187 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %188 = fsub double %186, %187
  %189 = call double @llvm.fma.f64(double %188, double noundef 0x3FE62E42FEFA39EF, double %169)
  %190 = fneg double %188
  %191 = call double @llvm.fma.f64(double %190, double noundef 0x3FE62E42FEFA39EF, double %189)
  %192 = fsub double %191, %169
  %193 = fsub double %184, %192
  %194 = call double @llvm.fma.f64(double %188, double noundef 0x3C7ABC9E3B39803F, double %193)
  %195 = fadd double %189, %194
  br label %__nv_log.exit.i

196:                                              ; preds = %144
  %197 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %198 = call i32 @llvm.nvvm.d2i.hi(double %.0) #139
  %199 = bitcast i32 %198 to float
  %200 = fcmp oeq float %199, 0.000000e+00
  %q.i.i.0 = select i1 %200, double 0xFFF0000000000000, double %197
  br label %__nv_log.exit.i

__nv_log.exit.i:                                  ; preds = %196, %160
  %q.i.i.1 = phi double [ %195, %160 ], [ %q.i.i.0, %196 ]
  %201 = fneg double %q.i.i.1
  %202 = call double @llvm.nvvm.rsqrt.approx.d(double %201) #139
  %203 = call double @llvm.fma.f64(double %202, double noundef 0x3FE8E2101C71B0BF, double noundef 0x3FFA2013964E259C)
  %204 = call double @llvm.fma.f64(double %203, double %202, double noundef 0x3FDABFE90921BE68)
  %205 = call double @llvm.fma.f64(double %204, double %202, double noundef 0x3F97E41314DE00D4)
  %206 = call double @llvm.fma.f64(double %205, double %202, double noundef 0x3F311BD487102E94)
  %207 = fadd double %202, 0x3FF59895C30BAA54
  %208 = call double @llvm.fma.f64(double %207, double %202, double noundef 0x3FFAE8E5956A143F)
  %209 = call double @llvm.fma.f64(double %208, double %202, double noundef 0x3FDACCE85FF7383D)
  %210 = call double @llvm.fma.f64(double %209, double %202, double noundef 0x3F97E43B6CAC34FE)
  %211 = call double @llvm.fma.f64(double %210, double %202, double noundef 0x3F311BD08289EB12)
  %212 = fmul double %202, %211
  %213 = fdiv double %206, %212
  br label %214

214:                                              ; preds = %__nv_log.exit.i, %75
  %t.i.0 = phi double [ %136, %75 ], [ %213, %__nv_log.exit.i ]
  %215 = fneg double %t.i.0
  %spec.select2 = select i1 %71, double %215, double %t.i.0
  br label %__nv_erfcinv.exit

__nv_erfcinv.exit:                                ; preds = %214, %5
  %t.i.2 = phi double [ %68, %5 ], [ %spec.select2, %214 ]
  %216 = fmul double %t.i.2, 0xBCA21165F626CDD5
  %217 = call double @llvm.fma.f64(double %t.i.2, double noundef 0xBFF6A09E667F3BCC, double %216)
  %218 = fadd double %217, 0.000000e+00
  ret double %218
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_erfc(double %a) local_unnamed_addr #116 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = and i32 %1, 2147483647
  %3 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %4 = icmp ult i32 %2, 2146435072
  br i1 %4, label %5, label %90

5:                                                ; preds = %0
  %6 = call double @llvm.nvvm.lohi.i2d(i32 %3, i32 %2) #142
  %7 = fadd double %6, -4.000000e+00
  %8 = fadd double %6, 4.000000e+00
  %9 = call double @llvm.nvvm.rcp.approx.ftz.d(double %8) #142
  %10 = fneg double %8
  %11 = call double @llvm.fma.f64(double %10, double %9, double noundef 1.000000e+00) #140
  %12 = call double @llvm.fma.f64(double %11, double %11, double %11) #140
  %13 = call double @llvm.fma.f64(double %12, double %9, double %9) #140
  %14 = fmul double %7, %13
  %15 = fadd double %14, 1.000000e+00
  %16 = call double @llvm.fma.f64(double %15, double noundef -4.000000e+00, double %6) #140
  %17 = fneg double %14
  %18 = call double @llvm.fma.f64(double %17, double %6, double %16) #140
  %19 = call double @llvm.fma.f64(double %13, double %18, double %14) #140
  %20 = call double @llvm.fma.f64(double %19, double noundef 0xBDF8774AD4E0BFD7, double noundef 0xBE44E1C6FD03D328) #140
  %21 = call double @llvm.fma.f64(double %20, double %19, double noundef 0xBE4330149F7A56B6) #140
  %22 = call double @llvm.fma.f64(double %21, double %19, double noundef 0x3E7BEDDED8376273) #140
  %23 = call double @llvm.fma.f64(double %22, double %19, double noundef 0x3E6F9254C3ABF22B) #140
  %24 = call double @llvm.fma.f64(double %23, double %19, double noundef 0xBEAB9068C2148CF0) #140
  %25 = call double @llvm.fma.f64(double %24, double %19, double noundef 0x3E94C6454DB34009) #140
  %26 = call double @llvm.fma.f64(double %25, double %19, double noundef 0x3ED7F1C378F2311D) #140
  %27 = call double @llvm.fma.f64(double %26, double %19, double noundef 0xBEE78E051C6D5C58) #140
  %28 = call double @llvm.fma.f64(double %27, double %19, double noundef 0xBEF995B4EAD14A90) #140
  %29 = call double @llvm.fma.f64(double %28, double %19, double noundef 0x3F23BE27CF0A29B2) #140
  %30 = call double @llvm.fma.f64(double %29, double %19, double noundef 0xBF2A1DEF3E81672E) #140
  %31 = call double @llvm.fma.f64(double %30, double %19, double noundef 0xBF48D4ABE68C1713) #140
  %32 = call double @llvm.fma.f64(double %31, double %19, double noundef 0x3F749C67210DD6B4) #140
  %33 = call double @llvm.fma.f64(double %32, double %19, double noundef 0xBF9096238568E357) #140
  %34 = call double @llvm.fma.f64(double %33, double %19, double noundef 0x3FA3079EDF8C2DC9) #140
  %35 = call double @llvm.fma.f64(double %34, double %19, double noundef 0xBFB0FB06DFF601FC) #140
  %36 = call double @llvm.fma.f64(double %35, double %19, double noundef 0x3FB7FEE004DFBCDC) #140
  %37 = call double @llvm.fma.f64(double %36, double %19, double noundef 0xBFB9DDB23C3DB8C6) #140
  %38 = call double @llvm.fma.f64(double %37, double %19, double noundef 0x3FB16ECEFCFA5FDA) #140
  %39 = call double @llvm.fma.f64(double %38, double %19, double noundef 0x3F8F7F5DF66FB6D6) #140
  %40 = call double @llvm.fma.f64(double %39, double %19, double noundef 0xBFC1DF1AD154A29D) #140
  %41 = call double @llvm.fma.f64(double %40, double %19, double noundef 0x3FF3BA5916E9FD7F) #140
  %42 = call double @llvm.fma.f64(double %6, double noundef 2.000000e+00, double noundef 1.000000e+00) #140
  %43 = call double @llvm.nvvm.rcp.approx.ftz.d(double %42) #142
  %44 = fneg double %42
  %45 = call double @llvm.fma.f64(double %44, double %43, double noundef 1.000000e+00) #140
  %46 = call double @llvm.fma.f64(double %45, double %45, double %45) #140
  %47 = call double @llvm.fma.f64(double %46, double %43, double %43) #140
  %48 = fmul double %47, %41
  %49 = fmul double %48, -2.000000e+00
  %50 = call double @llvm.fma.f64(double %6, double %49, double %41) #140
  %51 = fsub double %50, %48
  %52 = call double @llvm.fma.f64(double %51, double %47, double %48) #140
  %53 = fneg double %6
  %54 = fmul double %6, %53
  %55 = call double @llvm.fma.f64(double %54, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %56 = call i32 @llvm.nvvm.d2i.lo(double %55) #142
  %57 = fadd double %55, 0xC338000000000000
  %58 = call double @llvm.fma.f64(double %57, double noundef 0xBFE62E42FEFA39EF, double %54) #140
  %59 = call double @llvm.fma.f64(double %57, double noundef 0xBC7ABC9E3B39803F, double %58) #140
  %60 = call double @llvm.fma.f64(double %59, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %61 = call double @llvm.fma.f64(double %60, double %59, double noundef 0x3EC71DEE62401315) #140
  %62 = call double @llvm.fma.f64(double %61, double %59, double noundef 0x3EFA01997C89EB71) #140
  %63 = call double @llvm.fma.f64(double %62, double %59, double noundef 0x3F2A01A014761F65) #140
  %64 = call double @llvm.fma.f64(double %63, double %59, double noundef 0x3F56C16C1852B7AF) #140
  %65 = call double @llvm.fma.f64(double %64, double %59, double noundef 0x3F81111111122322) #140
  %66 = call double @llvm.fma.f64(double %65, double %59, double noundef 0x3FA55555555502A1) #140
  %67 = call double @llvm.fma.f64(double %66, double %59, double noundef 0x3FC5555555555511) #140
  %68 = call double @llvm.fma.f64(double %67, double %59, double noundef 0x3FE000000000000B) #140
  %69 = call double @llvm.fma.f64(double %68, double %59, double noundef 1.000000e+00) #140
  %70 = call double @llvm.fma.f64(double %69, double %59, double noundef 1.000000e+00) #140
  %71 = sdiv i32 %56, 2
  %72 = call i32 @llvm.nvvm.d2i.lo(double %70) #142
  %73 = call i32 @llvm.nvvm.d2i.hi(double %70) #142
  %74 = shl i32 %71, 20
  %75 = add i32 %73, %74
  %76 = call double @llvm.nvvm.lohi.i2d(i32 %72, i32 %75) #142
  %77 = sub nsw i32 %56, %71
  %78 = shl i32 %77, 20
  %79 = add nsw i32 %78, 1072693248
  %80 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %79) #142
  %81 = fmul double %80, %76
  %82 = fneg double %54
  %83 = call double @llvm.fma.f64(double %53, double %6, double %82) #140
  %84 = call double @llvm.fma.f64(double %81, double %83, double %81) #140
  %85 = fmul double %84, %52
  %86 = icmp ugt i32 %2, 1077624832
  %spec.select = select i1 %86, double 0.000000e+00, double %85
  %87 = icmp slt i32 %1, 0
  %88 = fsub double 2.000000e+00, %spec.select
  %89 = select i1 %87, double %88, double %spec.select
  br label %96

90:                                               ; preds = %0
  %91 = icmp eq i32 %2, 2146435072
  %92 = icmp eq i32 %3, 0
  %or.cond = select i1 %91, i1 %92, i1 false
  %93 = icmp slt i32 %1, 0
  %94 = select i1 %93, double 2.000000e+00, double 0.000000e+00
  %95 = fadd double %a, %a
  %t1.1 = select i1 %or.cond, double %94, double %95
  br label %96

96:                                               ; preds = %90, %5
  %t1.2 = phi double [ %89, %5 ], [ %t1.1, %90 ]
  ret double %t1.2
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_erfcx(double %a) local_unnamed_addr #117 {
__internal_fast_icmp_abs_lt.exit:
  %0 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %1 = bitcast i32 %0 to float
  %2 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not = icmp eq i32 %2, 0
  %3 = call float @llvm.nvvm.fabs.ftz.f(float %1) #139
  %4 = call float @llvm.fabs.f32(float %1)
  %.02 = select i1 %.not, float %4, float %3
  %5 = fcmp olt float %.02, 3.000000e+00
  br i1 %5, label %6, label %56

6:                                                ; preds = %__internal_fast_icmp_abs_lt.exit
  %7 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %8 = and i32 %0, 2147483647
  %9 = call double @llvm.nvvm.lohi.i2d(i32 %7, i32 %8) #139
  %10 = fadd double %9, -4.000000e+00
  %11 = fadd double %9, 4.000000e+00
  %12 = call double @llvm.nvvm.rcp.approx.ftz.d(double %11) #139
  %13 = fneg double %11
  %14 = call double @llvm.fma.f64(double %13, double %12, double noundef 1.000000e+00)
  %15 = call double @llvm.fma.f64(double %14, double %14, double %14)
  %16 = call double @llvm.fma.f64(double %15, double %12, double %12)
  %17 = fmul double %10, %16
  %18 = fadd double %17, 1.000000e+00
  %19 = call double @llvm.fma.f64(double %18, double noundef -4.000000e+00, double %9)
  %20 = fneg double %17
  %21 = call double @llvm.fma.f64(double %20, double %9, double %19)
  %22 = call double @llvm.fma.f64(double %16, double %21, double %17)
  %23 = call double @llvm.fma.f64(double %22, double noundef 0xBDF8774AD4E0BFD7, double noundef 0xBE44E1C6FD03D328)
  %24 = call double @llvm.fma.f64(double %23, double %22, double noundef 0xBE4330149F7A56B6)
  %25 = call double @llvm.fma.f64(double %24, double %22, double noundef 0x3E7BEDDED8376273)
  %26 = call double @llvm.fma.f64(double %25, double %22, double noundef 0x3E6F9254C3ABF22B)
  %27 = call double @llvm.fma.f64(double %26, double %22, double noundef 0xBEAB9068C2148CF0)
  %28 = call double @llvm.fma.f64(double %27, double %22, double noundef 0x3E94C6454DB34009)
  %29 = call double @llvm.fma.f64(double %28, double %22, double noundef 0x3ED7F1C378F2311D)
  %30 = call double @llvm.fma.f64(double %29, double %22, double noundef 0xBEE78E051C6D5C58)
  %31 = call double @llvm.fma.f64(double %30, double %22, double noundef 0xBEF995B4EAD14A90)
  %32 = call double @llvm.fma.f64(double %31, double %22, double noundef 0x3F23BE27CF0A29B2)
  %33 = call double @llvm.fma.f64(double %32, double %22, double noundef 0xBF2A1DEF3E81672E)
  %34 = call double @llvm.fma.f64(double %33, double %22, double noundef 0xBF48D4ABE68C1713)
  %35 = call double @llvm.fma.f64(double %34, double %22, double noundef 0x3F749C67210DD6B4)
  %36 = call double @llvm.fma.f64(double %35, double %22, double noundef 0xBF9096238568E357)
  %37 = call double @llvm.fma.f64(double %36, double %22, double noundef 0x3FA3079EDF8C2DC9)
  %38 = call double @llvm.fma.f64(double %37, double %22, double noundef 0xBFB0FB06DFF601FC)
  %39 = call double @llvm.fma.f64(double %38, double %22, double noundef 0x3FB7FEE004DFBCDC)
  %40 = call double @llvm.fma.f64(double %39, double %22, double noundef 0xBFB9DDB23C3DB8C6)
  %41 = call double @llvm.fma.f64(double %40, double %22, double noundef 0x3FB16ECEFCFA5FDA)
  %42 = call double @llvm.fma.f64(double %41, double %22, double noundef 0x3F8F7F5DF66FB6D6)
  %43 = call double @llvm.fma.f64(double %42, double %22, double noundef 0xBFC1DF1AD154A29D)
  %44 = call double @llvm.fma.f64(double %43, double %22, double noundef 0x3FF3BA5916E9FD7F)
  %45 = call double @llvm.fma.f64(double %9, double noundef 2.000000e+00, double noundef 1.000000e+00)
  %46 = call double @llvm.nvvm.rcp.approx.ftz.d(double %45) #139
  %47 = fneg double %45
  %48 = call double @llvm.fma.f64(double %47, double %46, double noundef 1.000000e+00)
  %49 = call double @llvm.fma.f64(double %48, double %48, double %48)
  %50 = call double @llvm.fma.f64(double %49, double %46, double %46)
  %51 = fmul double %50, %44
  %52 = fmul double %51, -2.000000e+00
  %53 = call double @llvm.fma.f64(double %9, double %52, double %44)
  %54 = fsub double %53, %51
  %55 = call double @llvm.fma.f64(double %54, double %50, double %51)
  br label %66

56:                                               ; preds = %__internal_fast_icmp_abs_lt.exit
  %57 = fdiv double 1.000000e+00, %a
  %58 = fmul double %57, %57
  %59 = call double @llvm.fma.f64(double %58, double noundef 0xC03D880000000000, double noundef 6.562500e+00)
  %60 = call double @llvm.fma.f64(double %59, double %58, double noundef -1.875000e+00)
  %61 = call double @llvm.fma.f64(double %60, double %58, double noundef 7.500000e-01)
  %62 = call double @llvm.fma.f64(double %61, double %58, double noundef -5.000000e-01)
  %63 = call double @llvm.fma.f64(double %62, double %58, double noundef 1.000000e+00)
  %64 = fmul double %57, 0x3FE20DD750429B6D
  %65 = fmul double %64, %63
  br label %66

66:                                               ; preds = %56, %6
  %t1.0 = phi double [ %55, %6 ], [ %65, %56 ]
  %67 = icmp slt i32 %0, 0
  br i1 %67, label %__internal_fast_icmp_abs_lt.exit3.i, label %115

__internal_fast_icmp_abs_lt.exit3.i:              ; preds = %66
  %68 = fmul double %a, %a
  %69 = fneg double %68
  %70 = call double @llvm.fma.f64(double %a, double %a, double %69)
  %71 = call double @llvm.fma.f64(double %68, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000)
  %72 = call i32 @llvm.nvvm.d2i.lo(double %71) #139
  %73 = fadd double %71, 0xC338000000000000
  %74 = call double @llvm.fma.f64(double %73, double noundef 0xBFE62E42FEFA39EF, double %68)
  %75 = call double @llvm.fma.f64(double %73, double noundef 0xBC7ABC9E3B39803F, double %74)
  %76 = call double @llvm.fma.f64(double %75, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA)
  %77 = call double @llvm.fma.f64(double %76, double %75, double noundef 0x3EC71DEE62401315)
  %78 = call double @llvm.fma.f64(double %77, double %75, double noundef 0x3EFA01997C89EB71)
  %79 = call double @llvm.fma.f64(double %78, double %75, double noundef 0x3F2A01A014761F65)
  %80 = call double @llvm.fma.f64(double %79, double %75, double noundef 0x3F56C16C1852B7AF)
  %81 = call double @llvm.fma.f64(double %80, double %75, double noundef 0x3F81111111122322)
  %82 = call double @llvm.fma.f64(double %81, double %75, double noundef 0x3FA55555555502A1)
  %83 = call double @llvm.fma.f64(double %82, double %75, double noundef 0x3FC5555555555511)
  %84 = call double @llvm.fma.f64(double %83, double %75, double noundef 0x3FE000000000000B)
  %85 = call double @llvm.fma.f64(double %84, double %75, double noundef 1.000000e+00)
  %86 = call double @llvm.fma.f64(double %85, double %75, double noundef 1.000000e+00)
  %87 = call i32 @llvm.nvvm.d2i.lo(double %86) #139
  %88 = call i32 @llvm.nvvm.d2i.hi(double %86) #139
  %89 = shl i32 %72, 20
  %90 = add i32 %88, %89
  %91 = call double @llvm.nvvm.lohi.i2d(i32 %87, i32 %90) #139
  %92 = call i32 @llvm.nvvm.d2i.hi(double %68) #139
  %93 = bitcast i32 %92 to float
  %94 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not1 = icmp eq i32 %94, 0
  %95 = call float @llvm.nvvm.fabs.ftz.f(float %93) #139
  %96 = call float @llvm.fabs.f32(float %93)
  %.01 = select i1 %.not1, float %96, float %95
  %97 = fcmp olt float %.01, 0x4010C46560000000
  br i1 %97, label %__nv_exp.exit, label %__internal_fast_icmp_abs_lt.exit.i

__internal_fast_icmp_abs_lt.exit.i:               ; preds = %__internal_fast_icmp_abs_lt.exit3.i
  %98 = fadd double %68, 0x7FF0000000000000
  %99 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not2 = icmp eq i32 %99, 0
  %.0 = select i1 %.not2, float %96, float %95
  %100 = fcmp olt float %.0, 0x4010E90000000000
  br i1 %100, label %101, label %__nv_exp.exit

101:                                              ; preds = %__internal_fast_icmp_abs_lt.exit.i
  %102 = sdiv i32 %72, 2
  %103 = shl i32 %102, 20
  %104 = add i32 %88, %103
  %105 = call double @llvm.nvvm.lohi.i2d(i32 %87, i32 %104) #139
  %106 = sub nsw i32 %72, %102
  %107 = shl i32 %106, 20
  %108 = add nsw i32 %107, 1072693248
  %109 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %108) #139
  %110 = fmul double %109, %105
  br label %__nv_exp.exit

__nv_exp.exit:                                    ; preds = %__internal_fast_icmp_abs_lt.exit.i, %101, %__internal_fast_icmp_abs_lt.exit3.i
  %z.i.2 = phi double [ %91, %__internal_fast_icmp_abs_lt.exit3.i ], [ %110, %101 ], [ %98, %__internal_fast_icmp_abs_lt.exit.i ]
  %111 = fadd double %z.i.2, %z.i.2
  %112 = call double @llvm.fma.f64(double %111, double %70, double %111)
  %113 = fsub double %112, %t1.0
  %114 = fcmp oeq double %111, 0x7FF0000000000000
  %t1.1 = select i1 %114, double %111, double %113
  br label %115

115:                                              ; preds = %__nv_exp.exit, %66
  %t1.2 = phi double [ %t1.1, %__nv_exp.exit ], [ %t1.0, %66 ]
  ret double %t1.2
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_normcdf(double %a) local_unnamed_addr #118 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ogt double %1, 3.850000e+01
  br i1 %2, label %3, label %11

3:                                                ; preds = %0
  %4 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %5 = call i32 @llvm.nvvm.d2i.lo(double noundef 3.850000e+01) #142
  %6 = call i32 @llvm.nvvm.d2i.hi(double noundef 3.850000e+01) #142
  %7 = and i32 %4, -2147483648
  %8 = and i32 %6, 2147483647
  %9 = or i32 %7, %8
  %10 = call double @llvm.nvvm.lohi.i2d(i32 %5, i32 %9) #142
  br label %11

11:                                               ; preds = %3, %0
  %.0 = phi double [ %10, %3 ], [ %a, %0 ]
  %12 = fmul double %.0, 0xBFE6A09E667F3BCD
  %13 = fneg double %12
  %14 = call double @llvm.fma.f64(double %.0, double noundef 0xBFE6A09E667F3BCD, double %13) #140
  %15 = call double @llvm.fma.f64(double %.0, double noundef 0x3C8BDD3413B26456, double %14) #140
  %16 = fadd double %12, %15
  %17 = call i32 @llvm.nvvm.d2i.hi(double %16) #142
  %18 = and i32 %17, 2147483647
  %19 = call i32 @llvm.nvvm.d2i.lo(double %16) #142
  %20 = icmp ult i32 %18, 2146435072
  br i1 %20, label %21, label %106

21:                                               ; preds = %11
  %22 = call double @llvm.nvvm.lohi.i2d(i32 %19, i32 %18) #142
  %23 = fadd double %22, -4.000000e+00
  %24 = fadd double %22, 4.000000e+00
  %25 = call double @llvm.nvvm.rcp.approx.ftz.d(double %24) #142
  %26 = fneg double %24
  %27 = call double @llvm.fma.f64(double %26, double %25, double noundef 1.000000e+00) #140
  %28 = call double @llvm.fma.f64(double %27, double %27, double %27) #140
  %29 = call double @llvm.fma.f64(double %28, double %25, double %25) #140
  %30 = fmul double %23, %29
  %31 = fadd double %30, 1.000000e+00
  %32 = call double @llvm.fma.f64(double %31, double noundef -4.000000e+00, double %22) #140
  %33 = fneg double %30
  %34 = call double @llvm.fma.f64(double %33, double %22, double %32) #140
  %35 = call double @llvm.fma.f64(double %29, double %34, double %30) #140
  %36 = call double @llvm.fma.f64(double %35, double noundef 0xBDF8774AD4E0BFD7, double noundef 0xBE44E1C6FD03D328) #140
  %37 = call double @llvm.fma.f64(double %36, double %35, double noundef 0xBE4330149F7A56B6) #140
  %38 = call double @llvm.fma.f64(double %37, double %35, double noundef 0x3E7BEDDED8376273) #140
  %39 = call double @llvm.fma.f64(double %38, double %35, double noundef 0x3E6F9254C3ABF22B) #140
  %40 = call double @llvm.fma.f64(double %39, double %35, double noundef 0xBEAB9068C2148CF0) #140
  %41 = call double @llvm.fma.f64(double %40, double %35, double noundef 0x3E94C6454DB34009) #140
  %42 = call double @llvm.fma.f64(double %41, double %35, double noundef 0x3ED7F1C378F2311D) #140
  %43 = call double @llvm.fma.f64(double %42, double %35, double noundef 0xBEE78E051C6D5C58) #140
  %44 = call double @llvm.fma.f64(double %43, double %35, double noundef 0xBEF995B4EAD14A90) #140
  %45 = call double @llvm.fma.f64(double %44, double %35, double noundef 0x3F23BE27CF0A29B2) #140
  %46 = call double @llvm.fma.f64(double %45, double %35, double noundef 0xBF2A1DEF3E81672E) #140
  %47 = call double @llvm.fma.f64(double %46, double %35, double noundef 0xBF48D4ABE68C1713) #140
  %48 = call double @llvm.fma.f64(double %47, double %35, double noundef 0x3F749C67210DD6B4) #140
  %49 = call double @llvm.fma.f64(double %48, double %35, double noundef 0xBF9096238568E357) #140
  %50 = call double @llvm.fma.f64(double %49, double %35, double noundef 0x3FA3079EDF8C2DC9) #140
  %51 = call double @llvm.fma.f64(double %50, double %35, double noundef 0xBFB0FB06DFF601FC) #140
  %52 = call double @llvm.fma.f64(double %51, double %35, double noundef 0x3FB7FEE004DFBCDC) #140
  %53 = call double @llvm.fma.f64(double %52, double %35, double noundef 0xBFB9DDB23C3DB8C6) #140
  %54 = call double @llvm.fma.f64(double %53, double %35, double noundef 0x3FB16ECEFCFA5FDA) #140
  %55 = call double @llvm.fma.f64(double %54, double %35, double noundef 0x3F8F7F5DF66FB6D6) #140
  %56 = call double @llvm.fma.f64(double %55, double %35, double noundef 0xBFC1DF1AD154A29D) #140
  %57 = call double @llvm.fma.f64(double %56, double %35, double noundef 0x3FF3BA5916E9FD7F) #140
  %58 = call double @llvm.fma.f64(double %22, double noundef 2.000000e+00, double noundef 1.000000e+00) #140
  %59 = call double @llvm.nvvm.rcp.approx.ftz.d(double %58) #142
  %60 = fneg double %58
  %61 = call double @llvm.fma.f64(double %60, double %59, double noundef 1.000000e+00) #140
  %62 = call double @llvm.fma.f64(double %61, double %61, double %61) #140
  %63 = call double @llvm.fma.f64(double %62, double %59, double %59) #140
  %64 = fmul double %63, %57
  %65 = fmul double %64, -2.000000e+00
  %66 = call double @llvm.fma.f64(double %22, double %65, double %57) #140
  %67 = fsub double %66, %64
  %68 = call double @llvm.fma.f64(double %67, double %63, double %64) #140
  %69 = fneg double %22
  %70 = fmul double %22, %69
  %71 = call double @llvm.fma.f64(double %70, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %72 = call i32 @llvm.nvvm.d2i.lo(double %71) #142
  %73 = fadd double %71, 0xC338000000000000
  %74 = call double @llvm.fma.f64(double %73, double noundef 0xBFE62E42FEFA39EF, double %70) #140
  %75 = call double @llvm.fma.f64(double %73, double noundef 0xBC7ABC9E3B39803F, double %74) #140
  %76 = call double @llvm.fma.f64(double %75, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %77 = call double @llvm.fma.f64(double %76, double %75, double noundef 0x3EC71DEE62401315) #140
  %78 = call double @llvm.fma.f64(double %77, double %75, double noundef 0x3EFA01997C89EB71) #140
  %79 = call double @llvm.fma.f64(double %78, double %75, double noundef 0x3F2A01A014761F65) #140
  %80 = call double @llvm.fma.f64(double %79, double %75, double noundef 0x3F56C16C1852B7AF) #140
  %81 = call double @llvm.fma.f64(double %80, double %75, double noundef 0x3F81111111122322) #140
  %82 = call double @llvm.fma.f64(double %81, double %75, double noundef 0x3FA55555555502A1) #140
  %83 = call double @llvm.fma.f64(double %82, double %75, double noundef 0x3FC5555555555511) #140
  %84 = call double @llvm.fma.f64(double %83, double %75, double noundef 0x3FE000000000000B) #140
  %85 = call double @llvm.fma.f64(double %84, double %75, double noundef 1.000000e+00) #140
  %86 = call double @llvm.fma.f64(double %85, double %75, double noundef 1.000000e+00) #140
  %87 = sdiv i32 %72, 2
  %88 = call i32 @llvm.nvvm.d2i.lo(double %86) #142
  %89 = call i32 @llvm.nvvm.d2i.hi(double %86) #142
  %90 = shl i32 %87, 20
  %91 = add i32 %89, %90
  %92 = call double @llvm.nvvm.lohi.i2d(i32 %88, i32 %91) #142
  %93 = sub nsw i32 %72, %87
  %94 = shl i32 %93, 20
  %95 = add nsw i32 %94, 1072693248
  %96 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %95) #142
  %97 = fmul double %96, %92
  %98 = fneg double %70
  %99 = call double @llvm.fma.f64(double %69, double %22, double %98) #140
  %100 = call double @llvm.fma.f64(double %97, double %99, double %97) #140
  %101 = fmul double %100, %68
  %102 = icmp ugt i32 %18, 1077624832
  %spec.select = select i1 %102, double 0.000000e+00, double %101
  %103 = icmp slt i32 %17, 0
  %104 = fsub double 2.000000e+00, %spec.select
  %105 = select i1 %103, double %104, double %spec.select
  br label %__nv_erfc.exit

106:                                              ; preds = %11
  %107 = icmp eq i32 %18, 2146435072
  %108 = icmp eq i32 %19, 0
  %or.cond = select i1 %107, i1 %108, i1 false
  %109 = icmp slt i32 %17, 0
  %110 = select i1 %109, double 2.000000e+00, double 0.000000e+00
  %111 = fadd double %16, %16
  %t1.i.1 = select i1 %or.cond, double %110, double %111
  br label %__nv_erfc.exit

__nv_erfc.exit:                                   ; preds = %106, %21
  %t1.i.2 = phi double [ %105, %21 ], [ %t1.i.1, %106 ]
  %112 = call i32 @llvm.nvvm.d2i.hi(double %.0) #142
  %113 = icmp ugt i32 %112, -1074790400
  br i1 %113, label %114, label %120

114:                                              ; preds = %__nv_erfc.exit
  %115 = fsub double %12, %16
  %116 = fadd double %15, %115
  %117 = fmul double %16, -2.000000e+00
  %118 = fmul double %117, %t1.i.2
  %119 = call double @llvm.fma.f64(double %118, double %116, double %t1.i.2) #140
  br label %120

120:                                              ; preds = %114, %__nv_erfc.exit
  %z.0 = phi double [ %119, %114 ], [ %t1.i.2, %__nv_erfc.exit ]
  %121 = fmul double %z.0, 5.000000e-01
  ret double %121
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_tgamma(double %a) local_unnamed_addr #119 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = bitcast i32 %1 to float
  %3 = fcmp ult double %a, 0.000000e+00
  br i1 %3, label %208, label %4

4:                                                ; preds = %0
  %5 = fcmp olt float %2, 2.250000e+00
  br i1 %5, label %6, label %36

6:                                                ; preds = %4
  %7 = fcmp oge float %2, 2.187500e+00
  %8 = fadd double %a, -1.000000e+00
  %s.0 = select i1 %7, double %8, double 1.000000e+00
  %t.0 = select i1 %7, double %8, double %a
  %9 = fcmp oge float %2, 2.062500e+00
  %10 = fneg double %s.0
  %11 = call double @llvm.fma.f64(double %t.0, double %s.0, double %10) #140
  %12 = fadd double %t.0, -1.000000e+00
  %s.1 = select i1 %9, double %11, double %s.0
  %t.1 = select i1 %9, double %12, double %t.0
  %13 = fcmp oge float %2, 1.937500e+00
  %14 = fneg double %s.1
  %15 = call double @llvm.fma.f64(double %t.1, double %s.1, double %14) #140
  %16 = fadd double %t.1, -1.000000e+00
  %s.2 = select i1 %13, double %15, double %s.1
  %t.2 = select i1 %13, double %16, double %t.1
  %17 = fcmp oge float %2, 1.750000e+00
  %18 = fadd double %t.2, -1.000000e+00
  %t.3 = select i1 %17, double %18, double %t.2
  %19 = call double @llvm.fma.f64(double %t.3, double noundef 0x3E381B4960FCF5C9, double noundef 0xBE8AF7049AE8A594) #140
  %20 = call double @llvm.fma.f64(double %19, double %t.3, double noundef 0x3EB301D46D4B22F5) #140
  %21 = call double @llvm.fma.f64(double %20, double %t.3, double noundef 0xBEB50272AEED0FC4) #140
  %22 = call double @llvm.fma.f64(double %21, double %t.3, double noundef 0xBEF51CE1A40516F8) #140
  %23 = call double @llvm.fma.f64(double %22, double %t.3, double noundef 0x3F20C8AA7419084C) #140
  %24 = call double @llvm.fma.f64(double %23, double %t.3, double noundef 0xBF2C3650196BAD8A) #140
  %25 = call double @llvm.fma.f64(double %24, double %t.3, double noundef 0xBF531711365A3E26) #140
  %26 = call double @llvm.fma.f64(double %25, double %t.3, double noundef 0x3F7D919C52A7DF35) #140
  %27 = call double @llvm.fma.f64(double %26, double %t.3, double noundef 0xBF83B4AF28386F4D) #140
  %28 = call double @llvm.fma.f64(double %27, double %t.3, double noundef 0xBFA59AF103C37B4D) #140
  %29 = call double @llvm.fma.f64(double %28, double %t.3, double noundef 0x3FC5512320B439EF) #140
  %30 = call double @llvm.fma.f64(double %29, double %t.3, double noundef 0xBFA5815E8FA26F4F) #140
  %31 = call double @llvm.fma.f64(double %30, double %t.3, double noundef 0xBFE4FCF4026AFA2B) #140
  %32 = call double @llvm.fma.f64(double %31, double %t.3, double noundef 0x3FE2788CFC6FB619) #140
  %33 = call double @llvm.fma.f64(double %32, double %t.3, double noundef 1.000000e+00) #140
  %34 = select i1 %17, double 1.000000e+00, double %a
  %t.4 = fmul double %34, %33
  %35 = fdiv double %s.2, %t.4
  br label %405

36:                                               ; preds = %4
  %37 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %38 = lshr i32 %1, 20
  %39 = icmp ult i32 %1, 1048576
  br i1 %39, label %40, label %46

40:                                               ; preds = %36
  %41 = fmul double %a, 0x4350000000000000
  %42 = call i32 @llvm.nvvm.d2i.hi(double %41) #142
  %43 = call i32 @llvm.nvvm.d2i.lo(double %41) #142
  %44 = lshr i32 %42, 20
  %45 = add nsw i32 %44, -54
  br label %46

46:                                               ; preds = %40, %36
  %ilo.i.i26.0 = phi i32 [ %43, %40 ], [ %37, %36 ]
  %ihi.i.i27.0 = phi i32 [ %42, %40 ], [ %1, %36 ]
  %expo.i.i.0 = phi i32 [ %45, %40 ], [ %38, %36 ]
  %47 = add nsw i32 %expo.i.i.0, -1023
  %48 = and i32 %ihi.i.i27.0, -2146435073
  %49 = or i32 %48, 1072693248
  %50 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.i26.0, i32 %49) #142
  %51 = icmp ugt i32 %49, 1073127582
  br i1 %51, label %52, label %__internal_log_ext_prec.exit.i

52:                                               ; preds = %46
  %53 = call i32 @llvm.nvvm.d2i.lo(double %50) #142
  %54 = call i32 @llvm.nvvm.d2i.hi(double %50) #142
  %55 = add i32 %54, -1048576
  %56 = call double @llvm.nvvm.lohi.i2d(i32 %53, i32 %55) #142
  %57 = add nsw i32 %expo.i.i.0, -1022
  br label %__internal_log_ext_prec.exit.i

__internal_log_ext_prec.exit.i:                   ; preds = %52, %46
  %m.i.i.0 = phi double [ %56, %52 ], [ %50, %46 ]
  %expo.i.i.1 = phi i32 [ %57, %52 ], [ %47, %46 ]
  %58 = fadd double %m.i.i.0, -1.000000e+00
  %59 = fadd double %m.i.i.0, 1.000000e+00
  %60 = call double @llvm.nvvm.rcp.approx.ftz.d(double %59) #142
  %61 = fneg double %59
  %62 = call double @llvm.fma.f64(double %61, double %60, double noundef 1.000000e+00) #140
  %63 = call double @llvm.fma.f64(double %62, double %62, double %62) #140
  %64 = call double @llvm.fma.f64(double %63, double %60, double %60) #140
  %65 = fmul double %58, %64
  %66 = fadd double %65, %65
  %67 = fmul double %66, %66
  %68 = call double @llvm.fma.f64(double %67, double noundef 0x3EB0F5FF7D2CAFE2, double noundef 0x3ED0F5D241AD3B5A) #140
  %69 = call double @llvm.fma.f64(double %68, double %67, double noundef 0x3EF3B20A75488A3F) #140
  %70 = call double @llvm.fma.f64(double %69, double %67, double noundef 0x3F1745CDE4FAECD5) #140
  %71 = call double @llvm.fma.f64(double %70, double %67, double noundef 0x3F3C71C7258A578B) #140
  %72 = call double @llvm.fma.f64(double %71, double %67, double noundef 0x3F6249249242B910) #140
  %73 = call double @llvm.fma.f64(double %72, double %67, double noundef 0x3F89999999999DFB) #140
  %74 = fmul double %67, %73
  %75 = fsub double %58, %66
  %76 = fmul double %75, 2.000000e+00
  %77 = fneg double %66
  %78 = call double @llvm.fma.f64(double %77, double %58, double %76) #140
  %79 = fmul double %64, %78
  %80 = fadd double %74, 0x3FB5555555555555
  %81 = fsub double 0x3FB5555555555555, %80
  %82 = fadd double %74, %81
  %83 = fadd double %82, 0.000000e+00
  %84 = fadd double %83, 0xBC46A4CB00B9E7B0
  %85 = fadd double %80, %84
  %86 = fsub double %80, %85
  %87 = fadd double %84, %86
  %88 = fneg double %67
  %89 = call double @llvm.fma.f64(double %66, double %66, double %88) #140
  %90 = call i32 @llvm.nvvm.d2i.lo(double %79) #142
  %91 = call i32 @llvm.nvvm.d2i.hi(double %79) #142
  %92 = add i32 %91, 1048576
  %93 = call double @llvm.nvvm.lohi.i2d(i32 %90, i32 %92) #142
  %94 = call double @llvm.fma.f64(double %66, double %93, double %89) #140
  %95 = fmul double %66, %67
  %96 = fneg double %95
  %97 = call double @llvm.fma.f64(double %67, double %66, double %96) #140
  %98 = call double @llvm.fma.f64(double %67, double %79, double %97) #140
  %99 = call double @llvm.fma.f64(double %94, double %66, double %98) #140
  %100 = fmul double %95, %85
  %101 = fneg double %100
  %102 = call double @llvm.fma.f64(double %85, double %95, double %101) #140
  %103 = call double @llvm.fma.f64(double %85, double %99, double %102) #140
  %104 = call double @llvm.fma.f64(double %87, double %95, double %103) #140
  %105 = fadd double %100, %104
  %106 = fsub double %100, %105
  %107 = fadd double %104, %106
  %108 = fadd double %66, %105
  %109 = fsub double %66, %108
  %110 = fadd double %105, %109
  %111 = fadd double %107, %110
  %112 = fadd double %79, %111
  %113 = fadd double %108, %112
  %114 = fsub double %108, %113
  %115 = fadd double %112, %114
  %116 = xor i32 %expo.i.i.1, -2147483648
  %117 = call double @llvm.nvvm.lohi.i2d(i32 %116, i32 noundef 1127219200) #142
  %118 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %119 = fsub double %117, %118
  %120 = call double @llvm.fma.f64(double %119, double noundef 0x3FE62E42FEFA39EF, double %113) #140
  %121 = fneg double %119
  %122 = call double @llvm.fma.f64(double %121, double noundef 0x3FE62E42FEFA39EF, double %120) #140
  %123 = fsub double %122, %113
  %124 = fsub double %115, %123
  %125 = call double @llvm.fma.f64(double %119, double noundef 0x3C7ABC9E3B39803F, double %124) #140
  %126 = fadd double %120, %125
  %127 = fsub double %120, %126
  %128 = fadd double %125, %127
  %129 = fadd double %a, -5.000000e-01
  %130 = fmul double %129, %126
  %131 = fneg double %130
  %132 = call double @llvm.fma.f64(double %126, double %129, double %131) #140
  %133 = call double @llvm.fma.f64(double %128, double %129, double %132) #140
  %134 = fadd double %130, %133
  %135 = fsub double %130, %134
  %136 = fadd double %133, %135
  %137 = fsub double %134, %a
  %138 = fsub double %134, %137
  %139 = fsub double %138, %a
  %140 = fadd double %136, %139
  %141 = fadd double %137, %140
  %142 = fsub double %137, %141
  %143 = fadd double %140, %142
  %144 = call double @llvm.fma.f64(double %141, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000) #140
  %145 = call i32 @llvm.nvvm.d2i.lo(double %144) #142
  %146 = fadd double %144, 0xC338000000000000
  %147 = call double @llvm.fma.f64(double %146, double noundef 0xBFE62E42FEFA39EF, double %141) #140
  %148 = call double @llvm.fma.f64(double %146, double noundef 0xBC7ABC9E3B39803F, double %147) #140
  %149 = call double @llvm.fma.f64(double %148, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA) #140
  %150 = call double @llvm.fma.f64(double %149, double %148, double noundef 0x3EC71DEE62401315) #140
  %151 = call double @llvm.fma.f64(double %150, double %148, double noundef 0x3EFA01997C89EB71) #140
  %152 = call double @llvm.fma.f64(double %151, double %148, double noundef 0x3F2A01A014761F65) #140
  %153 = call double @llvm.fma.f64(double %152, double %148, double noundef 0x3F56C16C1852B7AF) #140
  %154 = call double @llvm.fma.f64(double %153, double %148, double noundef 0x3F81111111122322) #140
  %155 = call double @llvm.fma.f64(double %154, double %148, double noundef 0x3FA55555555502A1) #140
  %156 = call double @llvm.fma.f64(double %155, double %148, double noundef 0x3FC5555555555511) #140
  %157 = call double @llvm.fma.f64(double %156, double %148, double noundef 0x3FE000000000000B) #140
  %158 = call double @llvm.fma.f64(double %157, double %148, double noundef 1.000000e+00) #140
  %159 = call double @llvm.fma.f64(double %158, double %148, double noundef 1.000000e+00) #140
  %160 = call i32 @llvm.nvvm.d2i.lo(double %159) #142
  %161 = call i32 @llvm.nvvm.d2i.hi(double %159) #142
  %162 = shl i32 %145, 20
  %163 = add i32 %161, %162
  %164 = call double @llvm.nvvm.lohi.i2d(i32 %160, i32 %163) #142
  %165 = call i32 @llvm.nvvm.d2i.hi(double %141) #142
  %166 = bitcast i32 %165 to float
  %167 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not6 = icmp eq i32 %167, 0
  %168 = call float @llvm.nvvm.fabs.ftz.f(float %166) #139
  %169 = call float @llvm.fabs.f32(float %166)
  %.05 = select i1 %.not6, float %169, float %168
  %170 = fcmp olt float %.05, 0x4010C46560000000
  br i1 %170, label %__nv_exp.exit.i, label %__internal_fast_icmp_abs_lt.exit.i.i

__internal_fast_icmp_abs_lt.exit.i.i:             ; preds = %__internal_log_ext_prec.exit.i
  %171 = fcmp olt double %141, 0.000000e+00
  %172 = fadd double %141, 0x7FF0000000000000
  %z.i5.i.0 = select i1 %171, double 0.000000e+00, double %172
  %173 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139
  %.not7 = icmp eq i32 %173, 0
  %.04 = select i1 %.not7, float %169, float %168
  %174 = fcmp olt float %.04, 0x4010E90000000000
  br i1 %174, label %175, label %__nv_exp.exit.i

175:                                              ; preds = %__internal_fast_icmp_abs_lt.exit.i.i
  %176 = sdiv i32 %145, 2
  %177 = shl i32 %176, 20
  %178 = add i32 %161, %177
  %179 = call double @llvm.nvvm.lohi.i2d(i32 %160, i32 %178) #139
  %180 = sub nsw i32 %145, %176
  %181 = shl i32 %180, 20
  %182 = add nsw i32 %181, 1072693248
  %183 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %182) #139
  %184 = fmul double %183, %179
  br label %__nv_exp.exit.i

__nv_exp.exit.i:                                  ; preds = %__internal_fast_icmp_abs_lt.exit.i.i, %175, %__internal_log_ext_prec.exit.i
  %z.i5.i.2 = phi double [ %164, %__internal_log_ext_prec.exit.i ], [ %184, %175 ], [ %z.i5.i.0, %__internal_fast_icmp_abs_lt.exit.i.i ]
  %185 = call double @llvm.fma.f64(double %143, double %z.i5.i.2, double %z.i5.i.2)
  %186 = fmul double %185, 0xBCAA6A0D6F814637
  %187 = call double @llvm.fma.f64(double %185, double noundef 0x40040D931FF62706, double %186)
  %188 = call double @llvm.nvvm.rcp.approx.ftz.d(double %a) #139
  %189 = fneg double %a
  %190 = call double @llvm.fma.f64(double %189, double %188, double noundef 1.000000e+00)
  %191 = call double @llvm.fma.f64(double %190, double %190, double %190)
  %192 = call double @llvm.fma.f64(double %191, double %188, double %188)
  %193 = call double @llvm.fma.f64(double %192, double noundef 0xBF64BEE47C38A637, double noundef 0x3F73C25DA81303D5)
  %194 = call double @llvm.fma.f64(double %193, double %192, double noundef 0xBF6B7C37A96CFC72)
  %195 = call double @llvm.fma.f64(double %194, double %192, double noundef 0x3F35A85ABDE1E324)
  %196 = call double @llvm.fma.f64(double %195, double %192, double noundef 0x3F4A8B28F07B3F05)
  %197 = call double @llvm.fma.f64(double %196, double %192, double noundef 0xBF0A15D1D45A282F)
  %198 = call double @llvm.fma.f64(double %197, double %192, double noundef 0xBF4367D3468CB5BE)
  %199 = call double @llvm.fma.f64(double %198, double %192, double noundef 0x3F12471B0E9F1005)
  %200 = call double @llvm.fma.f64(double %199, double %192, double noundef 0x3F49B1004744D5C4)
  %201 = call double @llvm.fma.f64(double %200, double %192, double noundef 0xBF2E13CE69AB4B7F)
  %202 = call double @llvm.fma.f64(double %201, double %192, double noundef 0xBF65F7268ECF8A01)
  %203 = call double @llvm.fma.f64(double %202, double %192, double noundef 0x3F6C71C71C71ACE0)
  %204 = call double @llvm.fma.f64(double %203, double %192, double noundef 0x3FB5555555555556)
  %205 = fmul double %192, %204
  %206 = call double @llvm.fma.f64(double %205, double %187, double %187)
  %207 = fcmp oge double %a, 0x406573FAE561F648
  %r.i.0 = select i1 %207, double 0x7FF0000000000000, double %206
  br label %405

208:                                              ; preds = %0
  %209 = fcmp olt double %a, 0.000000e+00
  br i1 %209, label %210, label %403

210:                                              ; preds = %208
  %211 = call double @llvm.trunc.f64(double %a)
  %212 = fcmp oeq double %211, %a
  br i1 %212, label %405, label %213

213:                                              ; preds = %210
  %214 = fcmp ogt float %2, -2.250000e+00
  br i1 %214, label %215, label %245

215:                                              ; preds = %213
  %216 = fcmp ole float %2, -2.187500e+00
  %217 = call double @llvm.fma.f64(double %a, double %a, double %a)
  %218 = fadd double %a, 1.000000e+00
  %s.4 = select i1 %216, double %217, double %a
  %t.5 = select i1 %216, double %218, double %a
  %219 = fcmp ole float %2, -2.062500e+00
  %220 = call double @llvm.fma.f64(double %t.5, double %s.4, double %s.4)
  %221 = fadd double %t.5, 1.000000e+00
  %s.5 = select i1 %219, double %220, double %s.4
  %t.6 = select i1 %219, double %221, double %t.5
  %222 = fcmp ole float %2, -1.937500e+00
  %223 = call double @llvm.fma.f64(double %t.6, double %s.5, double %s.5)
  %224 = fadd double %t.6, 1.000000e+00
  %s.6 = select i1 %222, double %223, double %s.5
  %t.7 = select i1 %222, double %224, double %t.6
  %225 = fcmp ole float %2, -1.750000e+00
  %226 = call double @llvm.fma.f64(double %t.7, double %s.6, double %s.6)
  %227 = fadd double %t.7, 1.000000e+00
  %s.7 = select i1 %225, double %226, double %s.6
  %t.8 = select i1 %225, double %227, double %t.7
  %228 = call double @llvm.fma.f64(double %t.8, double noundef 0x3E381B4960FCF5C9, double noundef 0xBE8AF7049AE8A594)
  %229 = call double @llvm.fma.f64(double %228, double %t.8, double noundef 0x3EB301D46D4B22F5)
  %230 = call double @llvm.fma.f64(double %229, double %t.8, double noundef 0xBEB50272AEED0FC4)
  %231 = call double @llvm.fma.f64(double %230, double %t.8, double noundef 0xBEF51CE1A40516F8)
  %232 = call double @llvm.fma.f64(double %231, double %t.8, double noundef 0x3F20C8AA7419084C)
  %233 = call double @llvm.fma.f64(double %232, double %t.8, double noundef 0xBF2C3650196BAD8A)
  %234 = call double @llvm.fma.f64(double %233, double %t.8, double noundef 0xBF531711365A3E26)
  %235 = call double @llvm.fma.f64(double %234, double %t.8, double noundef 0x3F7D919C52A7DF35)
  %236 = call double @llvm.fma.f64(double %235, double %t.8, double noundef 0xBF83B4AF28386F4D)
  %237 = call double @llvm.fma.f64(double %236, double %t.8, double noundef 0xBFA59AF103C37B4D)
  %238 = call double @llvm.fma.f64(double %237, double %t.8, double noundef 0x3FC5512320B439EF)
  %239 = call double @llvm.fma.f64(double %238, double %t.8, double noundef 0xBFA5815E8FA26F4F)
  %240 = call double @llvm.fma.f64(double %239, double %t.8, double noundef 0xBFE4FCF4026AFA2B)
  %241 = call double @llvm.fma.f64(double %240, double %t.8, double noundef 0x3FE2788CFC6FB619)
  %242 = call double @llvm.fma.f64(double %241, double %t.8, double noundef 1.000000e+00)
  %243 = fmul double %s.7, %242
  %244 = fdiv double 1.000000e+00, %243
  br label %405

245:                                              ; preds = %213
  %246 = fcmp ogt float %2, 0xC00CE40000000000
  br i1 %246, label %247, label %395

247:                                              ; preds = %245
  %248 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %249 = icmp eq i32 %248, 350
  br i1 %249, label %__nv_sincospi.exit.i, label %250

250:                                              ; preds = %247
  %251 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %252 = icmp eq i32 %251, 370
  br i1 %252, label %__nv_sincospi.exit.i, label %__internal_sin_cos_kerneld.exit.i

__nv_sincospi.exit.i:                             ; preds = %250, %247
  %253 = shl i32 %1, 1
  %254 = icmp ugt i32 %253, -2038431744
  %255 = fmul double %a, 0.000000e+00
  %spec.select = select i1 %254, double %255, double %a
  %256 = call i32 @llvm.nvvm.d2i.lo(double %spec.select) #139
  %257 = call i32 @llvm.nvvm.d2i.hi(double %spec.select) #139
  %258 = add i32 %257, 1048576
  %259 = call double @llvm.nvvm.lohi.i2d(i32 %256, i32 %258) #139
  %260 = call double @llvm.round.f64(double %259)
  %261 = fptosi double %260 to i64
  %262 = trunc i64 %261 to i32
  %263 = fneg double %260
  %264 = call double @llvm.fma.f64(double %263, double noundef 5.000000e-01, double %spec.select)
  %265 = fmul double %264, 0x3CA1A62633145C07
  %266 = call double @llvm.fma.f64(double %264, double noundef 0x400921FB54442D18, double %265)
  %267 = fmul double %266, %266
  %268 = call double @llvm.fma.f64(double %267, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %269 = call double @llvm.fma.f64(double %268, double %267, double noundef 0xBE927E4F8E06E6D9)
  %270 = call double @llvm.fma.f64(double %269, double %267, double noundef 0x3EFA01A019DDBCE9)
  %271 = call double @llvm.fma.f64(double %270, double %267, double noundef 0xBF56C16C16C15D47)
  %272 = call double @llvm.fma.f64(double %271, double %267, double noundef 0x3FA5555555555551)
  %273 = call double @llvm.fma.f64(double %272, double %267, double noundef -5.000000e-01)
  %274 = call double @llvm.fma.f64(double %273, double %267, double noundef 1.000000e+00)
  %275 = call double @llvm.fma.f64(double %267, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %276 = call double @llvm.fma.f64(double %275, double %267, double noundef 0x3EC71DE369ACE392)
  %277 = call double @llvm.fma.f64(double %276, double %267, double noundef 0xBF2A01A019DB62A1)
  %278 = call double @llvm.fma.f64(double %277, double %267, double noundef 0x3F81111111110818)
  %279 = call double @llvm.fma.f64(double %278, double %267, double noundef 0xBFC5555555555554)
  %280 = call double @llvm.fma.f64(double %279, double %267, double noundef 0.000000e+00)
  %281 = call double @llvm.fma.f64(double %280, double %266, double %266)
  %282 = and i32 %262, 1
  %.not4 = icmp eq i32 %282, 0
  %s.i.i.0 = select i1 %.not4, double %281, double %274
  %283 = and i32 %262, 2
  %.not5 = icmp eq i32 %283, 0
  %284 = call i32 @llvm.nvvm.d2i.hi(double %s.i.i.0) #139
  %285 = call i32 @llvm.nvvm.d2i.lo(double %s.i.i.0) #139
  %286 = xor i32 %284, -2147483648
  %287 = call double @llvm.nvvm.lohi.i2d(i32 %285, i32 %286) #139
  %s.i.i.1 = select i1 %.not5, double %s.i.i.0, double %287
  %288 = call double @llvm.trunc.f64(double %spec.select)
  %289 = fcmp oeq double %spec.select, %288
  %290 = fmul double %spec.select, 0.000000e+00
  %s.i.i.2 = select i1 %289, double %290, double %s.i.i.1
  br label %__nv_sinpi.exit

__internal_sin_cos_kerneld.exit.i:                ; preds = %250
  %291 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %292 = add i32 %1, 1048576
  %293 = call double @llvm.nvvm.lohi.i2d(i32 %291, i32 %292) #139
  %294 = call double @llvm.round.f64(double %293)
  %295 = fptosi double %294 to i64
  %296 = trunc i64 %295 to i32
  %297 = fneg double %294
  %298 = call double @llvm.fma.f64(double %297, double noundef 5.000000e-01, double %a)
  %299 = fmul double %298, 0x3CA1A62633145C07
  %300 = call double @llvm.fma.f64(double %298, double noundef 0x400921FB54442D18, double %299)
  %301 = shl i64 %295, 3
  %302 = and i64 %301, 8
  %303 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %302
  %304 = fmul double %300, %300
  %305 = and i32 %296, 1
  %.not = icmp eq i32 %305, 0
  %306 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %307 = getelementptr inbounds double, double addrspace(1)* %303, i64 1
  %308 = load double, double addrspace(1)* %307, align 8
  %309 = call double @llvm.fma.f64(double noundef %306, double %304, double %308)
  %310 = getelementptr inbounds double, double addrspace(1)* %303, i64 2
  %311 = load double, double addrspace(1)* %310, align 8
  %312 = call double @llvm.fma.f64(double %309, double %304, double %311)
  %313 = getelementptr inbounds double, double addrspace(1)* %303, i64 3
  %314 = load double, double addrspace(1)* %313, align 8
  %315 = call double @llvm.fma.f64(double %312, double %304, double %314)
  %316 = getelementptr inbounds double, double addrspace(1)* %303, i64 4
  %317 = load double, double addrspace(1)* %316, align 8
  %318 = call double @llvm.fma.f64(double %315, double %304, double %317)
  %319 = getelementptr inbounds double, double addrspace(1)* %303, i64 5
  %320 = load double, double addrspace(1)* %319, align 8
  %321 = call double @llvm.fma.f64(double %318, double %304, double %320)
  %322 = getelementptr inbounds double, double addrspace(1)* %303, i64 6
  %323 = load double, double addrspace(1)* %322, align 8
  %324 = call double @llvm.fma.f64(double %321, double %304, double %323)
  %325 = call double @llvm.fma.f64(double %324, double %300, double %300)
  %326 = call double @llvm.fma.f64(double %324, double %304, double noundef 1.000000e+00)
  %spec.select1 = select i1 %.not, double %325, double %326
  %327 = and i32 %296, 2
  %.not3 = icmp eq i32 %327, 0
  %328 = call double @llvm.fma.f64(double %spec.select1, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not3, double %spec.select1, double %328
  br label %__nv_sinpi.exit

__nv_sinpi.exit:                                  ; preds = %__internal_sin_cos_kerneld.exit.i, %__nv_sincospi.exit.i
  %.01 = phi double [ %s.i.i.2, %__nv_sincospi.exit.i ], [ %.1, %__internal_sin_cos_kerneld.exit.i ]
  %329 = call double @llvm.fabs.f64(double %a)
  %330 = call double @llvm.nvvm.rcp.approx.ftz.d(double %329) #139
  %331 = fneg double %329
  %332 = call double @llvm.fma.f64(double %331, double %330, double noundef 1.000000e+00)
  %333 = call double @llvm.fma.f64(double %332, double %332, double %332)
  %334 = call double @llvm.fma.f64(double %333, double %330, double %330)
  %335 = call double @llvm.fma.f64(double %334, double noundef 0xBF64BEE47C38A637, double noundef 0x3F73C25DA81303D5)
  %336 = call double @llvm.fma.f64(double %335, double %334, double noundef 0xBF6B7C37A96CFC72)
  %337 = call double @llvm.fma.f64(double %336, double %334, double noundef 0x3F35A85ABDE1E324)
  %338 = call double @llvm.fma.f64(double %337, double %334, double noundef 0x3F4A8B28F07B3F05)
  %339 = call double @llvm.fma.f64(double %338, double %334, double noundef 0xBF0A15D1D45A282F)
  %340 = call double @llvm.fma.f64(double %339, double %334, double noundef 0xBF4367D3468CB5BE)
  %341 = call double @llvm.fma.f64(double %340, double %334, double noundef 0x3F12471B0E9F1005)
  %342 = call double @llvm.fma.f64(double %341, double %334, double noundef 0x3F49B1004744D5C4)
  %343 = call double @llvm.fma.f64(double %342, double %334, double noundef 0xBF2E13CE69AB4B7F)
  %344 = call double @llvm.fma.f64(double %343, double %334, double noundef 0xBF65F7268ECF8A01)
  %345 = call double @llvm.fma.f64(double %344, double %334, double noundef 0x3F6C71C71C71ACE0)
  %346 = call double @llvm.fma.f64(double %345, double %334, double noundef 0x3FB5555555555556)
  %347 = fmul double %334, %346
  %348 = call double @llvm.fma.f64(double %347, double %.01, double %.01)
  %349 = fmul double %329, %348
  %350 = call double @llvm.fma.f64(double %329, double noundef 0x3FF71547652B82FE, double noundef 0x4338000000000000)
  %351 = call i32 @llvm.nvvm.d2i.lo(double %350) #139
  %352 = fadd double %350, 0xC338000000000000
  %353 = call double @llvm.fma.f64(double %352, double noundef 0xBFE62E42FEFA39EF, double %329)
  %354 = call double @llvm.fma.f64(double %352, double noundef 0xBC7ABC9E3B39803F, double %353)
  %355 = call double @llvm.fma.f64(double %354, double noundef 0x3E5ADE1569CE2BDF, double noundef 0x3E928AF3FCA213EA)
  %356 = call double @llvm.fma.f64(double %355, double %354, double noundef 0x3EC71DEE62401315)
  %357 = call double @llvm.fma.f64(double %356, double %354, double noundef 0x3EFA01997C89EB71)
  %358 = call double @llvm.fma.f64(double %357, double %354, double noundef 0x3F2A01A014761F65)
  %359 = call double @llvm.fma.f64(double %358, double %354, double noundef 0x3F56C16C1852B7AF)
  %360 = call double @llvm.fma.f64(double %359, double %354, double noundef 0x3F81111111122322)
  %361 = call double @llvm.fma.f64(double %360, double %354, double noundef 0x3FA55555555502A1)
  %362 = call double @llvm.fma.f64(double %361, double %354, double noundef 0x3FC5555555555511)
  %363 = call double @llvm.fma.f64(double %362, double %354, double noundef 0x3FE000000000000B)
  %364 = call double @llvm.fma.f64(double %363, double %354, double noundef 1.000000e+00)
  %365 = call double @llvm.fma.f64(double %364, double %354, double noundef 1.000000e+00)
  %366 = call i32 @llvm.abs.i32(i32 %351, i1 noundef false)
  %367 = icmp slt i32 %366, 1023
  br i1 %367, label %368, label %371

368:                                              ; preds = %__nv_sinpi.exit
  %369 = shl i32 %351, 20
  %370 = add i32 %369, 1072693248
  br label %__internal_exp_kernel.exit

371:                                              ; preds = %__nv_sinpi.exit
  %372 = add nsw i32 %351, 2046
  %373 = shl i32 %372, 19
  %374 = and i32 %373, -1048576
  %375 = shl i32 %372, 20
  %376 = sub i32 %375, %374
  %377 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %374) #139
  %378 = fmul double %377, %365
  br label %__internal_exp_kernel.exit

__internal_exp_kernel.exit:                       ; preds = %371, %368
  %.0 = phi double [ %365, %368 ], [ %378, %371 ]
  %k.i.i.0 = phi i32 [ %370, %368 ], [ %376, %371 ]
  %379 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %k.i.i.0) #139
  %380 = fmul double %.0, %379
  %381 = fmul double %380, 0xBC9A6A0D6F814637
  %382 = call double @llvm.fma.f64(double %380, double noundef 0x3FF40D931FF62706, double %381)
  %383 = fdiv double %382, %349
  %384 = fadd double %329, -5.000000e-01
  %385 = fcmp ugt float %2, 0xC00C3C8000000000
  %386 = call i32 @llvm.nvvm.d2i.lo(double %384) #139
  %387 = call i32 @llvm.nvvm.d2i.hi(double %384) #139
  %388 = add i32 %387, -1048576
  %389 = call double @llvm.nvvm.lohi.i2d(i32 %386, i32 %388) #139
  %t.9 = select i1 %385, double %384, double %389
  %390 = fneg double %t.9
  %391 = call fastcc double @__internal_accurate_pow(double %329, double %390) #139
  %392 = fcmp ole float %2, 0xC00C3C8000000000
  %393 = select i1 %392, double %391, double 1.000000e+00
  %s.8 = fmul double %393, %383
  %394 = fmul double %391, %s.8
  br label %405

395:                                              ; preds = %245
  %396 = call double @llvm.floor.f64(double %a)
  %397 = fmul double %396, 5.000000e-01
  %398 = call double @llvm.floor.f64(double %397)
  %399 = fmul double %398, 2.000000e+00
  %400 = fsub double %396, %399
  %401 = fcmp oeq double %400, 1.000000e+00
  %402 = select i1 %401, double -0.000000e+00, double 0.000000e+00
  br label %405

403:                                              ; preds = %208
  %404 = fadd double %a, %a
  br label %405

405:                                              ; preds = %210, %403, %215, %395, %__internal_exp_kernel.exit, %6, %__nv_exp.exit.i
  %s.13 = phi double [ %35, %6 ], [ %r.i.0, %__nv_exp.exit.i ], [ %404, %403 ], [ %244, %215 ], [ %394, %__internal_exp_kernel.exit ], [ %402, %395 ], [ 0xFFF8000000000000, %210 ]
  ret double %s.13
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_lgamma(double %a) local_unnamed_addr #120 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ugt double %1, 0x7FF0000000000000
  br i1 %2, label %163, label %3

3:                                                ; preds = %0
  %4 = call fastcc double @__internal_lgamma_pos(double %1) #143
  %5 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %6 = icmp slt i32 %5, 0
  br i1 %6, label %7, label %165

7:                                                ; preds = %3
  %8 = call double @llvm.trunc.f64(double %1) #140
  %9 = fcmp oeq double %1, %8
  br i1 %9, label %165, label %10

10:                                               ; preds = %7
  %11 = call i32 @llvm.nvvm.d2i.hi(double %1) #142
  %12 = icmp slt i32 %11, 1006632960
  br i1 %12, label %97, label %13

13:                                               ; preds = %10
  %14 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %15 = icmp eq i32 %14, 350
  br i1 %15, label %__nv_sincospi.exit.i, label %16

16:                                               ; preds = %13
  %17 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %18 = icmp eq i32 %17, 370
  br i1 %18, label %__nv_sincospi.exit.i, label %__internal_sin_cos_kerneld.exit.i

__nv_sincospi.exit.i:                             ; preds = %16, %13
  %19 = icmp ugt i32 %11, 1128267776
  %20 = fmul double %1, 0.000000e+00
  %spec.select = select i1 %19, double %20, double %1
  %21 = call i32 @llvm.nvvm.d2i.lo(double %spec.select) #139
  %22 = call i32 @llvm.nvvm.d2i.hi(double %spec.select) #139
  %23 = add i32 %22, 1048576
  %24 = call double @llvm.nvvm.lohi.i2d(i32 %21, i32 %23) #139
  %25 = call double @llvm.round.f64(double %24)
  %26 = fptosi double %25 to i64
  %27 = trunc i64 %26 to i32
  %28 = fneg double %25
  %29 = call double @llvm.fma.f64(double %28, double noundef 5.000000e-01, double %spec.select)
  %30 = fmul double %29, 0x3CA1A62633145C07
  %31 = call double @llvm.fma.f64(double %29, double noundef 0x400921FB54442D18, double %30)
  %32 = fmul double %31, %31
  %33 = call double @llvm.fma.f64(double %32, double noundef 0xBDA8FF8320FD8164, double noundef 0x3E21EEA7C1EF8528)
  %34 = call double @llvm.fma.f64(double %33, double %32, double noundef 0xBE927E4F8E06E6D9)
  %35 = call double @llvm.fma.f64(double %34, double %32, double noundef 0x3EFA01A019DDBCE9)
  %36 = call double @llvm.fma.f64(double %35, double %32, double noundef 0xBF56C16C16C15D47)
  %37 = call double @llvm.fma.f64(double %36, double %32, double noundef 0x3FA5555555555551)
  %38 = call double @llvm.fma.f64(double %37, double %32, double noundef -5.000000e-01)
  %39 = call double @llvm.fma.f64(double %38, double %32, double noundef 1.000000e+00)
  %40 = call double @llvm.fma.f64(double %32, double noundef 0x3DE5DB65F9785EBA, double noundef 0xBE5AE5F12CB0D246)
  %41 = call double @llvm.fma.f64(double %40, double %32, double noundef 0x3EC71DE369ACE392)
  %42 = call double @llvm.fma.f64(double %41, double %32, double noundef 0xBF2A01A019DB62A1)
  %43 = call double @llvm.fma.f64(double %42, double %32, double noundef 0x3F81111111110818)
  %44 = call double @llvm.fma.f64(double %43, double %32, double noundef 0xBFC5555555555554)
  %45 = call double @llvm.fma.f64(double %44, double %32, double noundef 0.000000e+00)
  %46 = call double @llvm.fma.f64(double %45, double %31, double %31)
  %47 = and i32 %27, 1
  %.not4 = icmp eq i32 %47, 0
  %s.i.i.0 = select i1 %.not4, double %46, double %39
  %48 = and i32 %27, 2
  %.not5 = icmp eq i32 %48, 0
  %49 = call i32 @llvm.nvvm.d2i.hi(double %s.i.i.0) #139
  %50 = call i32 @llvm.nvvm.d2i.lo(double %s.i.i.0) #139
  %51 = xor i32 %49, -2147483648
  %52 = call double @llvm.nvvm.lohi.i2d(i32 %50, i32 %51) #139
  %s.i.i.1 = select i1 %.not5, double %s.i.i.0, double %52
  %53 = call double @llvm.trunc.f64(double %spec.select)
  %54 = fcmp oeq double %spec.select, %53
  %55 = fmul double %spec.select, 0.000000e+00
  %s.i.i.2 = select i1 %54, double %55, double %s.i.i.1
  br label %__nv_sinpi.exit

__internal_sin_cos_kerneld.exit.i:                ; preds = %16
  %56 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %57 = add nuw i32 %11, 1048576
  %58 = call double @llvm.nvvm.lohi.i2d(i32 %56, i32 %57) #139
  %59 = call double @llvm.round.f64(double %58)
  %60 = fptosi double %59 to i64
  %61 = trunc i64 %60 to i32
  %62 = fneg double %59
  %63 = call double @llvm.fma.f64(double %62, double noundef 5.000000e-01, double %1)
  %64 = fmul double %63, 0x3CA1A62633145C07
  %65 = call double @llvm.fma.f64(double %63, double noundef 0x400921FB54442D18, double %64)
  %66 = shl i64 %60, 3
  %67 = and i64 %66, 8
  %68 = getelementptr inbounds [16 x double], [16 x double] addrspace(1)* @__cudart_sin_cos_coeffs, i64 0, i64 %67
  %69 = fmul double %65, %65
  %70 = and i32 %61, 1
  %.not = icmp eq i32 %70, 0
  %71 = select i1 %.not, double 0x3DE5DB65F9785EBA, double 0xBDA8FF8320FD8164
  %72 = getelementptr inbounds double, double addrspace(1)* %68, i64 1
  %73 = load double, double addrspace(1)* %72, align 8
  %74 = call double @llvm.fma.f64(double noundef %71, double %69, double %73)
  %75 = getelementptr inbounds double, double addrspace(1)* %68, i64 2
  %76 = load double, double addrspace(1)* %75, align 8
  %77 = call double @llvm.fma.f64(double %74, double %69, double %76)
  %78 = getelementptr inbounds double, double addrspace(1)* %68, i64 3
  %79 = load double, double addrspace(1)* %78, align 8
  %80 = call double @llvm.fma.f64(double %77, double %69, double %79)
  %81 = getelementptr inbounds double, double addrspace(1)* %68, i64 4
  %82 = load double, double addrspace(1)* %81, align 8
  %83 = call double @llvm.fma.f64(double %80, double %69, double %82)
  %84 = getelementptr inbounds double, double addrspace(1)* %68, i64 5
  %85 = load double, double addrspace(1)* %84, align 8
  %86 = call double @llvm.fma.f64(double %83, double %69, double %85)
  %87 = getelementptr inbounds double, double addrspace(1)* %68, i64 6
  %88 = load double, double addrspace(1)* %87, align 8
  %89 = call double @llvm.fma.f64(double %86, double %69, double %88)
  %90 = call double @llvm.fma.f64(double %89, double %65, double %65)
  %91 = call double @llvm.fma.f64(double %89, double %69, double noundef 1.000000e+00)
  %spec.select1 = select i1 %.not, double %90, double %91
  %92 = and i32 %61, 2
  %.not3 = icmp eq i32 %92, 0
  %93 = call double @llvm.fma.f64(double %spec.select1, double noundef -1.000000e+00, double noundef 0.000000e+00)
  %.1 = select i1 %.not3, double %spec.select1, double %93
  br label %__nv_sinpi.exit

__nv_sinpi.exit:                                  ; preds = %__internal_sin_cos_kerneld.exit.i, %__nv_sincospi.exit.i
  %.01 = phi double [ %s.i.i.2, %__nv_sincospi.exit.i ], [ %.1, %__internal_sin_cos_kerneld.exit.i ]
  %94 = fmul double %.01, %a
  %95 = call double @llvm.fabs.f64(double %94)
  %96 = fdiv double 0x400921FB54442D18, %95
  %.pre = call i32 @llvm.nvvm.d2i.hi(double %96) #139
  br label %97

97:                                               ; preds = %10, %__nv_sinpi.exit
  %.pre-phi = phi i32 [ %11, %10 ], [ %.pre, %__nv_sinpi.exit ]
  %u.0 = phi double [ %1, %10 ], [ %96, %__nv_sinpi.exit ]
  %98 = call i32 @llvm.nvvm.d2i.lo(double %u.0) #139
  %99 = icmp slt i32 %.pre-phi, 1048576
  br i1 %99, label %100, label %104

100:                                              ; preds = %97
  %101 = fmul double %u.0, 0x4350000000000000
  %102 = call i32 @llvm.nvvm.d2i.hi(double %101) #139
  %103 = call i32 @llvm.nvvm.d2i.lo(double %101) #139
  br label %104

104:                                              ; preds = %100, %97
  %.0 = phi double [ %101, %100 ], [ %u.0, %97 ]
  %ihi.i.0 = phi i32 [ %102, %100 ], [ %.pre-phi, %97 ]
  %ilo.i.0 = phi i32 [ %103, %100 ], [ %98, %97 ]
  %e.i.0 = phi i32 [ -1077, %100 ], [ -1023, %97 ]
  %105 = add i32 %ihi.i.0, -1
  %106 = icmp ult i32 %105, 2146435071
  br i1 %106, label %107, label %156

107:                                              ; preds = %104
  %108 = lshr i32 %ihi.i.0, 20
  %109 = add nsw i32 %e.i.0, %108
  %110 = and i32 %ihi.i.0, -2146435073
  %111 = or i32 %110, 1072693248
  %112 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %111) #139
  %113 = icmp sgt i32 %111, 1073127582
  br i1 %113, label %114, label %120

114:                                              ; preds = %107
  %115 = call i32 @llvm.nvvm.d2i.lo(double %112) #139
  %116 = call i32 @llvm.nvvm.d2i.hi(double %112) #139
  %117 = add i32 %116, -1048576
  %118 = call double @llvm.nvvm.lohi.i2d(i32 %115, i32 %117) #139
  %119 = add nsw i32 %109, 1
  br label %120

120:                                              ; preds = %114, %107
  %m.i.0 = phi double [ %118, %114 ], [ %112, %107 ]
  %e.i.1 = phi i32 [ %119, %114 ], [ %109, %107 ]
  %121 = fadd double %m.i.0, -1.000000e+00
  %122 = fadd double %m.i.0, 1.000000e+00
  %123 = call double @llvm.nvvm.rcp.approx.ftz.d(double %122) #139
  %124 = fneg double %122
  %125 = call double @llvm.fma.f64(double %124, double %123, double noundef 1.000000e+00)
  %126 = call double @llvm.fma.f64(double %125, double %125, double %125)
  %127 = call double @llvm.fma.f64(double %126, double %123, double %123)
  %128 = fmul double %121, %127
  %129 = fadd double %128, %128
  %130 = fmul double %129, %129
  %131 = call double @llvm.fma.f64(double %130, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04)
  %132 = call double @llvm.fma.f64(double %131, double %130, double noundef 0x3EF3B2669F02676F)
  %133 = call double @llvm.fma.f64(double %132, double %130, double noundef 0x3F1745CBA9AB0956)
  %134 = call double @llvm.fma.f64(double %133, double %130, double noundef 0x3F3C71C72D1B5154)
  %135 = call double @llvm.fma.f64(double %134, double %130, double noundef 0x3F624924923BE72D)
  %136 = call double @llvm.fma.f64(double %135, double %130, double noundef 0x3F8999999999A3C4)
  %137 = call double @llvm.fma.f64(double %136, double %130, double noundef 0x3FB5555555555554)
  %138 = fsub double %121, %129
  %139 = fmul double %138, 2.000000e+00
  %140 = fneg double %129
  %141 = call double @llvm.fma.f64(double %140, double %121, double %139)
  %142 = fmul double %127, %141
  %143 = fmul double %130, %137
  %144 = call double @llvm.fma.f64(double %143, double %129, double %142)
  %145 = xor i32 %e.i.1, -2147483648
  %146 = call double @llvm.nvvm.lohi.i2d(i32 %145, i32 noundef 1127219200) #139
  %147 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #139
  %148 = fsub double %146, %147
  %149 = call double @llvm.fma.f64(double %148, double noundef 0x3FE62E42FEFA39EF, double %129)
  %150 = fneg double %148
  %151 = call double @llvm.fma.f64(double %150, double noundef 0x3FE62E42FEFA39EF, double %149)
  %152 = fsub double %151, %129
  %153 = fsub double %144, %152
  %154 = call double @llvm.fma.f64(double %148, double noundef 0x3C7ABC9E3B39803F, double %153)
  %155 = fadd double %149, %154
  br label %__nv_log.exit

156:                                              ; preds = %104
  %157 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000)
  %158 = call i32 @llvm.nvvm.d2i.hi(double %.0) #139
  %159 = bitcast i32 %158 to float
  %160 = fcmp oeq float %159, 0.000000e+00
  %q.i.0 = select i1 %160, double 0xFFF0000000000000, double %157
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %156, %120
  %q.i.1 = phi double [ %155, %120 ], [ %q.i.0, %156 ]
  %161 = fneg double %q.i.1
  %162 = fsub double %q.i.1, %4
  %t.0 = select i1 %12, double %161, double %162
  br label %165

163:                                              ; preds = %0
  %164 = fadd double %a, %a
  br label %165

165:                                              ; preds = %7, %3, %__nv_log.exit, %163
  %t.3 = phi double [ %164, %163 ], [ %4, %3 ], [ %t.0, %__nv_log.exit ], [ 0x7FF0000000000000, %7 ]
  ret double %t.3
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define internal fastcc double @__internal_lgamma_pos(double %a) unnamed_addr #121 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = icmp sgt i32 %1, 1074266111
  br i1 %2, label %3, label %101

3:                                                ; preds = %0
  %4 = icmp ugt i32 %1, 1075838975
  br i1 %4, label %5, label %83

5:                                                ; preds = %3
  %6 = call double @llvm.nvvm.rcp.approx.ftz.d(double %a) #142
  %7 = fneg double %a
  %8 = call double @llvm.fma.f64(double %7, double %6, double noundef 1.000000e+00) #140
  %9 = call double @llvm.fma.f64(double %8, double %8, double %8) #140
  %10 = call double @llvm.fma.f64(double %9, double %6, double %6) #140
  %11 = fmul double %10, %10
  %12 = call double @llvm.fma.f64(double %11, double noundef 0xBF5AC321034783F9, double noundef 0x3F4B68B992738FBF) #140
  %13 = call double @llvm.fma.f64(double %12, double %11, double noundef 0xBF4380D01E4F7B8C) #140
  %14 = call double @llvm.fma.f64(double %13, double %11, double noundef 0x3F4A019FA29F7264) #140
  %15 = call double @llvm.fma.f64(double %14, double %11, double noundef 0xBF66C16C16B2ACEC) #140
  %16 = call double @llvm.fma.f64(double %15, double %11, double noundef 0x3FB5555555555545) #140
  %17 = call double @llvm.fma.f64(double %16, double %10, double noundef 0x3FED67F1C864BEAE) #140
  %18 = add nsw i32 %1, -1
  %19 = icmp ult i32 %18, 2146435071
  br i1 %19, label %20, label %70

20:                                               ; preds = %5
  %21 = call i32 @llvm.nvvm.d2i.lo(double %a) #142
  %22 = lshr i32 %1, 20
  %23 = add nsw i32 %22, -1023
  %24 = and i32 %1, -2146435073
  %25 = or i32 %24, 1072693248
  %26 = call double @llvm.nvvm.lohi.i2d(i32 %21, i32 %25) #142
  %27 = icmp sgt i32 %25, 1073127582
  br i1 %27, label %28, label %34

28:                                               ; preds = %20
  %29 = call i32 @llvm.nvvm.d2i.lo(double %26) #142
  %30 = call i32 @llvm.nvvm.d2i.hi(double %26) #142
  %31 = add i32 %30, -1048576
  %32 = call double @llvm.nvvm.lohi.i2d(i32 %29, i32 %31) #142
  %33 = add nsw i32 %22, -1022
  br label %34

34:                                               ; preds = %28, %20
  %m.i10.0 = phi double [ %32, %28 ], [ %26, %20 ]
  %e.i22.1 = phi i32 [ %33, %28 ], [ %23, %20 ]
  %35 = fadd double %m.i10.0, -1.000000e+00
  %36 = fadd double %m.i10.0, 1.000000e+00
  %37 = call double @llvm.nvvm.rcp.approx.ftz.d(double %36) #142
  %38 = fneg double %36
  %39 = call double @llvm.fma.f64(double %38, double %37, double noundef 1.000000e+00) #140
  %40 = call double @llvm.fma.f64(double %39, double %39, double %39) #140
  %41 = call double @llvm.fma.f64(double %40, double %37, double %37) #140
  %42 = fmul double %35, %41
  %43 = fadd double %42, %42
  %44 = fmul double %43, %43
  %45 = call double @llvm.fma.f64(double %44, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %46 = call double @llvm.fma.f64(double %45, double %44, double noundef 0x3EF3B2669F02676F) #140
  %47 = call double @llvm.fma.f64(double %46, double %44, double noundef 0x3F1745CBA9AB0956) #140
  %48 = call double @llvm.fma.f64(double %47, double %44, double noundef 0x3F3C71C72D1B5154) #140
  %49 = call double @llvm.fma.f64(double %48, double %44, double noundef 0x3F624924923BE72D) #140
  %50 = call double @llvm.fma.f64(double %49, double %44, double noundef 0x3F8999999999A3C4) #140
  %51 = call double @llvm.fma.f64(double %50, double %44, double noundef 0x3FB5555555555554) #140
  %52 = fsub double %35, %43
  %53 = fmul double %52, 2.000000e+00
  %54 = fneg double %43
  %55 = call double @llvm.fma.f64(double %54, double %35, double %53) #140
  %56 = fmul double %41, %55
  %57 = fmul double %44, %51
  %58 = call double @llvm.fma.f64(double %57, double %43, double %56) #140
  %59 = xor i32 %e.i22.1, -2147483648
  %60 = call double @llvm.nvvm.lohi.i2d(i32 %59, i32 noundef 1127219200) #142
  %61 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %62 = fsub double %60, %61
  %63 = call double @llvm.fma.f64(double %62, double noundef 0x3FE62E42FEFA39EF, double %43) #140
  %64 = fneg double %62
  %65 = call double @llvm.fma.f64(double %64, double noundef 0x3FE62E42FEFA39EF, double %63) #140
  %66 = fsub double %65, %43
  %67 = fsub double %58, %66
  %68 = call double @llvm.fma.f64(double %62, double noundef 0x3C7ABC9E3B39803F, double %67) #140
  %69 = fadd double %63, %68
  br label %__nv_log.exit23

70:                                               ; preds = %5
  %71 = call double @llvm.fma.f64(double %a, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %72 = bitcast i32 %1 to float
  %73 = fcmp oeq float %72, 0.000000e+00
  %q.i16.0 = select i1 %73, double 0xFFF0000000000000, double %71
  br label %__nv_log.exit23

__nv_log.exit23:                                  ; preds = %70, %34
  %q.i16.1 = phi double [ %69, %34 ], [ %q.i16.0, %70 ]
  %74 = call i32 @llvm.nvvm.d2i.lo(double %q.i16.1) #142
  %75 = call i32 @llvm.nvvm.d2i.hi(double %q.i16.1) #142
  %76 = add i32 %75, -1048576
  %77 = call double @llvm.nvvm.lohi.i2d(i32 %74, i32 %76) #142
  %78 = fadd double %a, -5.000000e-01
  %79 = call double @llvm.fma.f64(double %77, double %78, double %17) #140
  %80 = call double @llvm.fma.f64(double %77, double %78, double %7) #140
  %81 = fadd double %80, %79
  %82 = fcmp oeq double %a, 0x7FF0000000000000
  %t.0 = select i1 %82, double %a, double %81
  br label %232

83:                                               ; preds = %3
  %84 = fadd double %a, -3.000000e+00
  %85 = call double @llvm.fma.f64(double %84, double noundef 0xC0AF7040BB18FB05, double noundef 0xC1122B7730207EF3) #140
  %86 = call double @llvm.fma.f64(double %85, double %84, double noundef 0xC1585A0DB81DE7D0) #140
  %87 = call double @llvm.fma.f64(double %86, double %84, double noundef 0xC18A992B8BA94677) #140
  %88 = call double @llvm.fma.f64(double %87, double %84, double noundef 0xC1AAC5CB6957CC20) #140
  %89 = call double @llvm.fma.f64(double %88, double %84, double noundef 0xC1BC0E2B308774BE) #140
  %90 = call double @llvm.fma.f64(double %89, double %84, double noundef 0xC1C6BA13DCAE7F67) #140
  %91 = call double @llvm.fma.f64(double %90, double %84, double noundef 0xC1CCF33B9C3D120C) #140
  %92 = fadd double %84, 0xC08FF62E0BE189FE
  %93 = call double @llvm.fma.f64(double %92, double %84, double noundef 0xC10074FACE10C93F) #140
  %94 = call double @llvm.fma.f64(double %93, double %84, double noundef 0xC151B662F8D75791) #140
  %95 = call double @llvm.fma.f64(double %94, double %84, double noundef 0xC18EE64AB4D207F7) #140
  %96 = call double @llvm.fma.f64(double %95, double %84, double noundef 0xC1B9051687C9951A) #140
  %97 = call double @llvm.fma.f64(double %96, double %84, double noundef 0xC1D2B866BF0B853D) #140
  %98 = call double @llvm.fma.f64(double %97, double %84, double noundef 0xC1D4E2130E9DC133) #140
  %99 = fdiv double %91, %98
  %100 = fadd double %84, %99
  br label %232

101:                                              ; preds = %0
  %102 = icmp sgt i32 %1, 1073217535
  br i1 %102, label %103, label %125

103:                                              ; preds = %101
  %104 = fadd double %a, -2.000000e+00
  %105 = call double @llvm.fma.f64(double %104, double noundef 0x3E452636124338B3, double noundef 0xBE71FA71D78C0EE2) #140
  %106 = call double @llvm.fma.f64(double %105, double %104, double noundef 0x3E8D111F31E61306) #140
  %107 = call double @llvm.fma.f64(double %106, double %104, double noundef 0xBEA0502BBE1B2706) #140
  %108 = call double @llvm.fma.f64(double %107, double %104, double noundef 0x3EB06850B2970292) #140
  %109 = call double @llvm.fma.f64(double %108, double %104, double noundef 0xBEC108474875033D) #140
  %110 = call double @llvm.fma.f64(double %109, double %104, double noundef 0x3ED24ACCC62909DC) #140
  %111 = call double @llvm.fma.f64(double %110, double %104, double noundef 0xBEE3CB25209E63BE) #140
  %112 = call double @llvm.fma.f64(double %111, double %104, double noundef 0x3EF581CBBC8CDC7B) #140
  %113 = call double @llvm.fma.f64(double %112, double %104, double noundef 0xBF078E04B85C7597) #140
  %114 = call double @llvm.fma.f64(double %113, double %104, double noundef 0x3F1A12730CF45051) #140
  %115 = call double @llvm.fma.f64(double %114, double %104, double noundef 0xBF2D3FD354062012) #140
  %116 = call double @llvm.fma.f64(double %115, double %104, double noundef 0x3F40B36B0B4DE323) #140
  %117 = call double @llvm.fma.f64(double %116, double %104, double noundef 0xBF538AC5C6D0317A) #140
  %118 = call double @llvm.fma.f64(double %117, double %104, double noundef 0x3F67ADD6EAAB19FC) #140
  %119 = call double @llvm.fma.f64(double %118, double %104, double noundef 0xBF7E404FC20E4D5B) #140
  %120 = call double @llvm.fma.f64(double %119, double %104, double noundef 0x3F951322AC7DA390) #140
  %121 = call double @llvm.fma.f64(double %120, double %104, double noundef 0xBFB13E001A5578A3) #140
  %122 = call double @llvm.fma.f64(double %121, double %104, double noundef 0x3FD4A34CC4A60FA3) #140
  %123 = call double @llvm.fma.f64(double %122, double %104, double noundef 0x3FDB0EE6072093CF) #140
  %124 = fmul double %104, %123
  br label %232

125:                                              ; preds = %101
  %126 = icmp sgt i32 %1, 1072064101
  br i1 %126, label %127, label %151

127:                                              ; preds = %125
  %128 = fsub double 1.000000e+00, %a
  %129 = call double @llvm.fma.f64(double %128, double noundef 0x3F881F6D2A4C4310, double noundef 0x3FA3EB504359EB88) #140
  %130 = call double @llvm.fma.f64(double %129, double %128, double noundef 0x3FAE35D8DEB06317) #140
  %131 = call double @llvm.fma.f64(double %130, double %128, double noundef 0x3FAED469A8B6ECCE) #140
  %132 = call double @llvm.fma.f64(double %131, double %128, double noundef 0x3FACC1B1C357BEFE) #140
  %133 = call double @llvm.fma.f64(double %132, double %128, double noundef 0x3FAD7154DB67F79F) #140
  %134 = call double @llvm.fma.f64(double %133, double %128, double noundef 0x3FAFCC622CF2F7BB) #140
  %135 = call double @llvm.fma.f64(double %134, double %128, double noundef 0x3FB11747A4D1CC43) #140
  %136 = call double @llvm.fma.f64(double %135, double %128, double noundef 0x3FB24CE16A21B8AC) #140
  %137 = call double @llvm.fma.f64(double %136, double %128, double noundef 0x3FB3B1C21A7BCB00) #140
  %138 = call double @llvm.fma.f64(double %137, double %128, double noundef 0x3FB556723452ED57) #140
  %139 = call double @llvm.fma.f64(double %138, double %128, double noundef 0x3FB748C00891544F) #140
  %140 = call double @llvm.fma.f64(double %139, double %128, double noundef 0x3FB9A0207808CF40) #140
  %141 = call double @llvm.fma.f64(double %140, double %128, double noundef 0x3FBC80673B8AE26B) #140
  %142 = call double @llvm.fma.f64(double %141, double %128, double noundef 0x3FC010B364B7E555) #140
  %143 = call double @llvm.fma.f64(double %142, double %128, double noundef 0x3FC2703A1D239658) #140
  %144 = call double @llvm.fma.f64(double %143, double %128, double noundef 0x3FC5B40CB1137E6E) #140
  %145 = call double @llvm.fma.f64(double %144, double %128, double noundef 0x3FCA8B9C17AC4F03) #140
  %146 = call double @llvm.fma.f64(double %145, double %128, double noundef 0x3FD151322AC7CB52) #140
  %147 = call double @llvm.fma.f64(double %146, double %128, double noundef 0x3FD9A4D55BEAB1D4) #140
  %148 = call double @llvm.fma.f64(double %147, double %128, double noundef 0x3FEA51A6625307D6) #140
  %149 = call double @llvm.fma.f64(double %148, double %128, double noundef 0x3FE2788CFC6FB619) #140
  %150 = fmul double %128, %149
  br label %232

151:                                              ; preds = %125
  %152 = call double @llvm.fma.f64(double %a, double noundef 0xBE7844988BFE6590, double noundef 0x3EA7B77CEB0625E8) #140
  %153 = call double @llvm.fma.f64(double %152, double %a, double noundef 0xBE998C69C8710CC4) #140
  %154 = call double @llvm.fma.f64(double %153, double %a, double noundef 0xBEF6527A5A11CF6E) #140
  %155 = call double @llvm.fma.f64(double %154, double %a, double noundef 0x3F20EC2950B1B5DE) #140
  %156 = call double @llvm.fma.f64(double %155, double %a, double noundef 0xBF2C4D80C24BA278) #140
  %157 = call double @llvm.fma.f64(double %156, double %a, double noundef 0xBF5315B4E8CC0D09) #140
  %158 = call double @llvm.fma.f64(double %157, double %a, double noundef 0x3F7D917F15D50020) #140
  %159 = call double @llvm.fma.f64(double %158, double %a, double noundef 0xBF83B4ABB41CB6FA) #140
  %160 = call double @llvm.fma.f64(double %159, double %a, double noundef 0xBFA59AF1275B7120) #140
  %161 = call double @llvm.fma.f64(double %160, double %a, double noundef 0x3FC5512321A168A0) #140
  %162 = call double @llvm.fma.f64(double %161, double %a, double noundef 0xBFA5815E8FDCE74C) #140
  %163 = call double @llvm.fma.f64(double %162, double %a, double noundef 0xBFE4FCF4026ADD1A) #140
  %164 = call double @llvm.fma.f64(double %163, double %a, double noundef 0x3FE2788CFC6FB5C8) #140
  %165 = fmul double %164, %a
  %166 = call double @llvm.fma.f64(double %165, double %a, double %a) #140
  %167 = call i32 @llvm.nvvm.d2i.hi(double %166) #142
  %168 = call i32 @llvm.nvvm.d2i.lo(double %166) #142
  %169 = icmp slt i32 %167, 1048576
  br i1 %169, label %170, label %174

170:                                              ; preds = %151
  %171 = fmul double %166, 0x4350000000000000
  %172 = call i32 @llvm.nvvm.d2i.hi(double %171) #142
  %173 = call i32 @llvm.nvvm.d2i.lo(double %171) #142
  br label %174

174:                                              ; preds = %170, %151
  %.0 = phi double [ %171, %170 ], [ %166, %151 ]
  %ihi.i.0 = phi i32 [ %172, %170 ], [ %167, %151 ]
  %ilo.i.0 = phi i32 [ %173, %170 ], [ %168, %151 ]
  %e.i.0 = phi i32 [ -1077, %170 ], [ -1023, %151 ]
  %175 = add i32 %ihi.i.0, -1
  %176 = icmp ult i32 %175, 2146435071
  br i1 %176, label %177, label %226

177:                                              ; preds = %174
  %178 = lshr i32 %ihi.i.0, 20
  %179 = add nsw i32 %e.i.0, %178
  %180 = and i32 %ihi.i.0, -2146435073
  %181 = or i32 %180, 1072693248
  %182 = call double @llvm.nvvm.lohi.i2d(i32 %ilo.i.0, i32 %181) #142
  %183 = icmp sgt i32 %181, 1073127582
  br i1 %183, label %184, label %190

184:                                              ; preds = %177
  %185 = call i32 @llvm.nvvm.d2i.lo(double %182) #142
  %186 = call i32 @llvm.nvvm.d2i.hi(double %182) #142
  %187 = add i32 %186, -1048576
  %188 = call double @llvm.nvvm.lohi.i2d(i32 %185, i32 %187) #142
  %189 = add nsw i32 %179, 1
  br label %190

190:                                              ; preds = %184, %177
  %m.i.0 = phi double [ %188, %184 ], [ %182, %177 ]
  %e.i.1 = phi i32 [ %189, %184 ], [ %179, %177 ]
  %191 = fadd double %m.i.0, -1.000000e+00
  %192 = fadd double %m.i.0, 1.000000e+00
  %193 = call double @llvm.nvvm.rcp.approx.ftz.d(double %192) #142
  %194 = fneg double %192
  %195 = call double @llvm.fma.f64(double %194, double %193, double noundef 1.000000e+00) #140
  %196 = call double @llvm.fma.f64(double %195, double %195, double %195) #140
  %197 = call double @llvm.fma.f64(double %196, double %193, double %193) #140
  %198 = fmul double %191, %197
  %199 = fadd double %198, %198
  %200 = fmul double %199, %199
  %201 = call double @llvm.fma.f64(double %200, double noundef 0x3EB1380B3AE80F1E, double noundef 0x3ED0EE258B7A8B04) #140
  %202 = call double @llvm.fma.f64(double %201, double %200, double noundef 0x3EF3B2669F02676F) #140
  %203 = call double @llvm.fma.f64(double %202, double %200, double noundef 0x3F1745CBA9AB0956) #140
  %204 = call double @llvm.fma.f64(double %203, double %200, double noundef 0x3F3C71C72D1B5154) #140
  %205 = call double @llvm.fma.f64(double %204, double %200, double noundef 0x3F624924923BE72D) #140
  %206 = call double @llvm.fma.f64(double %205, double %200, double noundef 0x3F8999999999A3C4) #140
  %207 = call double @llvm.fma.f64(double %206, double %200, double noundef 0x3FB5555555555554) #140
  %208 = fsub double %191, %199
  %209 = fmul double %208, 2.000000e+00
  %210 = fneg double %199
  %211 = call double @llvm.fma.f64(double %210, double %191, double %209) #140
  %212 = fmul double %197, %211
  %213 = fmul double %200, %207
  %214 = call double @llvm.fma.f64(double %213, double %199, double %212) #140
  %215 = xor i32 %e.i.1, -2147483648
  %216 = call double @llvm.nvvm.lohi.i2d(i32 %215, i32 noundef 1127219200) #142
  %217 = call double @llvm.nvvm.lohi.i2d(i32 noundef -2147483648, i32 noundef 1127219200) #142
  %218 = fsub double %216, %217
  %219 = call double @llvm.fma.f64(double %218, double noundef 0x3FE62E42FEFA39EF, double %199) #140
  %220 = fneg double %218
  %221 = call double @llvm.fma.f64(double %220, double noundef 0x3FE62E42FEFA39EF, double %219) #140
  %222 = fsub double %221, %199
  %223 = fsub double %214, %222
  %224 = call double @llvm.fma.f64(double %218, double noundef 0x3C7ABC9E3B39803F, double %223) #140
  %225 = fadd double %219, %224
  br label %__nv_log.exit

226:                                              ; preds = %174
  %227 = call double @llvm.fma.f64(double %.0, double noundef 0x7FF0000000000000, double noundef 0x7FF0000000000000) #140
  %228 = call i32 @llvm.nvvm.d2i.hi(double %.0) #142
  %229 = bitcast i32 %228 to float
  %230 = fcmp oeq float %229, 0.000000e+00
  %q.i.0 = select i1 %230, double 0xFFF0000000000000, double %227
  br label %__nv_log.exit

__nv_log.exit:                                    ; preds = %226, %190
  %q.i.1 = phi double [ %225, %190 ], [ %q.i.0, %226 ]
  %231 = fneg double %q.i.1
  br label %232

232:                                              ; preds = %103, %__nv_log.exit, %127, %__nv_log.exit23, %83
  %t.4 = phi double [ %t.0, %__nv_log.exit23 ], [ %100, %83 ], [ %124, %103 ], [ %150, %127 ], [ %231, %__nv_log.exit ]
  ret double %t.4
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_ldexp(double %a, i32 %b) local_unnamed_addr #122 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = icmp slt i32 %b, 2200
  %min = select i1 %2, i32 %b, i32 2200
  %max.cond.inv = icmp sgt i32 %min, -2200
  %max = select i1 %max.cond.inv, i32 %min, i32 -2200
  %3 = fcmp oeq double %a, 0.000000e+00
  %4 = fcmp oeq double %1, 0x7FF0000000000000
  %or.cond = or i1 %3, %4
  %5 = icmp eq i32 %max, 0
  %or.cond2 = select i1 %or.cond, i1 true, i1 %5
  br i1 %or.cond2, label %6, label %9

6:                                                ; preds = %0
  %7 = fcmp one double %a, 0.000000e+00
  %8 = select i1 %7, double -0.000000e+00, double %a
  %.0 = fadd double %8, %a
  br label %43

9:                                                ; preds = %0
  %10 = call i32 @llvm.abs.i32(i32 %max, i1 noundef true) #140
  %11 = icmp ult i32 %10, 1022
  br i1 %11, label %12, label %17

12:                                               ; preds = %9
  %13 = shl i32 %max, 20
  %14 = add i32 %13, 1072693248
  %15 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %14) #142
  %16 = fmul double %15, %a
  br label %43

17:                                               ; preds = %9
  %18 = icmp ult i32 %10, 2044
  br i1 %18, label %19, label %30

19:                                               ; preds = %17
  %.lhs.trunc = trunc i32 %max to i16
  %20 = sdiv i16 %.lhs.trunc, 2
  %.sext = sext i16 %20 to i32
  %21 = shl nsw i32 %.sext, 20
  %22 = add i32 %21, 1072693248
  %23 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %22) #142
  %24 = fmul double %23, %a
  %25 = sub nsw i32 %max, %.sext
  %26 = shl i32 %25, 20
  %27 = add i32 %26, 1072693248
  %28 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %27) #142
  %29 = fmul double %24, %28
  br label %43

30:                                               ; preds = %17
  %.lhs.trunc3 = trunc i32 %max to i16
  %31 = sdiv i16 %.lhs.trunc3, 4
  %.sext4 = sext i16 %31 to i32
  %32 = shl nsw i32 %.sext4, 20
  %33 = add nsw i32 %32, 1072693248
  %34 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %33) #142
  %35 = fmul double %34, %a
  %36 = fmul double %34, %35
  %37 = fmul double %34, %36
  %.neg = mul nsw i32 %.sext4, -3
  %38 = add nsw i32 %.neg, %max
  %39 = shl i32 %38, 20
  %40 = add i32 %39, 1072693248
  %41 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %40) #142
  %42 = fmul double %41, %37
  br label %43

43:                                               ; preds = %12, %30, %19, %6
  %.3 = phi double [ %.0, %6 ], [ %16, %12 ], [ %29, %19 ], [ %42, %30 ]
  ret double %.3
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_scalbn(double %a, i32 %b) local_unnamed_addr #123 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = icmp slt i32 %b, 2200
  %min = select i1 %2, i32 %b, i32 2200
  %max.cond.inv = icmp sgt i32 %min, -2200
  %max = select i1 %max.cond.inv, i32 %min, i32 -2200
  %3 = fcmp oeq double %a, 0.000000e+00
  %4 = fcmp oeq double %1, 0x7FF0000000000000
  %or.cond = or i1 %3, %4
  %5 = icmp eq i32 %max, 0
  %or.cond2 = select i1 %or.cond, i1 true, i1 %5
  br i1 %or.cond2, label %6, label %9

6:                                                ; preds = %0
  %7 = fcmp one double %a, 0.000000e+00
  %8 = select i1 %7, double -0.000000e+00, double %a
  %.0 = fadd double %8, %a
  br label %__nv_ldexp.exit

9:                                                ; preds = %0
  %10 = call i32 @llvm.abs.i32(i32 %max, i1 noundef true) #140
  %11 = icmp ult i32 %10, 1022
  br i1 %11, label %12, label %17

12:                                               ; preds = %9
  %13 = shl i32 %max, 20
  %14 = add i32 %13, 1072693248
  %15 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %14) #142
  %16 = fmul double %15, %a
  br label %__nv_ldexp.exit

17:                                               ; preds = %9
  %18 = icmp ult i32 %10, 2044
  br i1 %18, label %19, label %30

19:                                               ; preds = %17
  %.lhs.trunc = trunc i32 %max to i16
  %20 = sdiv i16 %.lhs.trunc, 2
  %.sext = sext i16 %20 to i32
  %21 = shl nsw i32 %.sext, 20
  %22 = add i32 %21, 1072693248
  %23 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %22) #142
  %24 = fmul double %23, %a
  %25 = sub nsw i32 %max, %.sext
  %26 = shl i32 %25, 20
  %27 = add i32 %26, 1072693248
  %28 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %27) #142
  %29 = fmul double %24, %28
  br label %__nv_ldexp.exit

30:                                               ; preds = %17
  %.lhs.trunc3 = trunc i32 %max to i16
  %31 = sdiv i16 %.lhs.trunc3, 4
  %.sext4 = sext i16 %31 to i32
  %32 = shl nsw i32 %.sext4, 20
  %33 = add nsw i32 %32, 1072693248
  %34 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %33) #142
  %35 = fmul double %34, %a
  %36 = fmul double %34, %35
  %37 = fmul double %34, %36
  %.neg = mul nsw i32 %.sext4, -3
  %38 = add nsw i32 %.neg, %max
  %39 = shl i32 %38, 20
  %40 = add i32 %39, 1072693248
  %41 = call double @llvm.nvvm.lohi.i2d(i32 noundef 0, i32 %40) #142
  %42 = fmul double %41, %37
  br label %__nv_ldexp.exit

__nv_ldexp.exit:                                  ; preds = %12, %30, %19, %6
  %.3 = phi double [ %.0, %6 ], [ %16, %12 ], [ %29, %19 ], [ %42, %30 ]
  ret double %.3
}

; Function Attrs: argmemonly nofree noinline nosync nounwind willreturn writeonly
define dso_local double @__nv_frexp(double %a, i32* nocapture nofree nonnull writeonly align 4 dereferenceable(4) %b) local_unnamed_addr #124 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = and i32 %1, 2146435072
  %3 = icmp eq i32 %2, 0
  %4 = fmul double %a, 0x4350000000000000
  %5 = call i32 @llvm.nvvm.d2i.hi(double %4) #142
  %6 = and i32 %5, 2146435072
  %7 = add nsw i32 %6, -56623104
  %.0 = select i1 %3, double %4, double %a
  %expo.0 = select i1 %3, i32 %7, i32 %2
  %8 = fcmp oeq double %.0, 0.000000e+00
  %9 = icmp eq i32 %expo.0, 2146435072
  %or.cond = select i1 %8, i1 true, i1 %9
  br i1 %or.cond, label %10, label %12

10:                                               ; preds = %0
  %11 = fadd double %.0, %.0
  br label %19

12:                                               ; preds = %0
  %13 = bitcast double %.0 to i64
  %14 = and i64 %13, -9218868437227405313
  %15 = or i64 %14, 4602678819172646912
  %16 = bitcast i64 %15 to double
  %17 = add nsw i32 %expo.0, -1071644672
  %18 = ashr exact i32 %17, 20
  br label %19

19:                                               ; preds = %12, %10
  %.1 = phi double [ %11, %10 ], [ %16, %12 ]
  %expo.1 = phi i32 [ 0, %10 ], [ %18, %12 ]
  store i32 %expo.1, i32* %b, align 4, !noalias !68
  ret double %.1
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_modf(double %a, double* nocapture nofree writeonly %b) local_unnamed_addr #125 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = and i32 %1, 2146435072
  %.not = icmp eq i32 %2, 2146435072
  br i1 %.not, label %11, label %3

3:                                                ; preds = %0
  %4 = call double @llvm.trunc.f64(double %a) #140
  store double %4, double* %b, align 8, !noalias !68
  %5 = fsub double %a, %4
  %6 = call i32 @llvm.nvvm.d2i.lo(double %5) #142
  %7 = call i32 @llvm.nvvm.d2i.hi(double %5) #142
  %8 = and i32 %1, -2147483648
  %9 = or i32 %7, %8
  %10 = call double @llvm.nvvm.lohi.i2d(i32 %6, i32 %9) #142
  br label %37

11:                                               ; preds = %0
  %12 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %13 = icmp eq i32 %12, 200
  br i1 %13, label %20, label %14

14:                                               ; preds = %11
  %15 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %16 = icmp eq i32 %15, 350
  br i1 %16, label %20, label %17

17:                                               ; preds = %14
  %18 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %19 = icmp eq i32 %18, 370
  br i1 %19, label %20, label %23

20:                                               ; preds = %17, %14, %11
  %21 = call double @llvm.fabs.f64(double %a)
  %22 = fcmp oeq double %21, 0x7FF0000000000000
  br label %__nv_isinfd.exit

23:                                               ; preds = %17
  %24 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %25 = and i32 %1, 2147483647
  %26 = icmp eq i32 %25, 2146435072
  %27 = icmp eq i32 %24, 0
  %28 = select i1 %26, i1 %27, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %23, %20
  %.01.in = phi i1 [ %22, %20 ], [ %28, %23 ]
  br i1 %.01.in, label %29, label %35

29:                                               ; preds = %__nv_isinfd.exit
  store double %a, double* %b, align 8, !noalias !68
  %30 = call i32 @llvm.nvvm.d2i.lo(double noundef 0.000000e+00) #139
  %31 = call i32 @llvm.nvvm.d2i.hi(double noundef 0.000000e+00) #139
  %32 = and i32 %1, -2147483648
  %33 = or i32 %32, %31
  %34 = call double @llvm.nvvm.lohi.i2d(i32 %30, i32 %33) #139
  br label %37

35:                                               ; preds = %__nv_isinfd.exit
  %36 = fadd double %a, %a
  store double %36, double* %b, align 8, !noalias !68
  br label %37

37:                                               ; preds = %29, %35, %3
  %.1 = phi double [ %10, %3 ], [ %34, %29 ], [ %36, %35 ]
  ret double %.1
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_fmod(double %a, double %b) local_unnamed_addr #126 {
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2 = icmp eq i32 %1, 200
  br i1 %2, label %9, label %3

3:                                                ; preds = %0
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %5 = icmp eq i32 %4, 350
  br i1 %5, label %9, label %6

6:                                                ; preds = %3
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %8 = icmp eq i32 %7, 370
  br i1 %8, label %9, label %60

9:                                                ; preds = %6, %3, %0
  %10 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %11 = and i32 %10, 2147483647
  %12 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %13 = call i32 @llvm.nvvm.d2i.hi(double %b) #139
  %14 = and i32 %13, 2147483647
  %15 = call i32 @llvm.nvvm.d2i.lo(double %b) #139
  %16 = call double @llvm.nvvm.lohi.i2d(i32 %12, i32 %11) #139
  %17 = call double @llvm.nvvm.lohi.i2d(i32 %15, i32 %14) #139
  %18 = icmp ugt i32 %11, 2146435071
  %19 = icmp ugt i32 %14, 2146435071
  %or.cond = select i1 %18, i1 true, i1 %19
  br i1 %or.cond, label %20, label %27

20:                                               ; preds = %9
  %21 = fcmp ole double %16, 0x7FF0000000000000
  %22 = fcmp ole double %17, 0x7FF0000000000000
  %or.cond2 = select i1 %21, i1 %22, i1 false
  br i1 %or.cond2, label %25, label %23

23:                                               ; preds = %20
  %24 = fadd double %a, %b
  br label %146

25:                                               ; preds = %20
  %26 = fcmp oeq double %16, 0x7FF0000000000000
  %.01 = select i1 %26, double 0xFFF8000000000000, double %a
  br label %146

27:                                               ; preds = %9
  %28 = fcmp oeq double %17, 0.000000e+00
  br i1 %28, label %146, label %29

29:                                               ; preds = %27
  %30 = fcmp ult double %16, %17
  br i1 %30, label %146, label %31

31:                                               ; preds = %29
  %32 = call i32 @llvm.nvvm.d2i.hi(double %17) #139
  %33 = icmp slt i32 %32, 1048576
  br i1 %33, label %.preheader, label %41

.preheader:                                       ; preds = %31
  %34 = fcmp olt double %17, %16
  br i1 %34, label %.lr.ph9, label %._crit_edge10

.lr.ph9:                                          ; preds = %.preheader, %.lr.ph9
  %t.08 = phi double [ %35, %.lr.ph9 ], [ %17, %.preheader ]
  %35 = fadd double %t.08, %t.08
  %36 = fcmp olt double %35, %16
  %37 = call i32 @llvm.nvvm.d2i.hi(double %35) #139
  %38 = icmp slt i32 %37, 1048576
  %39 = select i1 %36, i1 %38, i1 false
  br i1 %39, label %.lr.ph9, label %._crit_edge10

._crit_edge10:                                    ; preds = %.lr.ph9, %.preheader
  %t.0.lcssa = phi double [ %17, %.preheader ], [ %35, %.lr.ph9 ]
  %.lcssa = phi i32 [ %32, %.preheader ], [ %37, %.lr.ph9 ]
  %40 = call i32 @llvm.nvvm.d2i.lo(double %t.0.lcssa) #139
  br label %41

41:                                               ; preds = %._crit_edge10, %31
  %bhi.0 = phi i32 [ %.lcssa, %._crit_edge10 ], [ %14, %31 ]
  %blo.0 = phi i32 [ %40, %._crit_edge10 ], [ %15, %31 ]
  %scaled_b.0 = phi double [ %t.0.lcssa, %._crit_edge10 ], [ 0.000000e+00, %31 ]
  %42 = call i32 @llvm.nvvm.d2i.hi(double %16) #139
  %43 = icmp sgt i32 %42, 1048575
  %44 = and i32 %bhi.0, 1048575
  %45 = and i32 %10, 2146435072
  %46 = or i32 %44, %45
  %47 = call double @llvm.nvvm.lohi.i2d(i32 %blo.0, i32 %46) #139
  %scaled_b.1 = select i1 %43, double %47, double %scaled_b.0
  %48 = fcmp ogt double %scaled_b.1, %16
  %49 = fmul double %scaled_b.1, 5.000000e-01
  %scaled_b.2 = select i1 %48, double %49, double %scaled_b.1
  %50 = fcmp ult double %scaled_b.2, %17
  br i1 %50, label %55, label %.lr.ph

.lr.ph:                                           ; preds = %41, %.lr.ph
  %scaled_b.37 = phi double [ %53, %.lr.ph ], [ %scaled_b.2, %41 ]
  %.26 = phi double [ %.3, %.lr.ph ], [ %16, %41 ]
  %51 = fcmp oge double %.26, %scaled_b.37
  %52 = select i1 %51, double %scaled_b.37, double 0.000000e+00
  %.3 = fsub double %.26, %52
  %53 = fmul double %scaled_b.37, 5.000000e-01
  %54 = fcmp ult double %53, %17
  br i1 %54, label %._crit_edge, label %.lr.ph

._crit_edge:                                      ; preds = %.lr.ph
  %.pre = call i32 @llvm.nvvm.d2i.hi(double %.3) #139
  br label %55

55:                                               ; preds = %._crit_edge, %41
  %.pre-phi = phi i32 [ %.pre, %._crit_edge ], [ %42, %41 ]
  %.2.lcssa = phi double [ %.3, %._crit_edge ], [ %16, %41 ]
  %56 = call i32 @llvm.nvvm.d2i.lo(double %.2.lcssa) #139
  %57 = and i32 %10, -2147483648
  %58 = or i32 %.pre-phi, %57
  %59 = call double @llvm.nvvm.lohi.i2d(i32 %56, i32 %58) #139
  br label %146

60:                                               ; preds = %6
  %61 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %62 = and i32 %61, 2147483647
  %63 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %64 = call i32 @llvm.nvvm.d2i.hi(double %b) #139
  %65 = and i32 %64, 2147483647
  %66 = call i32 @llvm.nvvm.d2i.lo(double %b) #139
  %67 = call double @llvm.nvvm.lohi.i2d(i32 %63, i32 %62) #139
  %68 = call double @llvm.nvvm.lohi.i2d(i32 %66, i32 %65) #139
  %69 = icmp ugt i32 %62, 2146435071
  %70 = icmp ugt i32 %65, 2146435071
  %or.cond3 = select i1 %69, i1 true, i1 %70
  br i1 %or.cond3, label %71, label %78

71:                                               ; preds = %60
  %72 = fcmp ole double %67, 0x7FF0000000000000
  %73 = fcmp ole double %68, 0x7FF0000000000000
  %or.cond5 = select i1 %72, i1 %73, i1 false
  br i1 %or.cond5, label %76, label %74

74:                                               ; preds = %71
  %75 = fadd double %a, %b
  br label %146

76:                                               ; preds = %71
  %77 = fcmp oeq double %67, 0x7FF0000000000000
  %.7 = select i1 %77, double 0xFFF8000000000000, double %a
  br label %146

78:                                               ; preds = %60
  %79 = fcmp oeq double %68, 0.000000e+00
  br i1 %79, label %146, label %80

80:                                               ; preds = %78
  %81 = fcmp ult double %67, %68
  br i1 %81, label %146, label %82

82:                                               ; preds = %80
  %83 = call i32 @llvm.nvvm.d2i.hi(double %67) #139
  %84 = lshr i32 %83, 20
  %85 = call i32 @llvm.nvvm.d2i.hi(double %68) #139
  %86 = lshr i32 %85, 20
  %87 = icmp ult i32 %83, 1048576
  br i1 %87, label %88, label %94

88:                                               ; preds = %82
  %89 = fmul double %67, 0x4350000000000000
  %90 = call i32 @llvm.nvvm.d2i.hi(double %89) #139
  %91 = lshr i32 %90, 20
  %92 = add nsw i32 %84, -54
  %93 = add nsw i32 %92, %91
  br label %94

94:                                               ; preds = %88, %82
  %.9 = phi double [ %89, %88 ], [ %67, %82 ]
  %expoa.0 = phi i32 [ %93, %88 ], [ %84, %82 ]
  %95 = icmp ult i32 %85, 1048576
  br i1 %95, label %96, label %102

96:                                               ; preds = %94
  %97 = fmul double %68, 0x4350000000000000
  %98 = call i32 @llvm.nvvm.d2i.hi(double %97) #139
  %99 = lshr i32 %98, 20
  %100 = add nsw i32 %86, -54
  %101 = add nsw i32 %100, %99
  br label %102

102:                                              ; preds = %96, %94
  %.0 = phi double [ %97, %96 ], [ %68, %94 ]
  %expob.0 = phi i32 [ %101, %96 ], [ %86, %94 ]
  %103 = bitcast double %.9 to i64
  %104 = bitcast double %.0 to i64
  %105 = and i64 %103, 4503599627370495
  %106 = or i64 %105, 4503599627370496
  %107 = and i64 %104, 4503599627370495
  %108 = or i64 %107, 4503599627370496
  %109 = sub nsw i32 %expoa.0, %expob.0
  br label %110

110:                                              ; preds = %110, %102
  %ia.0 = phi i64 [ %106, %102 ], [ %115, %110 ]
  %expob.1 = phi i32 [ %109, %102 ], [ %116, %110 ]
  %111 = sub i64 %ia.0, %108
  %112 = bitcast i64 %111 to double
  %113 = call i32 @llvm.nvvm.d2i.hi(double %112) #139
  %114 = icmp slt i32 %113, 0
  %spec.select = select i1 %114, i64 %ia.0, i64 %111
  %115 = shl i64 %spec.select, 1
  %116 = add nsw i32 %expob.1, -1
  %117 = icmp sgt i32 %expob.1, 0
  br i1 %117, label %110, label %118

118:                                              ; preds = %110
  %119 = and i64 %spec.select, 9223372036854775807
  %.not = icmp eq i64 %119, 0
  br i1 %.not, label %139, label %120

120:                                              ; preds = %118
  %121 = bitcast i64 %119 to double
  %122 = fmul double %121, 0x4350000000000000
  %123 = call i32 @llvm.nvvm.d2i.hi(double %122) #139
  %124 = lshr i32 %123, 20
  %125 = sub nsw i32 55, %124
  %126 = sub nsw i32 %expob.0, %125
  %127 = zext i32 %125 to i64
  %128 = shl i64 %119, %127
  %129 = icmp slt i32 %126, 1
  br i1 %129, label %130, label %134

130:                                              ; preds = %120
  %131 = sub nsw i32 1, %126
  %132 = zext i32 %131 to i64
  %133 = lshr i64 %128, %132
  br label %139

134:                                              ; preds = %120
  %135 = add nuw nsw i32 %126, 4095
  %136 = zext i32 %135 to i64
  %137 = shl i64 %136, 52
  %138 = add i64 %137, %128
  br label %139

139:                                              ; preds = %130, %134, %118
  %ia.3 = phi i64 [ 0, %118 ], [ %133, %130 ], [ %138, %134 ]
  %140 = bitcast i64 %ia.3 to double
  %141 = call i32 @llvm.nvvm.d2i.lo(double %140) #139
  %142 = call i32 @llvm.nvvm.d2i.hi(double %140) #139
  %143 = and i32 %61, -2147483648
  %144 = or i32 %142, %143
  %145 = call double @llvm.nvvm.lohi.i2d(i32 %141, i32 %144) #139
  br label %146

146:                                              ; preds = %78, %27, %76, %74, %139, %80, %25, %23, %55, %29
  %.02 = phi double [ %.01, %25 ], [ %24, %23 ], [ %59, %55 ], [ %a, %29 ], [ %.7, %76 ], [ %75, %74 ], [ %145, %139 ], [ %a, %80 ], [ 0xFFF8000000000000, %27 ], [ 0xFFF8000000000000, %78 ]
  ret double %.02
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_remainder(double %a, double %b) local_unnamed_addr #127 {
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2 = icmp eq i32 %1, 200
  br i1 %2, label %9, label %3

3:                                                ; preds = %0
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %5 = icmp eq i32 %4, 350
  br i1 %5, label %9, label %6

6:                                                ; preds = %3
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %8 = icmp eq i32 %7, 370
  br i1 %8, label %9, label %70

9:                                                ; preds = %6, %3, %0
  %10 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %11 = and i32 %10, 2147483647
  %12 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %13 = call i32 @llvm.nvvm.d2i.hi(double %b) #139
  %14 = and i32 %13, 2147483647
  %15 = call i32 @llvm.nvvm.d2i.lo(double %b) #139
  %16 = call double @llvm.nvvm.lohi.i2d(i32 %12, i32 %11) #139
  %17 = call double @llvm.nvvm.lohi.i2d(i32 %15, i32 %14) #139
  %18 = icmp ugt i32 %11, 2146435071
  %19 = icmp ugt i32 %14, 2146435071
  %or.cond = select i1 %18, i1 true, i1 %19
  br i1 %or.cond, label %20, label %27

20:                                               ; preds = %9
  %21 = fcmp ole double %16, 0x7FF0000000000000
  %22 = fcmp ole double %17, 0x7FF0000000000000
  %or.cond2 = select i1 %21, i1 %22, i1 false
  br i1 %or.cond2, label %25, label %23

23:                                               ; preds = %20
  %24 = fadd double %a, %b
  br label %165

25:                                               ; preds = %20
  %26 = fcmp oeq double %16, 0x7FF0000000000000
  %.0 = select i1 %26, double 0xFFF8000000000000, double %a
  br label %165

27:                                               ; preds = %9
  %28 = fcmp oeq double %17, 0.000000e+00
  br i1 %28, label %165, label %29

29:                                               ; preds = %27
  %30 = fcmp ult double %16, %17
  br i1 %30, label %56, label %31

31:                                               ; preds = %29
  %32 = call i32 @llvm.nvvm.d2i.hi(double %17) #139
  %33 = icmp slt i32 %32, 1048576
  br i1 %33, label %.preheader, label %41

.preheader:                                       ; preds = %31
  %34 = fcmp olt double %17, %16
  br i1 %34, label %.lr.ph16, label %._crit_edge17

.lr.ph16:                                         ; preds = %.preheader, %.lr.ph16
  %t.015 = phi double [ %35, %.lr.ph16 ], [ %17, %.preheader ]
  %35 = fadd double %t.015, %t.015
  %36 = fcmp olt double %35, %16
  %37 = call i32 @llvm.nvvm.d2i.hi(double %35) #139
  %38 = icmp slt i32 %37, 1048576
  %39 = select i1 %36, i1 %38, i1 false
  br i1 %39, label %.lr.ph16, label %._crit_edge17

._crit_edge17:                                    ; preds = %.lr.ph16, %.preheader
  %t.0.lcssa = phi double [ %17, %.preheader ], [ %35, %.lr.ph16 ]
  %.lcssa = phi i32 [ %32, %.preheader ], [ %37, %.lr.ph16 ]
  %40 = call i32 @llvm.nvvm.d2i.lo(double %t.0.lcssa) #139
  br label %41

41:                                               ; preds = %._crit_edge17, %31
  %bhi.0 = phi i32 [ %.lcssa, %._crit_edge17 ], [ %14, %31 ]
  %blo.0 = phi i32 [ %40, %._crit_edge17 ], [ %15, %31 ]
  %scaled_b.0 = phi double [ %t.0.lcssa, %._crit_edge17 ], [ 0.000000e+00, %31 ]
  %42 = call i32 @llvm.nvvm.d2i.hi(double %16) #139
  %43 = icmp sgt i32 %42, 1048575
  %44 = and i32 %bhi.0, 1048575
  %45 = and i32 %10, 2146435072
  %46 = or i32 %44, %45
  %47 = call double @llvm.nvvm.lohi.i2d(i32 %blo.0, i32 %46) #139
  %scaled_b.1 = select i1 %43, double %47, double %scaled_b.0
  %48 = fcmp ogt double %scaled_b.1, %16
  %49 = fmul double %scaled_b.1, 5.000000e-01
  %scaled_b.2 = select i1 %48, double %49, double %scaled_b.1
  %50 = fcmp ult double %scaled_b.2, %17
  br i1 %50, label %56, label %.lr.ph

.lr.ph:                                           ; preds = %41, %.lr.ph
  %scaled_b.312 = phi double [ %53, %.lr.ph ], [ %scaled_b.2, %41 ]
  %.211 = phi double [ %.3, %.lr.ph ], [ %16, %41 ]
  %51 = fcmp oge double %.211, %scaled_b.312
  %52 = select i1 %51, double %scaled_b.312, double 0.000000e+00
  %.3 = fsub double %.211, %52
  %53 = fmul double %scaled_b.312, 5.000000e-01
  %54 = fcmp ult double %53, %17
  br i1 %54, label %._crit_edge, label %.lr.ph

._crit_edge:                                      ; preds = %.lr.ph
  %55 = zext i1 %51 to i32
  br label %56

56:                                               ; preds = %41, %._crit_edge, %29
  %.4 = phi double [ %16, %29 ], [ %.3, %._crit_edge ], [ %16, %41 ]
  %quot.0 = phi i32 [ 0, %29 ], [ %55, %._crit_edge ], [ 0, %41 ]
  %57 = fadd double %.4, %.4
  %58 = fcmp ogt double %57, %17
  br i1 %58, label %62, label %59

59:                                               ; preds = %56
  %60 = fcmp oeq double %57, %17
  %61 = icmp ne i32 %quot.0, 0
  %or.cond4 = select i1 %60, i1 %61, i1 false
  br i1 %or.cond4, label %62, label %64

62:                                               ; preds = %59, %56
  %63 = fsub double %.4, %17
  br label %64

64:                                               ; preds = %62, %59
  %.5 = phi double [ %63, %62 ], [ %.4, %59 ]
  %65 = call i32 @llvm.nvvm.d2i.hi(double %.5) #139
  %66 = call i32 @llvm.nvvm.d2i.lo(double %.5) #139
  %67 = and i32 %10, -2147483648
  %68 = xor i32 %65, %67
  %69 = call double @llvm.nvvm.lohi.i2d(i32 %66, i32 %68) #139
  br label %165

70:                                               ; preds = %6
  %71 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %72 = and i32 %71, 2147483647
  %73 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %74 = call i32 @llvm.nvvm.d2i.hi(double %b) #139
  %75 = and i32 %74, 2147483647
  %76 = call i32 @llvm.nvvm.d2i.lo(double %b) #139
  %77 = call double @llvm.nvvm.lohi.i2d(i32 %73, i32 %72) #139
  %78 = call double @llvm.nvvm.lohi.i2d(i32 %76, i32 %75) #139
  %79 = icmp ugt i32 %72, 2146435071
  %80 = icmp ugt i32 %75, 2146435071
  %or.cond5 = select i1 %79, i1 true, i1 %80
  br i1 %or.cond5, label %81, label %88

81:                                               ; preds = %70
  %82 = fcmp ole double %77, 0x7FF0000000000000
  %83 = fcmp ole double %78, 0x7FF0000000000000
  %or.cond7 = select i1 %82, i1 %83, i1 false
  br i1 %or.cond7, label %86, label %84

84:                                               ; preds = %81
  %85 = fadd double %a, %b
  br label %165

86:                                               ; preds = %81
  %87 = fcmp oeq double %77, 0x7FF0000000000000
  %.8 = select i1 %87, double 0xFFF8000000000000, double %a
  br label %165

88:                                               ; preds = %70
  %89 = fcmp oeq double %78, 0.000000e+00
  br i1 %89, label %165, label %90

90:                                               ; preds = %88
  %91 = fcmp ult double %77, %78
  br i1 %91, label %151, label %92

92:                                               ; preds = %90
  %93 = call i32 @llvm.nvvm.d2i.hi(double %77) #139
  %94 = lshr i32 %93, 20
  %95 = call i32 @llvm.nvvm.d2i.hi(double %78) #139
  %96 = lshr i32 %95, 20
  %97 = icmp ult i32 %93, 1048576
  br i1 %97, label %98, label %104

98:                                               ; preds = %92
  %99 = fmul double %77, 0x4350000000000000
  %100 = call i32 @llvm.nvvm.d2i.hi(double %99) #139
  %101 = lshr i32 %100, 20
  %102 = add nsw i32 %94, -54
  %103 = add nsw i32 %102, %101
  br label %104

104:                                              ; preds = %98, %92
  %.10 = phi double [ %99, %98 ], [ %77, %92 ]
  %expoa.0 = phi i32 [ %103, %98 ], [ %94, %92 ]
  %105 = icmp ult i32 %95, 1048576
  br i1 %105, label %106, label %112

106:                                              ; preds = %104
  %107 = fmul double %78, 0x4350000000000000
  %108 = call i32 @llvm.nvvm.d2i.hi(double %107) #139
  %109 = lshr i32 %108, 20
  %110 = add nsw i32 %96, -54
  %111 = add nsw i32 %110, %109
  br label %112

112:                                              ; preds = %106, %104
  %t7.0 = phi double [ %107, %106 ], [ %78, %104 ]
  %expob.0 = phi i32 [ %111, %106 ], [ %96, %104 ]
  %113 = bitcast double %.10 to i64
  %114 = bitcast double %t7.0 to i64
  %115 = and i64 %113, 4503599627370495
  %116 = or i64 %115, 4503599627370496
  %117 = and i64 %114, 4503599627370495
  %118 = or i64 %117, 4503599627370496
  %119 = sub nsw i32 %expoa.0, %expob.0
  br label %120

120:                                              ; preds = %120, %112
  %ia.0 = phi i64 [ %116, %112 ], [ %125, %120 ]
  %expob.1 = phi i32 [ %119, %112 ], [ %126, %120 ]
  %121 = sub i64 %ia.0, %118
  %122 = bitcast i64 %121 to double
  %123 = call i32 @llvm.nvvm.d2i.hi(double %122) #139
  %124 = and i32 %123, -2147483648
  %.not = icmp eq i32 %124, 0
  %spec.select = select i1 %.not, i64 %121, i64 %ia.0
  %125 = shl i64 %spec.select, 1
  %126 = add nsw i32 %expob.1, -1
  %127 = icmp sgt i32 %expob.1, 0
  br i1 %127, label %120, label %128

128:                                              ; preds = %120
  %129 = and i64 %spec.select, 9223372036854775807
  %.not9 = icmp eq i64 %129, 0
  br i1 %.not9, label %149, label %130

130:                                              ; preds = %128
  %131 = bitcast i64 %129 to double
  %132 = fmul double %131, 0x4350000000000000
  %133 = call i32 @llvm.nvvm.d2i.hi(double %132) #139
  %134 = lshr i32 %133, 20
  %135 = sub nsw i32 55, %134
  %136 = sub nsw i32 %expob.0, %135
  %137 = zext i32 %135 to i64
  %138 = shl i64 %129, %137
  %139 = icmp slt i32 %136, 1
  br i1 %139, label %140, label %144

140:                                              ; preds = %130
  %141 = sub nsw i32 1, %136
  %142 = zext i32 %141 to i64
  %143 = lshr i64 %138, %142
  br label %149

144:                                              ; preds = %130
  %145 = add nuw nsw i32 %136, 4095
  %146 = zext i32 %145 to i64
  %147 = shl i64 %146, 52
  %148 = add i64 %147, %138
  br label %149

149:                                              ; preds = %140, %144, %128
  %ia.3 = phi i64 [ 0, %128 ], [ %143, %140 ], [ %148, %144 ]
  %150 = bitcast i64 %ia.3 to double
  br label %151

151:                                              ; preds = %149, %90
  %.11 = phi double [ %150, %149 ], [ %77, %90 ]
  %nquot0.0 = phi i32 [ %124, %149 ], [ 1, %90 ]
  %152 = fadd double %.11, %.11
  %153 = fcmp ogt double %152, %78
  br i1 %153, label %157, label %154

154:                                              ; preds = %151
  %155 = fcmp une double %152, %78
  %156 = icmp ne i32 %nquot0.0, 0
  %or.cond8 = select i1 %155, i1 true, i1 %156
  br i1 %or.cond8, label %159, label %157

157:                                              ; preds = %154, %151
  %158 = fsub double %.11, %78
  br label %159

159:                                              ; preds = %157, %154
  %.12 = phi double [ %158, %157 ], [ %.11, %154 ]
  %160 = and i32 %71, -2147483648
  %161 = call i32 @llvm.nvvm.d2i.hi(double %.12) #139
  %162 = xor i32 %161, %160
  %163 = call i32 @llvm.nvvm.d2i.lo(double %.12) #139
  %164 = call double @llvm.nvvm.lohi.i2d(i32 %163, i32 %162) #139
  br label %165

165:                                              ; preds = %88, %27, %86, %84, %159, %25, %23, %64
  %.01 = phi double [ %.0, %25 ], [ %24, %23 ], [ %69, %64 ], [ %.8, %86 ], [ %85, %84 ], [ %164, %159 ], [ 0xFFF8000000000000, %27 ], [ 0xFFF8000000000000, %88 ]
  ret double %.01
}

; Function Attrs: noinline nounwind
define dso_local double @__nv_remquo(double %a, double %b, i32* nocapture nofree writeonly %c) local_unnamed_addr #128 {
  %1 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %2 = icmp eq i32 %1, 200
  br i1 %2, label %9, label %3

3:                                                ; preds = %0
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %5 = icmp eq i32 %4, 350
  br i1 %5, label %9, label %6

6:                                                ; preds = %3
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %8 = icmp eq i32 %7, 370
  br i1 %8, label %9, label %77

9:                                                ; preds = %6, %3, %0
  %10 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %11 = call i32 @llvm.nvvm.d2i.hi(double %b) #139
  %12 = xor i32 %11, %10
  %.lobit.neg11 = ashr i32 %12, 31
  %13 = and i32 %10, 2147483647
  %14 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %15 = and i32 %11, 2147483647
  %16 = call i32 @llvm.nvvm.d2i.lo(double %b) #139
  %17 = call double @llvm.nvvm.lohi.i2d(i32 %14, i32 %13) #139
  %18 = call double @llvm.nvvm.lohi.i2d(i32 %16, i32 %15) #139
  %19 = icmp ugt i32 %13, 2146435071
  %20 = icmp ugt i32 %15, 2146435071
  %or.cond = select i1 %19, i1 true, i1 %20
  br i1 %or.cond, label %21, label %28

21:                                               ; preds = %9
  %22 = fcmp ole double %17, 0x7FF0000000000000
  %23 = fcmp ole double %18, 0x7FF0000000000000
  %or.cond2 = select i1 %22, i1 %23, i1 false
  br i1 %or.cond2, label %26, label %24

24:                                               ; preds = %21
  %25 = fadd double %a, %b
  br label %178

26:                                               ; preds = %21
  %27 = fcmp oeq double %17, 0x7FF0000000000000
  %.0 = select i1 %27, double 0xFFF8000000000000, double %a
  br label %178

28:                                               ; preds = %9
  %29 = fcmp oeq double %18, 0.000000e+00
  br i1 %29, label %178, label %30

30:                                               ; preds = %28
  %31 = fcmp ult double %17, %18
  br i1 %31, label %58, label %32

32:                                               ; preds = %30
  %33 = call i32 @llvm.nvvm.d2i.hi(double %18) #139
  %34 = icmp slt i32 %33, 1048576
  br i1 %34, label %.preheader, label %42

.preheader:                                       ; preds = %32
  %35 = fcmp olt double %18, %17
  br i1 %35, label %.lr.ph18, label %._crit_edge19

.lr.ph18:                                         ; preds = %.preheader, %.lr.ph18
  %t.017 = phi double [ %36, %.lr.ph18 ], [ %18, %.preheader ]
  %36 = fadd double %t.017, %t.017
  %37 = fcmp olt double %36, %17
  %38 = call i32 @llvm.nvvm.d2i.hi(double %36) #139
  %39 = icmp slt i32 %38, 1048576
  %40 = select i1 %37, i1 %39, i1 false
  br i1 %40, label %.lr.ph18, label %._crit_edge19

._crit_edge19:                                    ; preds = %.lr.ph18, %.preheader
  %t.0.lcssa = phi double [ %18, %.preheader ], [ %36, %.lr.ph18 ]
  %.lcssa = phi i32 [ %33, %.preheader ], [ %38, %.lr.ph18 ]
  %41 = call i32 @llvm.nvvm.d2i.lo(double %t.0.lcssa) #139
  br label %42

42:                                               ; preds = %._crit_edge19, %32
  %bhi.0 = phi i32 [ %.lcssa, %._crit_edge19 ], [ %15, %32 ]
  %blo.0 = phi i32 [ %41, %._crit_edge19 ], [ %16, %32 ]
  %scaled_b.0 = phi double [ %t.0.lcssa, %._crit_edge19 ], [ 0.000000e+00, %32 ]
  %43 = call i32 @llvm.nvvm.d2i.hi(double %17) #139
  %44 = icmp sgt i32 %43, 1048575
  %45 = and i32 %bhi.0, 1048575
  %46 = and i32 %10, 2146435072
  %47 = or i32 %45, %46
  %48 = call double @llvm.nvvm.lohi.i2d(i32 %blo.0, i32 %47) #139
  %scaled_b.1 = select i1 %44, double %48, double %scaled_b.0
  %49 = fcmp ogt double %scaled_b.1, %17
  %50 = fmul double %scaled_b.1, 5.000000e-01
  %scaled_b.2 = select i1 %49, double %50, double %scaled_b.1
  %51 = fcmp ult double %scaled_b.2, %18
  br i1 %51, label %58, label %.lr.ph

.lr.ph:                                           ; preds = %42, %.lr.ph
  %nquot.015 = phi i32 [ %nquot.1, %.lr.ph ], [ -1, %42 ]
  %scaled_b.314 = phi double [ %56, %.lr.ph ], [ %scaled_b.2, %42 ]
  %.213 = phi double [ %.3, %.lr.ph ], [ %17, %42 ]
  %52 = shl i32 %nquot.015, 1
  %53 = fcmp oge double %.213, %scaled_b.314
  %54 = select i1 %53, double %scaled_b.314, double 0.000000e+00
  %.3 = fsub double %.213, %54
  %not. = xor i1 %53, true
  %55 = zext i1 %not. to i32
  %nquot.1 = or i32 %52, %55
  %56 = fmul double %scaled_b.314, 5.000000e-01
  %57 = fcmp ult double %56, %18
  br i1 %57, label %._crit_edge, label %.lr.ph

._crit_edge:                                      ; preds = %.lr.ph
  %phi.bo22 = xor i32 %nquot.1, -1
  br label %58

58:                                               ; preds = %42, %._crit_edge, %30
  %.4 = phi double [ %17, %30 ], [ %.3, %._crit_edge ], [ %17, %42 ]
  %quot.0 = phi i32 [ 0, %30 ], [ %phi.bo22, %._crit_edge ], [ 0, %42 ]
  %59 = fadd double %.4, %.4
  %60 = fcmp ogt double %59, %18
  br i1 %60, label %65, label %61

61:                                               ; preds = %58
  %62 = fcmp oeq double %59, %18
  %63 = and i32 %quot.0, 1
  %64 = icmp ne i32 %63, 0
  %or.cond4 = select i1 %62, i1 %64, i1 false
  br i1 %or.cond4, label %65, label %68

65:                                               ; preds = %61, %58
  %66 = add i32 %quot.0, 1
  %67 = fsub double %.4, %18
  br label %68

68:                                               ; preds = %65, %61
  %.5 = phi double [ %67, %65 ], [ %.4, %61 ]
  %quot.1 = phi i32 [ %66, %65 ], [ %quot.0, %61 ]
  %69 = call i32 @llvm.nvvm.d2i.hi(double %.5) #139
  %70 = call i32 @llvm.nvvm.d2i.lo(double %.5) #139
  %71 = and i32 %10, -2147483648
  %72 = xor i32 %69, %71
  %73 = call double @llvm.nvvm.lohi.i2d(i32 %70, i32 %72) #139
  %74 = and i32 %quot.1, 7
  %75 = xor i32 %74, %.lobit.neg11
  %76 = sub nsw i32 %75, %.lobit.neg11
  br label %178

77:                                               ; preds = %6
  %78 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %79 = call i32 @llvm.nvvm.d2i.hi(double %b) #139
  %80 = xor i32 %79, %78
  %.lobit.neg = ashr i32 %80, 31
  %81 = and i32 %78, 2147483647
  %82 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %83 = and i32 %79, 2147483647
  %84 = call i32 @llvm.nvvm.d2i.lo(double %b) #139
  %85 = call double @llvm.nvvm.lohi.i2d(i32 %82, i32 %81) #139
  %86 = call double @llvm.nvvm.lohi.i2d(i32 %84, i32 %83) #139
  %87 = icmp ugt i32 %81, 2146435071
  %88 = icmp ugt i32 %83, 2146435071
  %or.cond5 = select i1 %87, i1 true, i1 %88
  br i1 %or.cond5, label %89, label %96

89:                                               ; preds = %77
  %90 = fcmp ole double %85, 0x7FF0000000000000
  %91 = fcmp ole double %86, 0x7FF0000000000000
  %or.cond7 = select i1 %90, i1 %91, i1 false
  br i1 %or.cond7, label %94, label %92

92:                                               ; preds = %89
  %93 = fadd double %a, %b
  br label %178

94:                                               ; preds = %89
  %95 = fcmp oeq double %85, 0x7FF0000000000000
  %.8 = select i1 %95, double 0xFFF8000000000000, double %a
  br label %178

96:                                               ; preds = %77
  %97 = fcmp oeq double %86, 0.000000e+00
  br i1 %97, label %178, label %98

98:                                               ; preds = %96
  %99 = fcmp ult double %85, %86
  br i1 %99, label %159, label %100

100:                                              ; preds = %98
  %101 = call i32 @llvm.nvvm.d2i.hi(double %85) #139
  %102 = lshr i32 %101, 20
  %103 = call i32 @llvm.nvvm.d2i.hi(double %86) #139
  %104 = lshr i32 %103, 20
  %105 = icmp ult i32 %101, 1048576
  br i1 %105, label %106, label %112

106:                                              ; preds = %100
  %107 = fmul double %85, 0x4350000000000000
  %108 = call i32 @llvm.nvvm.d2i.hi(double %107) #139
  %109 = lshr i32 %108, 20
  %110 = add nsw i32 %102, -54
  %111 = add nsw i32 %110, %109
  br label %112

112:                                              ; preds = %106, %100
  %.10 = phi double [ %107, %106 ], [ %85, %100 ]
  %expoa.0 = phi i32 [ %111, %106 ], [ %102, %100 ]
  %113 = icmp ult i32 %103, 1048576
  br i1 %113, label %114, label %120

114:                                              ; preds = %112
  %115 = fmul double %86, 0x4350000000000000
  %116 = call i32 @llvm.nvvm.d2i.hi(double %115) #139
  %117 = lshr i32 %116, 20
  %118 = add nsw i32 %104, -54
  %119 = add nsw i32 %118, %117
  br label %120

120:                                              ; preds = %114, %112
  %t10.0 = phi double [ %115, %114 ], [ %86, %112 ]
  %expob.0 = phi i32 [ %119, %114 ], [ %104, %112 ]
  %121 = bitcast double %.10 to i64
  %122 = bitcast double %t10.0 to i64
  %123 = and i64 %121, 4503599627370495
  %124 = or i64 %123, 4503599627370496
  %125 = and i64 %122, 4503599627370495
  %126 = or i64 %125, 4503599627370496
  %127 = sub nsw i32 %expoa.0, %expob.0
  br label %128

128:                                              ; preds = %128, %120
  %nquot4.0 = phi i32 [ -1, %120 ], [ %133, %128 ]
  %ia.0 = phi i64 [ %124, %120 ], [ %132, %128 ]
  %expob.1 = phi i32 [ %127, %120 ], [ %134, %128 ]
  %129 = sub i64 %ia.0, %126
  %130 = bitcast i64 %129 to double
  %131 = call i32 @llvm.nvvm.d2i.hi(double %130) #139
  %.not10 = icmp slt i32 %131, 0
  %spec.select = select i1 %.not10, i64 %ia.0, i64 %129
  %132 = shl i64 %spec.select, 1
  %133 = call i32 @llvm.fshl.i32(i32 %nquot4.0, i32 %131, i32 noundef 1)
  %134 = add nsw i32 %expob.1, -1
  %135 = icmp sgt i32 %expob.1, 0
  br i1 %135, label %128, label %136

136:                                              ; preds = %128
  %137 = and i64 %spec.select, 9223372036854775807
  %.not = icmp eq i64 %137, 0
  br i1 %.not, label %157, label %138

138:                                              ; preds = %136
  %139 = bitcast i64 %137 to double
  %140 = fmul double %139, 0x4350000000000000
  %141 = call i32 @llvm.nvvm.d2i.hi(double %140) #139
  %142 = lshr i32 %141, 20
  %143 = sub nsw i32 55, %142
  %144 = sub nsw i32 %expob.0, %143
  %145 = zext i32 %143 to i64
  %146 = shl i64 %137, %145
  %147 = icmp slt i32 %144, 1
  br i1 %147, label %148, label %152

148:                                              ; preds = %138
  %149 = sub nsw i32 1, %144
  %150 = zext i32 %149 to i64
  %151 = lshr i64 %146, %150
  br label %157

152:                                              ; preds = %138
  %153 = add nuw nsw i32 %144, 4095
  %154 = zext i32 %153 to i64
  %155 = shl i64 %154, 52
  %156 = add i64 %155, %146
  br label %157

157:                                              ; preds = %148, %152, %136
  %ia.3 = phi i64 [ 0, %136 ], [ %151, %148 ], [ %156, %152 ]
  %158 = bitcast i64 %ia.3 to double
  %phi.bo = xor i32 %133, -1
  br label %159

159:                                              ; preds = %157, %98
  %.11 = phi double [ %158, %157 ], [ %85, %98 ]
  %nquot4.1 = phi i32 [ %phi.bo, %157 ], [ 0, %98 ]
  %160 = fadd double %.11, %.11
  %161 = fcmp ogt double %160, %86
  br i1 %161, label %166, label %162

162:                                              ; preds = %159
  %163 = fcmp oeq double %160, %86
  %164 = and i32 %nquot4.1, 1
  %165 = icmp ne i32 %164, 0
  %or.cond9 = select i1 %163, i1 %165, i1 false
  br i1 %or.cond9, label %166, label %169

166:                                              ; preds = %162, %159
  %167 = add i32 %nquot4.1, 1
  %168 = fsub double %.11, %86
  br label %169

169:                                              ; preds = %166, %162
  %.12 = phi double [ %168, %166 ], [ %.11, %162 ]
  %quot3.0 = phi i32 [ %167, %166 ], [ %nquot4.1, %162 ]
  %170 = and i32 %78, -2147483648
  %171 = call i32 @llvm.nvvm.d2i.hi(double %.12) #139
  %172 = xor i32 %171, %170
  %173 = call i32 @llvm.nvvm.d2i.lo(double %.12) #139
  %174 = call double @llvm.nvvm.lohi.i2d(i32 %173, i32 %172) #139
  %175 = and i32 %quot3.0, 7
  %176 = xor i32 %175, %.lobit.neg
  %177 = sub nsw i32 %176, %.lobit.neg
  br label %178

178:                                              ; preds = %94, %92, %169, %96, %26, %24, %68, %28
  %storemerge = phi i32 [ 0, %26 ], [ 0, %24 ], [ %76, %68 ], [ 0, %28 ], [ 0, %94 ], [ 0, %92 ], [ %177, %169 ], [ 0, %96 ]
  %.01 = phi double [ %.0, %26 ], [ %25, %24 ], [ %73, %68 ], [ 0xFFF8000000000000, %28 ], [ %.8, %94 ], [ %93, %92 ], [ %174, %169 ], [ 0xFFF8000000000000, %96 ]
  store i32 %storemerge, i32* %c, align 4, !noalias !68
  ret double %.01
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_round(double %a) local_unnamed_addr #129 {
  %1 = call i32 @llvm.nvvm.d2i.hi(double %a) #142
  %2 = call i32 @llvm.nvvm.d2i.lo(double noundef 5.000000e-01) #142
  %3 = call i32 @llvm.nvvm.d2i.hi(double noundef 5.000000e-01) #142
  %4 = and i32 %1, -2147483648
  %5 = or i32 %4, %3
  %6 = call double @llvm.nvvm.lohi.i2d(i32 %2, i32 %5) #142
  %7 = call double @llvm.nvvm.add.rz.d(double %a, double %6) #142
  %8 = call double @llvm.trunc.f64(double %7) #140
  ret double %8
}

; Function Attrs: noinline nounwind
define dso_local i32 @__nv_ilogb(double %a) local_unnamed_addr #130 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ugt double %1, 0x7FF0000000000000
  br i1 %2, label %36, label %3

3:                                                ; preds = %0
  %4 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %5 = icmp eq i32 %4, 200
  br i1 %5, label %12, label %6

6:                                                ; preds = %3
  %7 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %8 = icmp eq i32 %7, 350
  br i1 %8, label %12, label %9

9:                                                ; preds = %6
  %10 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([12 x i8], [12 x i8]* @.str.1, i64 0, i64 0)) #139
  %11 = icmp eq i32 %10, 370
  br i1 %11, label %12, label %14

12:                                               ; preds = %9, %6, %3
  %13 = fcmp oeq double %1, 0x7FF0000000000000
  br label %__nv_isinfd.exit

14:                                               ; preds = %9
  %15 = call i32 @llvm.nvvm.d2i.lo(double %a) #139
  %16 = call i32 @llvm.nvvm.d2i.hi(double %a) #139
  %17 = and i32 %16, 2147483647
  %18 = icmp eq i32 %17, 2146435072
  %19 = icmp eq i32 %15, 0
  %20 = select i1 %18, i1 %19, i1 false
  br label %__nv_isinfd.exit

__nv_isinfd.exit:                                 ; preds = %14, %12
  %.01.in = phi i1 [ %13, %12 ], [ %20, %14 ]
  br i1 %.01.in, label %36, label %21

21:                                               ; preds = %__nv_isinfd.exit
  %22 = fcmp oeq double %a, 0.000000e+00
  br i1 %22, label %36, label %23

23:                                               ; preds = %21
  %24 = call i32 @llvm.nvvm.d2i.hi(double %1) #139
  %25 = icmp ugt i32 %24, 1048575
  br i1 %25, label %26, label %29

26:                                               ; preds = %23
  %27 = lshr i32 %24, 20
  %28 = add nsw i32 %27, -1023
  br label %36

29:                                               ; preds = %23
  %30 = zext i32 %24 to i64
  %31 = shl nuw nsw i64 %30, 32
  %32 = call i32 @llvm.nvvm.d2i.lo(double %1) #139
  %33 = zext i32 %32 to i64
  %34 = or i64 %31, %33
  %ctlz = call i64 @llvm.ctlz.i64(i64 %34, i1 noundef false), !range !80
  %ctlz.trunc = trunc i64 %ctlz to i32
  %35 = sub nuw nsw i32 -1011, %ctlz.trunc
  br label %36

36:                                               ; preds = %21, %__nv_isinfd.exit, %0, %29, %26
  %.0 = phi i32 [ %28, %26 ], [ %35, %29 ], [ -2147483648, %0 ], [ 2147483647, %__nv_isinfd.exit ], [ -2147483648, %21 ]
  ret i32 %.0
}

; Function Attrs: nofree noinline nosync nounwind readnone willreturn
define dso_local double @__nv_logb(double %a) local_unnamed_addr #131 {
  %1 = call double @llvm.fabs.f64(double %a) #140
  %2 = fcmp ugt double %1, 0x7FF0000000000000
  br i1 %2, label %3, label %5

3:                                                ; preds = %0
  %4 = fadd double %a, %a
  br label %24

5:                                                ; preds = %0
  %6 = fcmp oeq double %1, 0x7FF0000000000000
  br i1 %6, label %24, label %7

7:                                                ; preds = %5
  %8 = fcmp oeq double %a, 0.000000e+00
  br i1 %8, label %24, label %9

9:                                                ; preds = %7
  %10 = call i32 @llvm.nvvm.d2i.hi(double %1) #142
  %11 = icmp ugt i32 %10, 1048575
  br i1 %11, label %12, label %16

12:                                               ; preds = %9
  %13 = lshr i32 %10, 20
  %14 = add nsw i32 %13, -1023
  %15 = sitofp i32 %14 to double
  br label %24

16:                                               ; preds = %9
  %17 = zext i32 %10 to i64
  %18 = shl nuw nsw i64 %17, 32
  %19 = call i32 @llvm.nvvm.d2i.lo(double %1) #142
  %20 = zext i32 %19 to i64
  %21 = or i64 %18, %20
  %ctlz = call i64 @llvm.ctlz.i64(i64 %21, i1 noundef false) #140, !range !80
  %ctlz.trunc = trunc i64 %ctlz to i32
  %22 = sub nuw nsw i32 -1011, %ctlz.trunc
  %23 = sitofp i32 %22 to double
  br label %24

24:                                               ; preds = %7, %5, %16, %12, %3
  %.0 = phi double [ %4, %3 ], [ %15, %12 ], [ %23, %16 ], [ 0x7FF0000000000000, %5 ], [ 0xFFF0000000000000, %7 ]
  ret double %.0
}

; Function Attrs: noinline nosync nounwind readnone
define dso_local double @__nv_fdim(double %a, double %b) local_unnamed_addr #132 {
  %1 = call double asm "sub.rn.f64 $0, $1, $2;", "=d,d,d"(double %a, double %b) #141, !srcloc !90
  %2 = fcmp ole double %a, %b
  %t.0 = select i1 %2, double 0.000000e+00, double %1
  ret double %t.0
}

; Function Attrs: nofree
define internal fastcc void @gpu_report_exception(i64 noundef zeroext %0) unnamed_addr #133 !dbg !91 {
top:
  %1 = alloca %printf_args.0, align 8
  %2 = call {}*** @julia.get_pgcstack()
  %3 = bitcast %printf_args.0* %1 to i8*, !dbg !92
  call void @llvm.lifetime.start.p0i8(i64 noundef 8, i8* noundef nonnull %3), !dbg !92
  %4 = getelementptr inbounds %printf_args.0, %printf_args.0* %1, i64 0, i32 0, !dbg !92
  store i64 %0, i64* %4, align 8, !dbg !92, !noalias !68
  %5 = call i32 @vprintf(i8* noundef getelementptr inbounds ([108 x i8], [108 x i8]* @1, i64 0, i64 0), i8* noundef nonnull %3), !dbg !92
  call void @llvm.lifetime.end.p0i8(i64 noundef 8, i8* noundef nonnull %3), !dbg !92
  ret void, !dbg !99
}

; Function Attrs: nofree
define internal fastcc void @gpu_signal_exception() unnamed_addr #133 !dbg !100 {
top:
  %0 = call {}*** @julia.get_pgcstack()
  %state.i = call [1 x i64] @julia.gpu.state_getter(), !dbg !101
  %state.i.fca.0.extract = extractvalue [1 x i64] %state.i, 0, !dbg !101
  %.not = icmp eq i64 %state.i.fca.0.extract, 0, !dbg !110
  br i1 %.not, label %L12, label %L8, !dbg !110

L8:                                               ; preds = %top
  %1 = inttoptr i64 %state.i.fca.0.extract to i64*, !dbg !111
  store i64 1, i64* %1, align 1, !dbg !111, !tbaa !116, !noalias !68
  call void @llvm.nvvm.membar.sys(), !dbg !120
  br label %L15, !dbg !123

L12:                                              ; preds = %top
  %2 = call i32 @vprintf(i8* noundef getelementptr inbounds ([110 x i8], [110 x i8]* @2, i64 0, i64 0), i8* noundef null), !dbg !124
  br label %L15, !dbg !124

L15:                                              ; preds = %L12, %L8
  ret void, !dbg !130
}

; Function Attrs: readnone
declare [1 x i64] @julia.gpu.state_getter() local_unnamed_addr #134

; Function Attrs: nounwind
declare void @llvm.nvvm.membar.sys() #135

; Function Attrs: argmemonly nofree nosync nounwind willreturn
declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #136

; Function Attrs: argmemonly nofree nosync nounwind willreturn
declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture) #136

define void @julia_gbsa_force_2_kernel__25091_inner24({ i8 addrspace(1)*, i64, [2 x i64], i64 } %0, { i8 addrspace(1)*, i64, [1 x i64], i64 } %1, { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, [1 x [1 x [3 x float]]] %3, float %4, { i8 addrspace(1)*, i64, [1 x i64], i64 } %5, { i8 addrspace(1)*, i64, [1 x i64], i64 } %6, { i8 addrspace(1)*, i64, [1 x i64], i64 } %7, { i8 addrspace(1)*, i64, [1 x i64], i64 } %8, { i8 addrspace(1)*, i64, [2 x i64], i64 } %9) local_unnamed_addr #137 !dbg !131 {
entry:
  %10 = alloca [1 x [2 x [1 x i64]]], align 8
  %11 = alloca [2 x i64], align 8
  %12 = alloca [1 x [2 x [1 x i64]]], align 8
  %13 = alloca [2 x i64], align 8
  %.fca.0.extract68 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 0, !dbg !132
  %.fca.2.0.extract70 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 2, 0, !dbg !132
  %.fca.2.1.extract71 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 2, 1, !dbg !132
  %.fca.0.extract60 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %1, 0, !dbg !132
  %.fca.0.extract47 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, 0, !dbg !132
  %.fca.2.0.extract49 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, 2, 0, !dbg !132
  %.fca.0.0.0.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 0, !dbg !132
  %.fca.0.0.1.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 1, !dbg !132
  %.fca.0.0.2.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 2, !dbg !132
  %.fca.0.extract29 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %5, 0, !dbg !132
  %.fca.0.extract21 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %6, 0, !dbg !132
  %.fca.0.extract13 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %7, 0, !dbg !132
  %.fca.0.extract6 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %8, 0, !dbg !132
  %.fca.0.extract = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %9, 0, !dbg !132
  %.fca.2.0.extract = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %9, 2, 0, !dbg !132
  %14 = bitcast [1 x [2 x [1 x i64]]]* %10 to i8*
  call void @llvm.lifetime.start.p0i8(i64 noundef 16, i8* noundef nonnull align 8 dereferenceable(16) %14) #144
  %15 = bitcast [2 x i64]* %11 to i8*
  call void @llvm.lifetime.start.p0i8(i64 noundef 16, i8* noundef nonnull align 8 dereferenceable(16) %15) #144
  %16 = bitcast [1 x [2 x [1 x i64]]]* %12 to i8*
  call void @llvm.lifetime.start.p0i8(i64 noundef 16, i8* noundef nonnull align 8 dereferenceable(16) %16) #144
  %17 = bitcast [2 x i64]* %13 to i8*
  call void @llvm.lifetime.start.p0i8(i64 noundef 16, i8* noundef nonnull align 8 dereferenceable(16) %17) #144
  %18 = call {}*** @julia.get_pgcstack()
  %19 = mul i64 %.fca.2.0.extract49, %.fca.2.0.extract49, !dbg !133
  %20 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x(), !dbg !141, !range !153
  %21 = zext i32 %20 to i64, !dbg !154
  %22 = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x(), !dbg !157, !range !164
  %23 = zext i32 %22 to i64, !dbg !165
  %24 = mul nuw nsw i64 %21, %23, !dbg !168
  %25 = call i32 @llvm.nvvm.read.ptx.sreg.tid.x(), !dbg !170, !range !177
  %26 = add nuw nsw i32 %25, 1, !dbg !178
  %27 = zext i32 %26 to i64, !dbg !180
  %28 = add nuw nsw i64 %24, %27, !dbg !182
  %.not = icmp sgt i64 %28, %19, !dbg !184
  br i1 %.not, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L33.i, !dbg !187

L33.i:                                            ; preds = %entry
  %.not153 = icmp eq i64 %.fca.2.0.extract49, 0, !dbg !188
  br i1 %.not153, label %fail.i, label %pass.i, !dbg !188

L50.i:                                            ; preds = %after_srem.i
  %29 = add nsw i64 %219, -1, !dbg !197
  %30 = bitcast i8 addrspace(1)* %.fca.0.extract47 to [1 x [3 x float]] addrspace(1)*, !dbg !215
  %.unpack.elt = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %30, i64 %29, i64 0, i64 0, !dbg !215
  %.unpack.unpack = load float, float addrspace(1)* %.unpack.elt, align 4, !dbg !215, !tbaa !216
  %.unpack.elt154 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %30, i64 %29, i64 0, i64 1, !dbg !215
  %.unpack.unpack155 = load float, float addrspace(1)* %.unpack.elt154, align 4, !dbg !215, !tbaa !216
  %.unpack.elt156 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %30, i64 %29, i64 0, i64 2, !dbg !215
  %.unpack.unpack157 = load float, float addrspace(1)* %.unpack.elt156, align 4, !dbg !215, !tbaa !216
  %.unpack.elt159 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %30, i64 %222, i64 0, i64 0, !dbg !215
  %.unpack.unpack160 = load float, float addrspace(1)* %.unpack.elt159, align 4, !dbg !215, !tbaa !216
  %.unpack.elt161 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %30, i64 %222, i64 0, i64 1, !dbg !215
  %.unpack.unpack162 = load float, float addrspace(1)* %.unpack.elt161, align 4, !dbg !215, !tbaa !216
  %.unpack.elt163 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %30, i64 %222, i64 0, i64 2, !dbg !215
  %.unpack.unpack164 = load float, float addrspace(1)* %.unpack.elt163, align 4, !dbg !215, !tbaa !216
  %31 = fcmp uge float %.unpack.unpack, %.unpack.unpack160, !dbg !219
  br i1 %31, label %L109.i, label %L99.i, !dbg !229

L99.i:                                            ; preds = %L50.i
  %32 = fsub float %.unpack.unpack160, %.unpack.unpack, !dbg !230
  %33 = fsub float %.unpack.unpack, %.unpack.unpack160, !dbg !230
  %34 = fadd float %.fca.0.0.0.extract, %33, !dbg !233
  %35 = fcmp uge float %32, %34, !dbg !235
  br i1 %35, label %L106.i, label %L119.i, !dbg !236

L106.i:                                           ; preds = %L99.i
  %36 = fsub float %32, %.fca.0.0.0.extract, !dbg !230
  br label %L119.i, !dbg !236

L109.i:                                           ; preds = %L50.i
  %37 = fsub float %.unpack.unpack, %.unpack.unpack160, !dbg !237
  %38 = fsub float %.unpack.unpack160, %.unpack.unpack, !dbg !237
  %39 = fadd float %.fca.0.0.0.extract, %38, !dbg !239
  %40 = fcmp uge float %37, %39, !dbg !240
  %. = select i1 %40, float %39, float %38, !dbg !241
  br label %L119.i, !dbg !241

L119.i:                                           ; preds = %L109.i, %L99.i, %L106.i
  %value_phi.i = phi float [ %36, %L106.i ], [ %32, %L99.i ], [ %., %L109.i ]
  %41 = fcmp uge float %.unpack.unpack155, %.unpack.unpack162, !dbg !219
  br i1 %41, label %L139.i, label %L129.i, !dbg !229

L129.i:                                           ; preds = %L119.i
  %42 = fsub float %.unpack.unpack162, %.unpack.unpack155, !dbg !230
  %43 = fsub float %.unpack.unpack155, %.unpack.unpack162, !dbg !230
  %44 = fadd float %.fca.0.0.1.extract, %43, !dbg !233
  %45 = fcmp uge float %42, %44, !dbg !235
  br i1 %45, label %L136.i, label %L149.i, !dbg !236

L136.i:                                           ; preds = %L129.i
  %46 = fsub float %42, %.fca.0.0.1.extract, !dbg !230
  br label %L149.i, !dbg !236

L139.i:                                           ; preds = %L119.i
  %47 = fsub float %.unpack.unpack155, %.unpack.unpack162, !dbg !237
  %48 = fsub float %.unpack.unpack162, %.unpack.unpack155, !dbg !237
  %49 = fadd float %.fca.0.0.1.extract, %48, !dbg !239
  %50 = fcmp uge float %47, %49, !dbg !240
  %.185 = select i1 %50, float %49, float %48, !dbg !241
  br label %L149.i, !dbg !241

L149.i:                                           ; preds = %L139.i, %L129.i, %L136.i
  %value_phi3.i = phi float [ %46, %L136.i ], [ %42, %L129.i ], [ %.185, %L139.i ]
  %51 = fcmp uge float %.unpack.unpack157, %.unpack.unpack164, !dbg !219
  br i1 %51, label %L169.i, label %L159.i, !dbg !229

L159.i:                                           ; preds = %L149.i
  %52 = fsub float %.unpack.unpack164, %.unpack.unpack157, !dbg !230
  %53 = fsub float %.unpack.unpack157, %.unpack.unpack164, !dbg !230
  %54 = fadd float %.fca.0.0.2.extract, %53, !dbg !233
  %55 = fcmp uge float %52, %54, !dbg !235
  br i1 %55, label %L166.i, label %L179.i, !dbg !236

L166.i:                                           ; preds = %L159.i
  %56 = fsub float %52, %.fca.0.0.2.extract, !dbg !230
  br label %L179.i, !dbg !236

L169.i:                                           ; preds = %L149.i
  %57 = fsub float %.unpack.unpack157, %.unpack.unpack164, !dbg !237
  %58 = fsub float %.unpack.unpack164, %.unpack.unpack157, !dbg !237
  %59 = fadd float %.fca.0.0.2.extract, %58, !dbg !239
  %60 = fcmp uge float %57, %59, !dbg !240
  %.186 = select i1 %60, float %59, float %58, !dbg !241
  br label %L179.i, !dbg !241

L179.i:                                           ; preds = %L169.i, %L159.i, %L166.i
  %value_phi4.i = phi float [ %56, %L166.i ], [ %52, %L159.i ], [ %.186, %L169.i ]
  %61 = fmul float %value_phi.i, %value_phi.i, !dbg !242
  %62 = fmul float %value_phi3.i, %value_phi3.i, !dbg !242
  %63 = fadd float %61, %62, !dbg !258
  %64 = fmul float %value_phi4.i, %value_phi4.i, !dbg !242
  %65 = fadd float %63, %64, !dbg !258
  %66 = call float @__nv_sqrtf(float %65) #139, !dbg !259
  %67 = fcmp ule float %66, 0.000000e+00, !dbg !262
  br i1 %67, label %L197.i, label %L190.i, !dbg !264

L190.i:                                           ; preds = %L179.i
  %68 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139, !dbg !265
  %.not166 = icmp eq i32 %68, 0, !dbg !265
  %69 = call float @llvm.nvvm.fabs.ftz.f(float %66) #139, !dbg !265
  %70 = call float @llvm.fabs.f32(float %66), !dbg !265
  %.0.i.i = select i1 %.not166, float %70, float %69, !dbg !265
  %71 = fcmp uge float %.0.i.i, 0x7FF0000000000000, !dbg !265
  br i1 %71, label %L197.i, label %L236.i, !dbg !264

L197.i:                                           ; preds = %L190.i, %L179.i
  %72 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139, !dbg !267
  %.not170 = icmp eq i32 %72, 0, !dbg !267
  %73 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi.i) #139, !dbg !267
  %74 = call float @llvm.fabs.f32(float %value_phi.i), !dbg !267
  %.0.i35.i = select i1 %.not170, float %74, float %73, !dbg !267
  %75 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139, !dbg !276
  %.not171 = icmp eq i32 %75, 0, !dbg !276
  %76 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi3.i) #139, !dbg !276
  %77 = call float @llvm.fabs.f32(float %value_phi3.i), !dbg !276
  %.0.i41.i = select i1 %.not171, float %77, float %76, !dbg !276
  %.inv = fcmp olt float %.0.i35.i, %.0.i41.i, !dbg !279
  %78 = select i1 %.inv, float %.0.i41.i, float %.0.i35.i, !dbg !279
  %79 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139, !dbg !276
  %.not171.1 = icmp eq i32 %79, 0, !dbg !276
  %80 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi4.i) #139, !dbg !276
  %81 = call float @llvm.fabs.f32(float %value_phi4.i), !dbg !276
  %.0.i41.i.1 = select i1 %.not171.1, float %81, float %80, !dbg !276
  %.inv.1 = fcmp olt float %78, %.0.i41.i.1, !dbg !279
  %82 = select i1 %.inv.1, float %.0.i41.i.1, float %78, !dbg !279
  %83 = fcmp une float %82, 0.000000e+00, !dbg !285
  br i1 %83, label %L223.i, label %L221.i, !dbg !290

L221.i:                                           ; preds = %L197.i
  %84 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139, !dbg !291
  %.not173 = icmp eq i32 %84, 0, !dbg !291
  %85 = call float @llvm.nvvm.fabs.ftz.f(float noundef 0.000000e+00) #139, !dbg !291
  %.0.i36.i = select i1 %.not173, float 0.000000e+00, float %85, !dbg !291
  br label %L236.i, !dbg !297

L223.i:                                           ; preds = %L197.i
  %86 = fdiv float %value_phi.i, %82, !dbg !301
  %87 = fmul float %86, %86, !dbg !304
  %88 = fdiv float %value_phi3.i, %82, !dbg !301
  %89 = fmul float %88, %88, !dbg !304
  %90 = fadd float %87, %89, !dbg !307
  %91 = fdiv float %value_phi4.i, %82, !dbg !301
  %92 = fmul float %91, %91, !dbg !304
  %93 = fadd float %92, %90, !dbg !307
  %94 = call float @__nv_sqrtf(float %93) #139, !dbg !308
  %95 = fmul float %82, %94, !dbg !309
  br label %L236.i, !dbg !297

L236.i:                                           ; preds = %L221.i, %L223.i, %L190.i
  %value_phi5.i = phi float [ %66, %L190.i ], [ %.0.i36.i, %L221.i ], [ %95, %L223.i ]
  %96 = fcmp une float %4, 0.000000e+00, !dbg !310
  %97 = fcmp ugt float %value_phi5.i, %4
  %or.cond = select i1 %96, i1 %97, i1 false, !dbg !313
  br i1 %or.cond, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L245.i, !dbg !313

L245.i:                                           ; preds = %L236.i
  %98 = shl nsw i64 %219, 2, !dbg !314
  %99 = add i64 %98, -4, !dbg !314
  %100 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract29, i64 %99, !dbg !321
  %101 = bitcast i8 addrspace(1)* %100 to float addrspace(1)*, !dbg !323
  %102 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %101, i32 noundef 4), !dbg !323
  %103 = shl nuw nsw i64 %222, 2, !dbg !314
  %104 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract21, i64 %103, !dbg !321
  %105 = bitcast i8 addrspace(1)* %104 to float addrspace(1)*, !dbg !323
  %106 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %105, i32 noundef 4), !dbg !323
  %107 = fadd float %value_phi5.i, %106, !dbg !329
  %108 = fcmp uge float %102, %107, !dbg !331
  br i1 %108, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L302.i, !dbg !333

L302.i:                                           ; preds = %L245.i
  %109 = fsub float %value_phi5.i, %106, !dbg !334
  %110 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139, !dbg !336
  %.not167 = icmp eq i32 %110, 0, !dbg !336
  %111 = call float @llvm.nvvm.fabs.ftz.f(float %109) #139, !dbg !336
  %112 = call float @llvm.fabs.f32(float %109), !dbg !336
  %.0.i38.i = select i1 %.not167, float %112, float %111, !dbg !336
  %113 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #139, !dbg !337
  %.not168 = icmp eq i32 %113, 0, !dbg !337
  %114 = call float @llvm.nvvm.fmax.ftz.f(float %102, float %.0.i38.i) #139, !dbg !337
  %115 = call float @llvm.maxnum.f32(float %102, float %.0.i38.i), !dbg !337
  %.0.i40.i = select i1 %.not168, float %115, float %114, !dbg !337
  %116 = fdiv float 1.000000e+00, %.0.i40.i, !dbg !339
  %117 = fdiv float 1.000000e+00, %107, !dbg !342
  %118 = fdiv float 1.000000e+00, %value_phi5.i, !dbg !345
  %119 = fmul float %118, %118, !dbg !348
  %120 = fmul float %106, %106, !dbg !351
  %121 = fmul float %119, %120, !dbg !354
  %122 = fadd float %121, 1.000000e+00, !dbg !355
  %123 = fmul float %116, %116, !dbg !351
  %124 = fmul float %117, %117, !dbg !351
  %125 = fsub float %123, %124, !dbg !359
  %126 = fmul float %122, %125, !dbg !354
  %127 = fmul float %126, 1.250000e-01, !dbg !360
  %128 = fdiv float %117, %116, !dbg !363
  %129 = call float @__nv_logf(float %128) #139, !dbg !364
  %130 = fmul float %119, %129, !dbg !354
  %131 = fmul float %130, 2.500000e-01, !dbg !360
  %132 = fadd float %131, %127, !dbg !366
  %133 = bitcast i8 addrspace(1)* %.fca.0.extract60 to float addrspace(1)*, !dbg !367
  %134 = getelementptr inbounds float, float addrspace(1)* %133, i64 %29, !dbg !367
  %135 = load float, float addrspace(1)* %134, align 4, !dbg !367, !tbaa !216
  %136 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract13, i64 %99, !dbg !377
  %137 = bitcast i8 addrspace(1)* %136 to float addrspace(1)*, !dbg !382
  %138 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %137, i32 noundef 4), !dbg !382
  %139 = fmul float %138, %138, !dbg !387
  %140 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract6, i64 %99, !dbg !377
  %141 = bitcast i8 addrspace(1)* %140 to float addrspace(1)*, !dbg !382
  %142 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %141, i32 noundef 4), !dbg !382
  %143 = fmul float %135, %139, !dbg !389
  %144 = fmul float %142, %143, !dbg !389
  %145 = icmp sgt i64 %.fca.2.0.extract, 0, !dbg !393
  %146 = select i1 %145, i64 %.fca.2.0.extract, i64 0, !dbg !393
  %147 = mul i64 %222, %146, !dbg !418
  %148 = add i64 %147, %219, !dbg !423
  %149 = shl i64 %148, 2, !dbg !424
  %150 = add i64 %149, -4, !dbg !424
  %151 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract, i64 %150, !dbg !429
  %152 = bitcast i8 addrspace(1)* %151 to float addrspace(1)*, !dbg !430
  %153 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %152, i32 noundef 4), !dbg !430
  %154 = fsub float %132, %153, !dbg !435
  %155 = fmul float %144, %154, !dbg !436
  %156 = fmul float %118, %155, !dbg !436
  %157 = fneg float %value_phi.i, !dbg !438
  %158 = fneg float %value_phi3.i, !dbg !438
  %159 = fneg float %value_phi4.i, !dbg !438
  %160 = fmul float %156, %157, !dbg !449
  %161 = fmul float %156, %158, !dbg !449
  %162 = fmul float %156, %159, !dbg !449
  %163 = icmp sgt i64 %.fca.2.0.extract70, 0
  %164 = select i1 %163, i64 %.fca.2.0.extract70, i64 0
  %165 = getelementptr inbounds [1 x [2 x [1 x i64]]], [1 x [2 x [1 x i64]]]* %10, i64 0, i64 0, i64 0, i64 0
  %166 = icmp sgt i64 %.fca.2.1.extract71, 0
  %167 = select i1 %166, i64 %.fca.2.1.extract71, i64 0
  %168 = getelementptr inbounds [1 x [2 x [1 x i64]]], [1 x [2 x [1 x i64]]]* %10, i64 0, i64 0, i64 1, i64 0
  %169 = getelementptr inbounds [2 x i64], [2 x i64]* %11, i64 0, i64 0
  %170 = getelementptr inbounds [2 x i64], [2 x i64]* %11, i64 0, i64 1
  %171 = icmp sgt i64 %219, 0
  %172 = icmp sle i64 %219, %167
  %173 = and i1 %171, %172
  %174 = mul i64 %29, %164
  %175 = getelementptr inbounds [1 x [2 x [1 x i64]]], [1 x [2 x [1 x i64]]]* %12, i64 0, i64 0, i64 0, i64 0
  %176 = getelementptr inbounds [1 x [2 x [1 x i64]]], [1 x [2 x [1 x i64]]]* %12, i64 0, i64 0, i64 1, i64 0
  %177 = getelementptr inbounds [2 x i64], [2 x i64]* %13, i64 0, i64 0
  %178 = getelementptr inbounds [2 x i64], [2 x i64]* %13, i64 0, i64 1
  %179 = icmp ult i64 %222, %167
  %180 = mul i64 %222, %164
  br i1 %173, label %L302.i.split.us, label %L302.i.L302.i.split_crit_edge, !dbg !457

L302.i.L302.i.split_crit_edge:                    ; preds = %L302.i
  store i64 %164, i64* %165, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %167, i64* %168, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 1, i64* %169, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %219, i64* %170, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  br label %L524.i, !dbg !487

L302.i.split.us:                                  ; preds = %L302.i
  store i64 %164, i64* %165, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %167, i64* %168, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 1, i64* %169, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %219, i64* %170, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  br i1 %163, label %L523.i.us, label %L524.i, !dbg !487

L523.i.us:                                        ; preds = %L302.i.split.us
  %181 = shl i64 %174, 2, !dbg !488
  %182 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %181, !dbg !494
  %183 = bitcast i8 addrspace(1)* %182 to float addrspace(1)*, !dbg !495
  %184 = atomicrmw fadd float addrspace(1)* %183, float %160 monotonic, align 4, !dbg !495
  store i64 %164, i64* %175, align 8, !dbg !507, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %167, i64* %176, align 8, !dbg !507, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 1, i64* %177, align 8, !dbg !516, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %223, i64* %178, align 8, !dbg !516, !tbaa !473, !alias.scope !475, !noalias !478
  br i1 %179, label %L615.i.us, label %L616.i, !dbg !519

L615.i.us:                                        ; preds = %L523.i.us
  %185 = fneg float %160, !dbg !520
  %186 = shl i64 %180, 2, !dbg !521
  %187 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %186, !dbg !525
  %188 = bitcast i8 addrspace(1)* %187 to float addrspace(1)*, !dbg !526
  %189 = atomicrmw fadd float addrspace(1)* %188, float %185 monotonic, align 4, !dbg !526
  store i64 %164, i64* %165, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %167, i64* %168, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 2, i64* %169, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %219, i64* %170, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  %190 = icmp ugt i64 %164, 1, !dbg !532
  br i1 %190, label %L523.i.us.1, label %L524.i, !dbg !487

L523.i.us.1:                                      ; preds = %L615.i.us
  %191 = shl i64 %174, 2, !dbg !488
  %192 = add i64 %191, 4, !dbg !488
  %193 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %192, !dbg !494
  %194 = bitcast i8 addrspace(1)* %193 to float addrspace(1)*, !dbg !495
  %195 = atomicrmw fadd float addrspace(1)* %194, float %161 monotonic, align 4, !dbg !495
  store i64 %164, i64* %175, align 8, !dbg !507, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %167, i64* %176, align 8, !dbg !507, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 2, i64* %177, align 8, !dbg !516, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %223, i64* %178, align 8, !dbg !516, !tbaa !473, !alias.scope !475, !noalias !478
  %196 = fneg float %161, !dbg !520
  %197 = shl i64 %180, 2, !dbg !521
  %198 = add i64 %197, 4, !dbg !521
  %199 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %198, !dbg !525
  %200 = bitcast i8 addrspace(1)* %199 to float addrspace(1)*, !dbg !526
  %201 = atomicrmw fadd float addrspace(1)* %200, float %196 monotonic, align 4, !dbg !526
  store i64 %164, i64* %165, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %167, i64* %168, align 8, !dbg !458, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 3, i64* %169, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %219, i64* %170, align 8, !dbg !483, !tbaa !473, !alias.scope !475, !noalias !478
  %.not187 = icmp eq i64 %164, 2, !dbg !532
  br i1 %.not187, label %L524.i, label %L523.i.us.2, !dbg !487

L523.i.us.2:                                      ; preds = %L523.i.us.1
  %202 = shl i64 %174, 2, !dbg !488
  %203 = add i64 %202, 8, !dbg !488
  %204 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %203, !dbg !494
  %205 = bitcast i8 addrspace(1)* %204 to float addrspace(1)*, !dbg !495
  %206 = atomicrmw fadd float addrspace(1)* %205, float %162 monotonic, align 4, !dbg !495
  store i64 %164, i64* %175, align 8, !dbg !507, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %167, i64* %176, align 8, !dbg !507, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 3, i64* %177, align 8, !dbg !516, !tbaa !473, !alias.scope !475, !noalias !478
  store i64 %223, i64* %178, align 8, !dbg !516, !tbaa !473, !alias.scope !475, !noalias !478
  %207 = fneg float %162, !dbg !520
  %208 = shl i64 %180, 2, !dbg !521
  %209 = add i64 %208, 8, !dbg !521
  %210 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %209, !dbg !525
  %211 = bitcast i8 addrspace(1)* %210 to float addrspace(1)*, !dbg !526
  %212 = atomicrmw fadd float addrspace(1)* %211, float %207 monotonic, align 4, !dbg !526
  br label %julia_gbsa_force_2_kernel__25091_inner.exit, !dbg !539

L524.i:                                           ; preds = %L302.i.split.us, %L615.i.us, %L523.i.us.1, %L302.i.L302.i.split_crit_edge
  call fastcc void @julia__throw_boundserror_25134() #145, !dbg !487
  unreachable, !dbg !487

L616.i:                                           ; preds = %L523.i.us
  call fastcc void @julia__throw_boundserror_25134() #145, !dbg !519
  unreachable, !dbg !519

fail.i:                                           ; preds = %L33.i
  call fastcc void @gpu_report_exception(i64 noundef ptrtoint ([10 x i8]* @exception6 to i64)), !dbg !188
  call fastcc void @gpu_signal_exception(), !dbg !188
  call void asm sideeffect "exit;", ""() #139, !dbg !188
  unreachable, !dbg !188

pass.i:                                           ; preds = %L33.i
  %213 = sdiv i64 %28, %.fca.2.0.extract49, !dbg !188
  %214 = icmp sgt i64 %.fca.2.0.extract49, 0, !dbg !540
  %215 = mul i64 %213, %.fca.2.0.extract49, !dbg !545
  %216 = icmp ne i64 %215, %28, !dbg !546
  %217 = and i1 %214, %216, !dbg !550
  %218 = zext i1 %217 to i64, !dbg !553
  %219 = add nsw i64 %213, %218, !dbg !562
  %cond = icmp eq i64 %.fca.2.0.extract49, -1, !dbg !564
  br i1 %cond, label %after_srem.i, label %oksrem.i, !dbg !564

oksrem.i:                                         ; preds = %pass.i
  %220 = add nsw i64 %28, -1, !dbg !566
  %221 = srem i64 %220, %.fca.2.0.extract49, !dbg !564
  br label %after_srem.i, !dbg !564

after_srem.i:                                     ; preds = %pass.i, %oksrem.i
  %222 = phi i64 [ %221, %oksrem.i ], [ 0, %pass.i ], !dbg !564
  %223 = add nuw nsw i64 %222, 1, !dbg !567
  %224 = icmp eq i64 %219, %223, !dbg !568
  br i1 %224, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L50.i, !dbg !571

julia_gbsa_force_2_kernel__25091_inner.exit:      ; preds = %L523.i.us.2, %L236.i, %entry, %L245.i, %after_srem.i
  call void @llvm.lifetime.end.p0i8(i64 noundef 16, i8* noundef nonnull %14), !dbg !539
  call void @llvm.lifetime.end.p0i8(i64 noundef 16, i8* noundef nonnull %15), !dbg !539
  call void @llvm.lifetime.end.p0i8(i64 noundef 16, i8* noundef nonnull %16), !dbg !539
  call void @llvm.lifetime.end.p0i8(i64 noundef 16, i8* noundef nonnull %17), !dbg !539
  ret void, !dbg !132
}

; Function Attrs: argmemonly nofree nounwind willreturn
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #2

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.sqrt.f32(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.sqrt.f64(double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.fabs.f32(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.fma.f32(float, float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.fshl.i32(i32, i32, i32) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.trunc.f32(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.maxnum.f32(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.minnum.f32(float, float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.round.f32(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare float @llvm.floor.f32(float) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i32 @llvm.abs.i32(i32, i1 immarg) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare i64 @llvm.fshl.i64(i64, i64, i64) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.fma.f64(double, double, double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.fabs.f64(double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.maxnum.f64(double, double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.minnum.f64(double, double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.trunc.f64(double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.floor.f64(double) #6

; Function Attrs: nofree nosync nounwind readnone speculatable willreturn
declare double @llvm.round.f64(double) #6

; Function Attrs: mustprogress willreturn
define void @preprocess_julia_gbsa_force_2_kernel__25091_inner24({ i8 addrspace(1)*, i64, [2 x i64], i64 } %0, { i8 addrspace(1)*, i64, [1 x i64], i64 } %1, { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, [1 x [1 x [3 x float]]] %3, float %4, { i8 addrspace(1)*, i64, [1 x i64], i64 } %5, { i8 addrspace(1)*, i64, [1 x i64], i64 } %6, { i8 addrspace(1)*, i64, [1 x i64], i64 } %7, { i8 addrspace(1)*, i64, [1 x i64], i64 } %8, { i8 addrspace(1)*, i64, [2 x i64], i64 } %9) local_unnamed_addr #138 !dbg !572 {
entry:
  %.fca.0.extract68 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 0, !dbg !573
  %.fca.2.0.extract70 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 2, 0, !dbg !573
  %.fca.2.1.extract71 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 2, 1, !dbg !573
  %.fca.0.extract60 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %1, 0, !dbg !573
  %.fca.0.extract47 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, 0, !dbg !573
  %.fca.2.0.extract49 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, 2, 0, !dbg !573
  %.fca.0.0.0.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 0, !dbg !573
  %.fca.0.0.1.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 1, !dbg !573
  %.fca.0.0.2.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 2, !dbg !573
  %.fca.0.extract29 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %5, 0, !dbg !573
  %.fca.0.extract21 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %6, 0, !dbg !573
  %.fca.0.extract13 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %7, 0, !dbg !573
  %.fca.0.extract6 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %8, 0, !dbg !573
  %.fca.0.extract = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %9, 0, !dbg !573
  %.fca.2.0.extract = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %9, 2, 0, !dbg !573
  %10 = call {}*** @julia.get_pgcstack() #146
  %11 = mul i64 %.fca.2.0.extract49, %.fca.2.0.extract49, !dbg !574
  %12 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x() #146, !dbg !578, !range !153
  %13 = zext i32 %12 to i64, !dbg !584
  %14 = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x() #146, !dbg !586, !range !164
  %15 = zext i32 %14 to i64, !dbg !591
  %16 = mul nuw nsw i64 %13, %15, !dbg !593
  %17 = call i32 @llvm.nvvm.read.ptx.sreg.tid.x() #146, !dbg !595, !range !177
  %18 = add nuw nsw i32 %17, 1, !dbg !600
  %19 = zext i32 %18 to i64, !dbg !601
  %20 = add nuw nsw i64 %16, %19, !dbg !603
  %.not = icmp sgt i64 %20, %11, !dbg !605
  br i1 %.not, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L33.i, !dbg !607

L33.i:                                            ; preds = %entry
  %.not153 = icmp eq i64 %.fca.2.0.extract49, 0, !dbg !608
  br i1 %.not153, label %fail.i, label %pass.i, !dbg !608

L50.i:                                            ; preds = %after_srem.i
  %21 = add nsw i64 %203, -1, !dbg !613
  %22 = bitcast i8 addrspace(1)* %.fca.0.extract47 to [1 x [3 x float]] addrspace(1)*, !dbg !622
  %.unpack.elt = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %21, i64 0, i64 0, !dbg !622
  %.unpack.unpack = load float, float addrspace(1)* %.unpack.elt, align 4, !dbg !622, !tbaa !216
  %.unpack.elt154 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %21, i64 0, i64 1, !dbg !622
  %.unpack.unpack155 = load float, float addrspace(1)* %.unpack.elt154, align 4, !dbg !622, !tbaa !216
  %.unpack.elt156 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %21, i64 0, i64 2, !dbg !622
  %.unpack.unpack157 = load float, float addrspace(1)* %.unpack.elt156, align 4, !dbg !622, !tbaa !216
  %.unpack.elt159 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %206, i64 0, i64 0, !dbg !622
  %.unpack.unpack160 = load float, float addrspace(1)* %.unpack.elt159, align 4, !dbg !622, !tbaa !216
  %.unpack.elt161 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %206, i64 0, i64 1, !dbg !622
  %.unpack.unpack162 = load float, float addrspace(1)* %.unpack.elt161, align 4, !dbg !622, !tbaa !216
  %.unpack.elt163 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %206, i64 0, i64 2, !dbg !622
  %.unpack.unpack164 = load float, float addrspace(1)* %.unpack.elt163, align 4, !dbg !622, !tbaa !216
  %23 = fcmp uge float %.unpack.unpack, %.unpack.unpack160, !dbg !623
  br i1 %23, label %L109.i, label %L99.i, !dbg !627

L99.i:                                            ; preds = %L50.i
  %24 = fsub float %.unpack.unpack160, %.unpack.unpack, !dbg !628
  %25 = fsub float %.unpack.unpack, %.unpack.unpack160, !dbg !628
  %26 = fadd float %.fca.0.0.0.extract, %25, !dbg !630
  %27 = fcmp uge float %24, %26, !dbg !631
  br i1 %27, label %L106.i, label %L119.i, !dbg !632

L106.i:                                           ; preds = %L99.i
  %28 = fsub float %24, %.fca.0.0.0.extract, !dbg !628
  br label %L119.i, !dbg !632

L109.i:                                           ; preds = %L50.i
  %29 = fsub float %.unpack.unpack, %.unpack.unpack160, !dbg !633
  %30 = fsub float %.unpack.unpack160, %.unpack.unpack, !dbg !633
  %31 = fadd float %.fca.0.0.0.extract, %30, !dbg !635
  %32 = fcmp uge float %29, %31, !dbg !636
  %. = select i1 %32, float %31, float %30, !dbg !637
  br label %L119.i, !dbg !637

L119.i:                                           ; preds = %L109.i, %L106.i, %L99.i
  %value_phi.i = phi float [ %28, %L106.i ], [ %24, %L99.i ], [ %., %L109.i ]
  %33 = fcmp uge float %.unpack.unpack155, %.unpack.unpack162, !dbg !623
  br i1 %33, label %L139.i, label %L129.i, !dbg !627

L129.i:                                           ; preds = %L119.i
  %34 = fsub float %.unpack.unpack162, %.unpack.unpack155, !dbg !628
  %35 = fsub float %.unpack.unpack155, %.unpack.unpack162, !dbg !628
  %36 = fadd float %.fca.0.0.1.extract, %35, !dbg !630
  %37 = fcmp uge float %34, %36, !dbg !631
  br i1 %37, label %L136.i, label %L149.i, !dbg !632

L136.i:                                           ; preds = %L129.i
  %38 = fsub float %34, %.fca.0.0.1.extract, !dbg !628
  br label %L149.i, !dbg !632

L139.i:                                           ; preds = %L119.i
  %39 = fsub float %.unpack.unpack155, %.unpack.unpack162, !dbg !633
  %40 = fsub float %.unpack.unpack162, %.unpack.unpack155, !dbg !633
  %41 = fadd float %.fca.0.0.1.extract, %40, !dbg !635
  %42 = fcmp uge float %39, %41, !dbg !636
  %.185 = select i1 %42, float %41, float %40, !dbg !637
  br label %L149.i, !dbg !637

L149.i:                                           ; preds = %L139.i, %L136.i, %L129.i
  %value_phi3.i = phi float [ %38, %L136.i ], [ %34, %L129.i ], [ %.185, %L139.i ]
  %43 = fcmp uge float %.unpack.unpack157, %.unpack.unpack164, !dbg !623
  br i1 %43, label %L169.i, label %L159.i, !dbg !627

L159.i:                                           ; preds = %L149.i
  %44 = fsub float %.unpack.unpack164, %.unpack.unpack157, !dbg !628
  %45 = fsub float %.unpack.unpack157, %.unpack.unpack164, !dbg !628
  %46 = fadd float %.fca.0.0.2.extract, %45, !dbg !630
  %47 = fcmp uge float %44, %46, !dbg !631
  br i1 %47, label %L166.i, label %L179.i, !dbg !632

L166.i:                                           ; preds = %L159.i
  %48 = fsub float %44, %.fca.0.0.2.extract, !dbg !628
  br label %L179.i, !dbg !632

L169.i:                                           ; preds = %L149.i
  %49 = fsub float %.unpack.unpack157, %.unpack.unpack164, !dbg !633
  %50 = fsub float %.unpack.unpack164, %.unpack.unpack157, !dbg !633
  %51 = fadd float %.fca.0.0.2.extract, %50, !dbg !635
  %52 = fcmp uge float %49, %51, !dbg !636
  %.186 = select i1 %52, float %51, float %50, !dbg !637
  br label %L179.i, !dbg !637

L179.i:                                           ; preds = %L169.i, %L166.i, %L159.i
  %value_phi4.i = phi float [ %48, %L166.i ], [ %44, %L159.i ], [ %.186, %L169.i ]
  %53 = fmul float %value_phi.i, %value_phi.i, !dbg !638
  %54 = fmul float %value_phi3.i, %value_phi3.i, !dbg !638
  %55 = fadd float %53, %54, !dbg !645
  %56 = fmul float %value_phi4.i, %value_phi4.i, !dbg !638
  %57 = fadd float %55, %56, !dbg !645
  %58 = call float @__nv_sqrtf(float %57) #147, !dbg !646
  %59 = fcmp ule float %58, 0.000000e+00, !dbg !647
  br i1 %59, label %L197.i, label %L190.i, !dbg !649

L190.i:                                           ; preds = %L179.i
  %60 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !650
  %.not166 = icmp eq i32 %60, 0, !dbg !650
  %61 = call float @llvm.nvvm.fabs.ftz.f(float %58) #147, !dbg !650
  %62 = call float @llvm.fabs.f32(float %58) #146, !dbg !650
  %.0.i.i = select i1 %.not166, float %62, float %61, !dbg !650
  %63 = fcmp uge float %.0.i.i, 0x7FF0000000000000, !dbg !650
  br i1 %63, label %L197.i, label %L236.i, !dbg !649

L197.i:                                           ; preds = %L190.i, %L179.i
  %64 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !651
  %.not170 = icmp eq i32 %64, 0, !dbg !651
  %65 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi.i) #147, !dbg !651
  %66 = call float @llvm.fabs.f32(float %value_phi.i) #146, !dbg !651
  %.0.i35.i = select i1 %.not170, float %66, float %65, !dbg !651
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !657
  %.not171 = icmp eq i32 %67, 0, !dbg !657
  %68 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi3.i) #147, !dbg !657
  %69 = call float @llvm.fabs.f32(float %value_phi3.i) #146, !dbg !657
  %.0.i41.i = select i1 %.not171, float %69, float %68, !dbg !657
  %.inv = fcmp olt float %.0.i35.i, %.0.i41.i, !dbg !660
  %70 = select i1 %.inv, float %.0.i41.i, float %.0.i35.i, !dbg !660
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !657
  %.not171.1 = icmp eq i32 %71, 0, !dbg !657
  %72 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi4.i) #147, !dbg !657
  %73 = call float @llvm.fabs.f32(float %value_phi4.i) #146, !dbg !657
  %.0.i41.i.1 = select i1 %.not171.1, float %73, float %72, !dbg !657
  %.inv.1 = fcmp olt float %70, %.0.i41.i.1, !dbg !660
  %74 = select i1 %.inv.1, float %.0.i41.i.1, float %70, !dbg !660
  %75 = fcmp une float %74, 0.000000e+00, !dbg !662
  br i1 %75, label %L223.i, label %L221.i, !dbg !665

L221.i:                                           ; preds = %L197.i
  %76 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !666
  %.not173 = icmp eq i32 %76, 0, !dbg !666
  %77 = call float @llvm.nvvm.fabs.ftz.f(float noundef 0.000000e+00) #147, !dbg !666
  %.0.i36.i = select i1 %.not173, float 0.000000e+00, float %77, !dbg !666
  br label %L236.i, !dbg !670

L223.i:                                           ; preds = %L197.i
  %78 = fdiv float %value_phi.i, %74, !dbg !672
  %79 = fmul float %78, %78, !dbg !674
  %80 = fdiv float %value_phi3.i, %74, !dbg !672
  %81 = fmul float %80, %80, !dbg !674
  %82 = fadd float %79, %81, !dbg !677
  %83 = fdiv float %value_phi4.i, %74, !dbg !672
  %84 = fmul float %83, %83, !dbg !674
  %85 = fadd float %84, %82, !dbg !677
  %86 = call float @__nv_sqrtf(float %85) #147, !dbg !678
  %87 = fmul float %74, %86, !dbg !679
  br label %L236.i, !dbg !670

L236.i:                                           ; preds = %L223.i, %L221.i, %L190.i
  %value_phi5.i = phi float [ %58, %L190.i ], [ %.0.i36.i, %L221.i ], [ %87, %L223.i ]
  %88 = fcmp une float %4, 0.000000e+00, !dbg !680
  %89 = fcmp ugt float %value_phi5.i, %4
  %or.cond = select i1 %88, i1 %89, i1 false, !dbg !683
  br i1 %or.cond, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L245.i, !dbg !683

L245.i:                                           ; preds = %L236.i
  %90 = shl nsw i64 %203, 2, !dbg !684
  %91 = add i64 %90, -4, !dbg !684
  %92 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract29, i64 %91, !dbg !690
  %93 = bitcast i8 addrspace(1)* %92 to float addrspace(1)*, !dbg !691
  %94 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %93, i32 noundef 4) #146, !dbg !691
  %95 = shl nuw nsw i64 %206, 2, !dbg !684
  %96 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract21, i64 %95, !dbg !690
  %97 = bitcast i8 addrspace(1)* %96 to float addrspace(1)*, !dbg !691
  %98 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %97, i32 noundef 4) #146, !dbg !691
  %99 = fadd float %value_phi5.i, %98, !dbg !696
  %100 = fcmp uge float %94, %99, !dbg !698
  br i1 %100, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L302.i, !dbg !700

L302.i:                                           ; preds = %L245.i
  %101 = fsub float %value_phi5.i, %98, !dbg !701
  %102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !703
  %.not167 = icmp eq i32 %102, 0, !dbg !703
  %103 = call float @llvm.nvvm.fabs.ftz.f(float %101) #147, !dbg !703
  %104 = call float @llvm.fabs.f32(float %101) #146, !dbg !703
  %.0.i38.i = select i1 %.not167, float %104, float %103, !dbg !703
  %105 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !704
  %.not168 = icmp eq i32 %105, 0, !dbg !704
  %106 = call float @llvm.nvvm.fmax.ftz.f(float %94, float %.0.i38.i) #147, !dbg !704
  %107 = call float @llvm.maxnum.f32(float %94, float %.0.i38.i) #146, !dbg !704
  %.0.i40.i = select i1 %.not168, float %107, float %106, !dbg !704
  %108 = fdiv float 1.000000e+00, %.0.i40.i, !dbg !705
  %109 = fdiv float 1.000000e+00, %99, !dbg !707
  %110 = fdiv float 1.000000e+00, %value_phi5.i, !dbg !710
  %111 = fmul float %110, %110, !dbg !713
  %112 = fmul float %98, %98, !dbg !716
  %113 = fmul float %111, %112, !dbg !719
  %114 = fadd float %113, 1.000000e+00, !dbg !720
  %115 = fmul float %108, %108, !dbg !716
  %116 = fmul float %109, %109, !dbg !716
  %117 = fsub float %115, %116, !dbg !722
  %118 = fmul float %114, %117, !dbg !719
  %119 = fmul float %118, 1.250000e-01, !dbg !723
  %120 = fdiv float %109, %108, !dbg !725
  %121 = call float @__nv_logf(float %120) #147, !dbg !726
  %122 = fmul float %111, %121, !dbg !719
  %123 = fmul float %122, 2.500000e-01, !dbg !723
  %124 = fadd float %123, %119, !dbg !727
  %125 = bitcast i8 addrspace(1)* %.fca.0.extract60 to float addrspace(1)*, !dbg !728
  %126 = getelementptr inbounds float, float addrspace(1)* %125, i64 %21, !dbg !728
  %127 = load float, float addrspace(1)* %126, align 4, !dbg !728, !tbaa !216
  %128 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract13, i64 %91, !dbg !736
  %129 = bitcast i8 addrspace(1)* %128 to float addrspace(1)*, !dbg !741
  %130 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %129, i32 noundef 4) #146, !dbg !741
  %131 = fmul float %130, %130, !dbg !746
  %132 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract6, i64 %91, !dbg !736
  %133 = bitcast i8 addrspace(1)* %132 to float addrspace(1)*, !dbg !741
  %134 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %133, i32 noundef 4) #146, !dbg !741
  %135 = fmul float %127, %131, !dbg !748
  %136 = fmul float %134, %135, !dbg !748
  %137 = icmp sgt i64 %.fca.2.0.extract, 0, !dbg !750
  %138 = select i1 %137, i64 %.fca.2.0.extract, i64 0, !dbg !750
  %139 = mul i64 %206, %138, !dbg !763
  %140 = add i64 %139, %203, !dbg !767
  %141 = shl i64 %140, 2, !dbg !768
  %142 = add i64 %141, -4, !dbg !768
  %143 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract, i64 %142, !dbg !773
  %144 = bitcast i8 addrspace(1)* %143 to float addrspace(1)*, !dbg !774
  %145 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %144, i32 noundef 4) #146, !dbg !774
  %146 = fsub float %124, %145, !dbg !779
  %147 = fmul float %136, %146, !dbg !780
  %148 = fmul float %110, %147, !dbg !780
  %149 = fneg float %value_phi.i, !dbg !782
  %150 = fneg float %value_phi3.i, !dbg !782
  %151 = fneg float %value_phi4.i, !dbg !782
  %152 = fmul float %148, %149, !dbg !788
  %153 = fmul float %148, %150, !dbg !788
  %154 = fmul float %148, %151, !dbg !788
  %155 = icmp sgt i64 %.fca.2.0.extract70, 0
  %156 = select i1 %155, i64 %.fca.2.0.extract70, i64 0
  %157 = icmp sgt i64 %.fca.2.1.extract71, 0
  %158 = select i1 %157, i64 %.fca.2.1.extract71, i64 0
  %159 = icmp sgt i64 %203, 0
  %160 = icmp sle i64 %203, %158
  %161 = and i1 %159, %160
  %162 = mul i64 %21, %156
  %163 = icmp ult i64 %206, %158
  %164 = mul i64 %206, %156
  %.not14 = xor i1 %161, true, !dbg !794
  %.not15 = xor i1 %155, true, !dbg !794
  %brmerge = select i1 %.not14, i1 true, i1 %.not15, !dbg !794
  br i1 %brmerge, label %L524.i, label %L523.i.us, !dbg !794

L523.i.us:                                        ; preds = %L302.i
  %165 = shl i64 %162, 2, !dbg !795
  %166 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %165, !dbg !801
  %167 = bitcast i8 addrspace(1)* %166 to float addrspace(1)*, !dbg !802
  %168 = atomicrmw fadd float addrspace(1)* %167, float %152 monotonic, align 4, !dbg !802
  br i1 %163, label %L615.i.us, label %L616.i, !dbg !808

L615.i.us:                                        ; preds = %L523.i.us
  %169 = fneg float %152, !dbg !814
  %170 = shl i64 %164, 2, !dbg !815
  %171 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %170, !dbg !819
  %172 = bitcast i8 addrspace(1)* %171 to float addrspace(1)*, !dbg !820
  %173 = atomicrmw fadd float addrspace(1)* %172, float %169 monotonic, align 4, !dbg !820
  %174 = icmp ugt i64 %156, 1, !dbg !826
  br i1 %174, label %L523.i.us.1, label %L524.i, !dbg !834

L523.i.us.1:                                      ; preds = %L615.i.us
  %175 = shl i64 %162, 2, !dbg !795
  %176 = add i64 %175, 4, !dbg !795
  %177 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %176, !dbg !801
  %178 = bitcast i8 addrspace(1)* %177 to float addrspace(1)*, !dbg !802
  %179 = atomicrmw fadd float addrspace(1)* %178, float %153 monotonic, align 4, !dbg !802
  %180 = fneg float %153, !dbg !814
  %181 = shl i64 %164, 2, !dbg !815
  %182 = add i64 %181, 4, !dbg !815
  %183 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %182, !dbg !819
  %184 = bitcast i8 addrspace(1)* %183 to float addrspace(1)*, !dbg !820
  %185 = atomicrmw fadd float addrspace(1)* %184, float %180 monotonic, align 4, !dbg !820
  %.not187 = icmp eq i64 %156, 2, !dbg !826
  br i1 %.not187, label %L524.i, label %L523.i.us.2, !dbg !834

L523.i.us.2:                                      ; preds = %L523.i.us.1
  %186 = shl i64 %162, 2, !dbg !795
  %187 = add i64 %186, 8, !dbg !795
  %188 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %187, !dbg !801
  %189 = bitcast i8 addrspace(1)* %188 to float addrspace(1)*, !dbg !802
  %190 = atomicrmw fadd float addrspace(1)* %189, float %154 monotonic, align 4, !dbg !802
  %191 = fneg float %154, !dbg !814
  %192 = shl i64 %164, 2, !dbg !815
  %193 = add i64 %192, 8, !dbg !815
  %194 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %193, !dbg !819
  %195 = bitcast i8 addrspace(1)* %194 to float addrspace(1)*, !dbg !820
  %196 = atomicrmw fadd float addrspace(1)* %195, float %191 monotonic, align 4, !dbg !820
  br label %julia_gbsa_force_2_kernel__25091_inner.exit, !dbg !835

L524.i:                                           ; preds = %L302.i, %L523.i.us.1, %L615.i.us
  call fastcc void @julia__throw_boundserror_25134() #148, !dbg !834
  unreachable, !dbg !834

L616.i:                                           ; preds = %L523.i.us
  call fastcc void @julia__throw_boundserror_25134() #148, !dbg !808
  unreachable, !dbg !808

fail.i:                                           ; preds = %L33.i
  call fastcc void @gpu_report_exception(i64 noundef ptrtoint ([10 x i8]* @exception6 to i64)) #146, !dbg !608
  call fastcc void @gpu_signal_exception() #146, !dbg !608
  unreachable, !dbg !608

pass.i:                                           ; preds = %L33.i
  %197 = sdiv i64 %20, %.fca.2.0.extract49, !dbg !608
  %198 = icmp sgt i64 %.fca.2.0.extract49, 0, !dbg !836
  %199 = mul i64 %197, %.fca.2.0.extract49, !dbg !839
  %200 = icmp ne i64 %199, %20, !dbg !840
  %201 = and i1 %198, %200, !dbg !842
  %202 = zext i1 %201 to i64, !dbg !843
  %203 = add nsw i64 %197, %202, !dbg !848
  %cond = icmp eq i64 %.fca.2.0.extract49, -1, !dbg !850
  br i1 %cond, label %after_srem.i, label %oksrem.i, !dbg !850

oksrem.i:                                         ; preds = %pass.i
  %204 = add nsw i64 %20, -1, !dbg !852
  %205 = srem i64 %204, %.fca.2.0.extract49, !dbg !850
  br label %after_srem.i, !dbg !850

after_srem.i:                                     ; preds = %oksrem.i, %pass.i
  %206 = phi i64 [ %205, %oksrem.i ], [ 0, %pass.i ], !dbg !850
  %207 = add nuw nsw i64 %206, 1, !dbg !853
  %208 = icmp eq i64 %203, %207, !dbg !854
  br i1 %208, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L50.i, !dbg !857

julia_gbsa_force_2_kernel__25091_inner.exit:      ; preds = %after_srem.i, %L523.i.us.2, %L245.i, %L236.i, %entry
  ret void, !dbg !573
}

; Function Attrs: mustprogress willreturn
define internal void @diffejulia_gbsa_force_2_kernel__25091_inner24({ i8 addrspace(1)*, i64, [2 x i64], i64 } %0, { i8 addrspace(1)*, i64, [2 x i64], i64 } %"'", { i8 addrspace(1)*, i64, [1 x i64], i64 } %1, { i8 addrspace(1)*, i64, [1 x i64], i64 } %"'1", { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, { i8 addrspace(1)*, i64, [1 x i64], i64 } %"'2", [1 x [1 x [3 x float]]] %3, float %4, { i8 addrspace(1)*, i64, [1 x i64], i64 } %5, { i8 addrspace(1)*, i64, [1 x i64], i64 } %"'3", { i8 addrspace(1)*, i64, [1 x i64], i64 } %6, { i8 addrspace(1)*, i64, [1 x i64], i64 } %"'4", { i8 addrspace(1)*, i64, [1 x i64], i64 } %7, { i8 addrspace(1)*, i64, [1 x i64], i64 } %"'5", { i8 addrspace(1)*, i64, [1 x i64], i64 } %8, { i8 addrspace(1)*, i64, [1 x i64], i64 } %"'6", { i8 addrspace(1)*, i64, [2 x i64], i64 } %9, { i8 addrspace(1)*, i64, [2 x i64], i64 } %"'7") local_unnamed_addr #138 !dbg !858 {
entry:
  %.fca.0.extract68 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 0, !dbg !859
  %.fca.2.0.extract70 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 2, 0, !dbg !859
  %.fca.2.1.extract71 = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %0, 2, 1, !dbg !859
  %.fca.0.extract60 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %1, 0, !dbg !859
  %".fca.0.extract47'ipev" = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %"'2", 0, !dbg !859
  %.fca.0.extract47 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, 0, !dbg !859
  %.fca.2.0.extract49 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %2, 2, 0, !dbg !859
  %.fca.0.0.0.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 0, !dbg !859
  %.fca.0.0.1.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 1, !dbg !859
  %.fca.0.0.2.extract = extractvalue [1 x [1 x [3 x float]]] %3, 0, 0, 2, !dbg !859
  %.fca.0.extract29 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %5, 0, !dbg !859
  %.fca.0.extract21 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %6, 0, !dbg !859
  %.fca.0.extract13 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %7, 0, !dbg !859
  %.fca.0.extract6 = extractvalue { i8 addrspace(1)*, i64, [1 x i64], i64 } %8, 0, !dbg !859
  %.fca.0.extract = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %9, 0, !dbg !859
  %.fca.2.0.extract = extractvalue { i8 addrspace(1)*, i64, [2 x i64], i64 } %9, 2, 0, !dbg !859
  %10 = call {}*** @julia.get_pgcstack() #146
  %11 = mul i64 %.fca.2.0.extract49, %.fca.2.0.extract49, !dbg !860
  %12 = call i32 @llvm.nvvm.read.ptx.sreg.ctaid.x() #146, !dbg !864, !range !153
  %13 = zext i32 %12 to i64, !dbg !870
  %14 = call i32 @llvm.nvvm.read.ptx.sreg.ntid.x() #146, !dbg !872, !range !164
  %15 = zext i32 %14 to i64, !dbg !877
  %16 = mul nuw nsw i64 %13, %15, !dbg !879
  %17 = call i32 @llvm.nvvm.read.ptx.sreg.tid.x() #146, !dbg !881, !range !177
  %18 = add nuw nsw i32 %17, 1, !dbg !886
  %19 = zext i32 %18 to i64, !dbg !887
  %20 = add nuw nsw i64 %16, %19, !dbg !889
  %.not = icmp sgt i64 %20, %11, !dbg !891
  br i1 %.not, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L33.i, !dbg !893

L33.i:                                            ; preds = %entry
  %.not153 = icmp eq i64 %.fca.2.0.extract49, 0, !dbg !894
  br i1 %.not153, label %fail.i, label %pass.i, !dbg !894

L50.i:                                            ; preds = %after_srem.i
  %21 = add nsw i64 %203, -1, !dbg !899
  %"'ipc" = bitcast i8 addrspace(1)* %".fca.0.extract47'ipev" to [1 x [3 x float]] addrspace(1)*, !dbg !908
  %22 = bitcast i8 addrspace(1)* %.fca.0.extract47 to [1 x [3 x float]] addrspace(1)*, !dbg !908
  %".unpack.elt'ipg" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc", i64 %21, i64 0, i64 0, !dbg !908
  %.unpack.elt = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %21, i64 0, i64 0, !dbg !908
  %.unpack.unpack = load float, float addrspace(1)* %.unpack.elt, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912
  store float %.unpack.unpack, float* %.unpack.unpack_cache, align 4, !dbg !908, !tbaa !216, !invariant.group !914
  %".unpack.elt154'ipg" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc", i64 %21, i64 0, i64 1, !dbg !908
  %.unpack.elt154 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %21, i64 0, i64 1, !dbg !908
  %.unpack.unpack155 = load float, float addrspace(1)* %.unpack.elt154, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912
  %".unpack.elt156'ipg" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc", i64 %21, i64 0, i64 2, !dbg !908
  %.unpack.elt156 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %21, i64 0, i64 2, !dbg !908
  %.unpack.unpack157 = load float, float addrspace(1)* %.unpack.elt156, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912
  %".unpack.elt159'ipg" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc", i64 %206, i64 0, i64 0, !dbg !908
  %.unpack.elt159 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %206, i64 0, i64 0, !dbg !908
  %.unpack.unpack160 = load float, float addrspace(1)* %.unpack.elt159, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912
  store float %.unpack.unpack160, float* %.unpack.unpack160_cache, align 4, !dbg !908, !tbaa !216, !invariant.group !915
  store float %.unpack.unpack157, float* %.unpack.unpack157_cache, align 4, !dbg !908, !tbaa !216, !invariant.group !916
  store float %.unpack.unpack155, float* %.unpack.unpack155_cache, align 4, !dbg !908, !tbaa !216, !invariant.group !917
  %".unpack.elt161'ipg" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc", i64 %206, i64 0, i64 1, !dbg !908
  %.unpack.elt161 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %206, i64 0, i64 1, !dbg !908
  %.unpack.unpack162 = load float, float addrspace(1)* %.unpack.elt161, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912
  store float %.unpack.unpack162, float* %.unpack.unpack162_cache, align 4, !dbg !908, !tbaa !216, !invariant.group !918
  %".unpack.elt163'ipg" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc", i64 %206, i64 0, i64 2, !dbg !908
  %.unpack.elt163 = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %22, i64 %206, i64 0, i64 2, !dbg !908
  %.unpack.unpack164 = load float, float addrspace(1)* %.unpack.elt163, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912
  store float %.unpack.unpack164, float* %.unpack.unpack164_cache, align 4, !dbg !919, !tbaa !216, !invariant.group !923
  %23 = fcmp uge float %.unpack.unpack, %.unpack.unpack160, !dbg !919
  br i1 %23, label %L109.i, label %L99.i, !dbg !924

L99.i:                                            ; preds = %L50.i
  %24 = fsub float %.unpack.unpack160, %.unpack.unpack, !dbg !925
  %25 = fsub float %.unpack.unpack, %.unpack.unpack160, !dbg !925
  %26 = fadd float %.fca.0.0.0.extract, %25, !dbg !927
  %27 = fcmp uge float %24, %26, !dbg !928
  br i1 %27, label %L106.i, label %L119.i, !dbg !929

L106.i:                                           ; preds = %L99.i
  %28 = fsub float %24, %.fca.0.0.0.extract, !dbg !925
  br label %L119.i, !dbg !929

L109.i:                                           ; preds = %L50.i
  %29 = fsub float %.unpack.unpack, %.unpack.unpack160, !dbg !930
  %30 = fsub float %.unpack.unpack160, %.unpack.unpack, !dbg !930
  %31 = fadd float %.fca.0.0.0.extract, %30, !dbg !932
  %32 = fcmp uge float %29, %31, !dbg !933
  %. = select i1 %32, float %31, float %30, !dbg !934
  br label %L119.i, !dbg !934

L119.i:                                           ; preds = %L109.i, %L106.i, %L99.i
  %value_phi.i = phi float [ %28, %L106.i ], [ %24, %L99.i ], [ %., %L109.i ]
  %33 = fcmp uge float %.unpack.unpack155, %.unpack.unpack162, !dbg !919
  br i1 %33, label %L139.i, label %L129.i, !dbg !924

L129.i:                                           ; preds = %L119.i
  %34 = fsub float %.unpack.unpack162, %.unpack.unpack155, !dbg !925
  %35 = fsub float %.unpack.unpack155, %.unpack.unpack162, !dbg !925
  %36 = fadd float %.fca.0.0.1.extract, %35, !dbg !927
  %37 = fcmp uge float %34, %36, !dbg !928
  br i1 %37, label %L136.i, label %L149.i, !dbg !929

L136.i:                                           ; preds = %L129.i
  %38 = fsub float %34, %.fca.0.0.1.extract, !dbg !925
  br label %L149.i, !dbg !929

L139.i:                                           ; preds = %L119.i
  %39 = fsub float %.unpack.unpack155, %.unpack.unpack162, !dbg !930
  %40 = fsub float %.unpack.unpack162, %.unpack.unpack155, !dbg !930
  %41 = fadd float %.fca.0.0.1.extract, %40, !dbg !932
  %42 = fcmp uge float %39, %41, !dbg !933
  %.185 = select i1 %42, float %41, float %40, !dbg !934
  br label %L149.i, !dbg !934

L149.i:                                           ; preds = %L139.i, %L136.i, %L129.i
  %value_phi3.i = phi float [ %38, %L136.i ], [ %34, %L129.i ], [ %.185, %L139.i ]
  %43 = fcmp uge float %.unpack.unpack157, %.unpack.unpack164, !dbg !919
  br i1 %43, label %L169.i, label %L159.i, !dbg !924

L159.i:                                           ; preds = %L149.i
  %44 = fsub float %.unpack.unpack164, %.unpack.unpack157, !dbg !925
  %45 = fsub float %.unpack.unpack157, %.unpack.unpack164, !dbg !925
  %46 = fadd float %.fca.0.0.2.extract, %45, !dbg !927
  %47 = fcmp uge float %44, %46, !dbg !928
  br i1 %47, label %L166.i, label %L179.i, !dbg !929

L166.i:                                           ; preds = %L159.i
  %48 = fsub float %44, %.fca.0.0.2.extract, !dbg !925
  br label %L179.i, !dbg !929

L169.i:                                           ; preds = %L149.i
  %49 = fsub float %.unpack.unpack157, %.unpack.unpack164, !dbg !930
  %50 = fsub float %.unpack.unpack164, %.unpack.unpack157, !dbg !930
  %51 = fadd float %.fca.0.0.2.extract, %50, !dbg !932
  %52 = fcmp uge float %49, %51, !dbg !933
  %.186 = select i1 %52, float %51, float %50, !dbg !934
  br label %L179.i, !dbg !934

L179.i:                                           ; preds = %L169.i, %L166.i, %L159.i
  %value_phi4.i = phi float [ %48, %L166.i ], [ %44, %L159.i ], [ %.186, %L169.i ]
  %53 = fmul float %value_phi.i, %value_phi.i, !dbg !935
  %54 = fmul float %value_phi3.i, %value_phi3.i, !dbg !935
  %55 = fadd float %53, %54, !dbg !942
  %56 = fmul float %value_phi4.i, %value_phi4.i, !dbg !935
  %57 = fadd float %55, %56, !dbg !942
  %58 = call float @__nv_sqrtf(float %57) #147, !dbg !943
  %59 = fcmp ule float %58, 0.000000e+00, !dbg !944
  br i1 %59, label %L197.i, label %L190.i, !dbg !946

L190.i:                                           ; preds = %L179.i
  %60 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !947
  %.not166 = icmp eq i32 %60, 0, !dbg !947
  %61 = call float @llvm.nvvm.fabs.ftz.f(float %58) #147, !dbg !947
  %62 = call float @llvm.fabs.f32(float %58) #146, !dbg !947
  %.0.i.i = select i1 %.not166, float %62, float %61, !dbg !947
  %63 = fcmp uge float %.0.i.i, 0x7FF0000000000000, !dbg !947
  br i1 %63, label %L197.i, label %L236.i, !dbg !946

L197.i:                                           ; preds = %L190.i, %L179.i
  %64 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !948
  %"'ip_phi" = phi i32 , !dbg !948
  %.not170 = icmp eq i32 %64, 0, !dbg !948
  store i1 %.not170, i1* %.not170_cache, align 1, !dbg !948, !invariant.group !954
  %65 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi.i) #147, !dbg !948
  %66 = call float @llvm.fabs.f32(float %value_phi.i) #146, !dbg !948
  %.0.i35.i = select i1 %.not170, float %66, float %65, !dbg !948
  %67 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !955
  %"'ip_phi8" = phi i32 , !dbg !955
  %.not171 = icmp eq i32 %67, 0, !dbg !955
  store i1 %.not171, i1* %.not171_cache, align 1, !dbg !955, !invariant.group !958
  %68 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi3.i) #147, !dbg !955
  %69 = call float @llvm.fabs.f32(float %value_phi3.i) #146, !dbg !955
  %.0.i41.i = select i1 %.not171, float %69, float %68, !dbg !955
  %.inv = fcmp olt float %.0.i35.i, %.0.i41.i, !dbg !959
  %70 = select i1 %.inv, float %.0.i41.i, float %.0.i35.i, !dbg !959
  %71 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !955
  %"'ip_phi9" = phi i32 , !dbg !955
  %.not171.1 = icmp eq i32 %71, 0, !dbg !955
  store i1 %.not171.1, i1* %.not171.1_cache, align 1, !dbg !955, !invariant.group !961
  %72 = call float @llvm.nvvm.fabs.ftz.f(float %value_phi4.i) #147, !dbg !955
  %73 = call float @llvm.fabs.f32(float %value_phi4.i) #146, !dbg !955
  %.0.i41.i.1 = select i1 %.not171.1, float %73, float %72, !dbg !955
  %.inv.1 = fcmp olt float %70, %.0.i41.i.1, !dbg !959
  %74 = select i1 %.inv.1, float %.0.i41.i.1, float %70, !dbg !959
  %75 = fcmp une float %74, 0.000000e+00, !dbg !962
  br i1 %75, label %L223.i, label %L221.i, !dbg !965

L221.i:                                           ; preds = %L197.i
  %76 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !966
  %"'ip_phi10" = phi i32 , !dbg !966
  %.not173 = icmp eq i32 %76, 0, !dbg !966
  %77 = call float @llvm.nvvm.fabs.ftz.f(float noundef 0.000000e+00) #147, !dbg !966
  %.0.i36.i = select i1 %.not173, float 0.000000e+00, float %77, !dbg !966
  br label %L236.i, !dbg !970

L223.i:                                           ; preds = %L197.i
  %78 = fdiv float %value_phi.i, %74, !dbg !972
  %79 = fmul float %78, %78, !dbg !974
  %80 = fdiv float %value_phi3.i, %74, !dbg !972
  %81 = fmul float %80, %80, !dbg !974
  %82 = fadd float %79, %81, !dbg !977
  %83 = fdiv float %value_phi4.i, %74, !dbg !972
  %84 = fmul float %83, %83, !dbg !974
  %85 = fadd float %84, %82, !dbg !977
  %86 = call float @__nv_sqrtf(float %85) #147, !dbg !978
  %87 = fmul float %74, %86, !dbg !979
  br label %L236.i, !dbg !970

L236.i:                                           ; preds = %L223.i, %L221.i, %L190.i
  %value_phi5.i = phi float [ %58, %L190.i ], [ %.0.i36.i, %L221.i ], [ %87, %L223.i ]
  %88 = fcmp une float %4, 0.000000e+00, !dbg !980
  %89 = fcmp ugt float %value_phi5.i, %4
  %or.cond = select i1 %88, i1 %89, i1 false, !dbg !983
  br i1 %or.cond, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L245.i, !dbg !983

L245.i:                                           ; preds = %L236.i
  %90 = shl nsw i64 %203, 2, !dbg !984
  %91 = add i64 %90, -4, !dbg !984
  %92 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract29, i64 %91, !dbg !990
  %93 = bitcast i8 addrspace(1)* %92 to float addrspace(1)*, !dbg !991
  %94 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %93, i32 noundef 4) #146, !dbg !991
  %95 = shl nuw nsw i64 %206, 2, !dbg !984
  %96 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract21, i64 %95, !dbg !990
  %97 = bitcast i8 addrspace(1)* %96 to float addrspace(1)*, !dbg !991
  %98 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %97, i32 noundef 4) #146, !dbg !991
  %99 = fadd float %value_phi5.i, %98, !dbg !996
  %100 = fcmp uge float %94, %99, !dbg !998
  br i1 %100, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L302.i, !dbg !1000

L302.i:                                           ; preds = %L245.i
  %101 = fsub float %value_phi5.i, %98, !dbg !1001
  %102 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !1003
  %"'ip_phi11" = phi i32 , !dbg !1003
  %.not167 = icmp eq i32 %102, 0, !dbg !1003
  %103 = call float @llvm.nvvm.fabs.ftz.f(float %101) #147, !dbg !1003
  %104 = call float @llvm.fabs.f32(float %101) #146, !dbg !1003
  %.0.i38.i = select i1 %.not167, float %104, float %103, !dbg !1003
  %105 = call i32 @__nvvm_reflect(i8* noundef getelementptr inbounds ([11 x i8], [11 x i8]* @.str, i64 0, i64 0)) #147, !dbg !1004
  %"'ip_phi12" = phi i32 , !dbg !1004
  %.not168 = icmp eq i32 %105, 0, !dbg !1004
  %106 = call float @llvm.nvvm.fmax.ftz.f(float %94, float %.0.i38.i) #147, !dbg !1004
  %107 = call float @llvm.maxnum.f32(float %94, float %.0.i38.i) #146, !dbg !1004
  %.0.i40.i = select i1 %.not168, float %107, float %106, !dbg !1004
  %108 = fdiv float 1.000000e+00, %.0.i40.i, !dbg !1005
  %109 = fdiv float 1.000000e+00, %99, !dbg !1007
  %110 = fdiv float 1.000000e+00, %value_phi5.i, !dbg !1010
  %111 = fmul float %110, %110, !dbg !1013
  %112 = fmul float %98, %98, !dbg !1016
  %113 = fmul float %111, %112, !dbg !1019
  %114 = fadd float %113, 1.000000e+00, !dbg !1020
  %115 = fmul float %108, %108, !dbg !1016
  %116 = fmul float %109, %109, !dbg !1016
  %117 = fsub float %115, %116, !dbg !1022
  %118 = fmul float %114, %117, !dbg !1019
  %119 = fmul float %118, 1.250000e-01, !dbg !1023
  %120 = fdiv float %109, %108, !dbg !1025
  %121 = call float @__nv_logf(float %120) #147, !dbg !1026
  %122 = fmul float %111, %121, !dbg !1019
  %123 = fmul float %122, 2.500000e-01, !dbg !1023
  %124 = fadd float %123, %119, !dbg !1027
  %125 = bitcast i8 addrspace(1)* %.fca.0.extract60 to float addrspace(1)*, !dbg !1028
  %126 = getelementptr inbounds float, float addrspace(1)* %125, i64 %21, !dbg !1028
  %127 = load float, float addrspace(1)* %126, align 4, !dbg !1028, !tbaa !216
  %128 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract13, i64 %91, !dbg !1036
  %129 = bitcast i8 addrspace(1)* %128 to float addrspace(1)*, !dbg !1041
  %130 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %129, i32 noundef 4) #146, !dbg !1041
  %131 = fmul float %130, %130, !dbg !1046
  %132 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract6, i64 %91, !dbg !1036
  %133 = bitcast i8 addrspace(1)* %132 to float addrspace(1)*, !dbg !1041
  %134 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %133, i32 noundef 4) #146, !dbg !1041
  %135 = fmul float %127, %131, !dbg !1048
  %136 = fmul float %134, %135, !dbg !1048
  %137 = icmp sgt i64 %.fca.2.0.extract, 0, !dbg !1050
  %138 = select i1 %137, i64 %.fca.2.0.extract, i64 0, !dbg !1050
  %139 = mul i64 %206, %138, !dbg !1063
  %140 = add i64 %139, %203, !dbg !1067
  %141 = shl i64 %140, 2, !dbg !1068
  %142 = add i64 %141, -4, !dbg !1068
  %143 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract, i64 %142, !dbg !1073
  %144 = bitcast i8 addrspace(1)* %143 to float addrspace(1)*, !dbg !1074
  %145 = call float @llvm.nvvm.ldg.global.f.f32.p1f32(float addrspace(1)* %144, i32 noundef 4) #146, !dbg !1074
  %146 = fsub float %124, %145, !dbg !1079
  %147 = fmul float %136, %146, !dbg !1080
  %148 = fmul float %110, %147, !dbg !1080
  %149 = fneg float %value_phi.i, !dbg !1082
  %150 = fneg float %value_phi3.i, !dbg !1082
  %151 = fneg float %value_phi4.i, !dbg !1082
  %152 = fmul float %148, %149, !dbg !1088
  %153 = fmul float %148, %150, !dbg !1088
  %154 = fmul float %148, %151, !dbg !1088
  %155 = icmp sgt i64 %.fca.2.0.extract70, 0
  %156 = select i1 %155, i64 %.fca.2.0.extract70, i64 0
  %157 = icmp sgt i64 %.fca.2.1.extract71, 0
  %158 = select i1 %157, i64 %.fca.2.1.extract71, i64 0
  %159 = icmp sgt i64 %203, 0
  %160 = icmp sle i64 %203, %158
  %161 = and i1 %159, %160
  %162 = mul i64 %21, %156
  %163 = icmp ult i64 %206, %158
  %164 = mul i64 %206, %156
  %.not14 = xor i1 %161, true, !dbg !1094
  %.not15 = xor i1 %155, true, !dbg !1094
  %brmerge = select i1 %.not14, i1 true, i1 %.not15, !dbg !1094
  br i1 %brmerge, label %L524.i, label %L523.i.us, !dbg !1094

L523.i.us:                                        ; preds = %L302.i
  %165 = shl i64 %162, 2, !dbg !1095
  %166 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %165, !dbg !1101
  %167 = bitcast i8 addrspace(1)* %166 to float addrspace(1)*, !dbg !1102
  %168 = atomicrmw fadd float addrspace(1)* %167, float %152 monotonic, align 4, !dbg !1102
  br i1 %163, label %L615.i.us, label %L616.i, !dbg !1108

L615.i.us:                                        ; preds = %L523.i.us
  %169 = fneg float %152, !dbg !1114
  %170 = shl i64 %164, 2, !dbg !1115
  %171 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %170, !dbg !1119
  %172 = bitcast i8 addrspace(1)* %171 to float addrspace(1)*, !dbg !1120
  %173 = atomicrmw fadd float addrspace(1)* %172, float %169 monotonic, align 4, !dbg !1120
  %174 = icmp ugt i64 %156, 1, !dbg !1126
  br i1 %174, label %L523.i.us.1, label %L524.i, !dbg !1134

L523.i.us.1:                                      ; preds = %L615.i.us
  %175 = shl i64 %162, 2, !dbg !1095
  %176 = add i64 %175, 4, !dbg !1095
  %177 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %176, !dbg !1101
  %178 = bitcast i8 addrspace(1)* %177 to float addrspace(1)*, !dbg !1102
  %179 = atomicrmw fadd float addrspace(1)* %178, float %153 monotonic, align 4, !dbg !1102
  %180 = fneg float %153, !dbg !1114
  %181 = shl i64 %164, 2, !dbg !1115
  %182 = add i64 %181, 4, !dbg !1115
  %183 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %182, !dbg !1119
  %184 = bitcast i8 addrspace(1)* %183 to float addrspace(1)*, !dbg !1120
  %185 = atomicrmw fadd float addrspace(1)* %184, float %180 monotonic, align 4, !dbg !1120
  %.not187 = icmp eq i64 %156, 2, !dbg !1126
  br i1 %.not187, label %L524.i, label %L523.i.us.2, !dbg !1134

L523.i.us.2:                                      ; preds = %L523.i.us.1
  %186 = shl i64 %162, 2, !dbg !1095
  %187 = add i64 %186, 8, !dbg !1095
  %188 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %187, !dbg !1101
  %189 = bitcast i8 addrspace(1)* %188 to float addrspace(1)*, !dbg !1102
  %190 = atomicrmw fadd float addrspace(1)* %189, float %154 monotonic, align 4, !dbg !1102
  %191 = fneg float %154, !dbg !1114
  %192 = shl i64 %164, 2, !dbg !1115
  %193 = add i64 %192, 8, !dbg !1115
  %194 = getelementptr i8, i8 addrspace(1)* %.fca.0.extract68, i64 %193, !dbg !1119
  %195 = bitcast i8 addrspace(1)* %194 to float addrspace(1)*, !dbg !1120
  %196 = atomicrmw fadd float addrspace(1)* %195, float %191 monotonic, align 4, !dbg !1120
  br label %julia_gbsa_force_2_kernel__25091_inner.exit, !dbg !1135

L524.i:                                           ; preds = %L523.i.us.1, %L615.i.us, %L302.i
  call fastcc void @julia__throw_boundserror_25134() #148, !dbg !1134
  unreachable, !dbg !1134

L616.i:                                           ; preds = %L523.i.us
  call fastcc void @julia__throw_boundserror_25134() #148, !dbg !1108
  unreachable, !dbg !1108

fail.i:                                           ; preds = %L33.i
  call fastcc void @gpu_report_exception(i64 noundef ptrtoint ([10 x i8]* @exception6 to i64)) #146, !dbg !894
  call fastcc void @gpu_signal_exception() #146, !dbg !894
  unreachable, !dbg !894

pass.i:                                           ; preds = %L33.i
  %197 = sdiv i64 %20, %.fca.2.0.extract49, !dbg !894
  %198 = icmp sgt i64 %.fca.2.0.extract49, 0, !dbg !1136
  %199 = mul i64 %197, %.fca.2.0.extract49, !dbg !1139
  %200 = icmp ne i64 %199, %20, !dbg !1140
  %201 = and i1 %198, %200, !dbg !1142
  %202 = zext i1 %201 to i64, !dbg !1143
  %203 = add nsw i64 %197, %202, !dbg !1148
  %cond = icmp eq i64 %.fca.2.0.extract49, -1, !dbg !1150
  br i1 %cond, label %after_srem.i, label %oksrem.i, !dbg !1150

oksrem.i:                                         ; preds = %pass.i
  %204 = add nsw i64 %20, -1, !dbg !1152
  %205 = srem i64 %204, %.fca.2.0.extract49, !dbg !1150
  br label %after_srem.i, !dbg !1150

after_srem.i:                                     ; preds = %oksrem.i, %pass.i
  %206 = phi i64 [ %205, %oksrem.i ], [ 0, %pass.i ], !dbg !1150
  %207 = add nuw nsw i64 %206, 1, !dbg !1153
  %208 = icmp eq i64 %203, %207, !dbg !1154
  br i1 %208, label %julia_gbsa_force_2_kernel__25091_inner.exit, label %L50.i, !dbg !1157

julia_gbsa_force_2_kernel__25091_inner.exit:      ; preds = %after_srem.i, %L523.i.us.2, %L245.i, %L236.i, %entry
  br label %invertjulia_gbsa_force_2_kernel__25091_inner.exit, !dbg !859

allocsForInversion:                               ; No predecessors!
  %".unpack.unpack164'de" = alloca float, align 4
  %209 = getelementptr float, float* %".unpack.unpack164'de", i64 0
  store float 0.000000e+00, float* %209, align 4
  %".unpack.unpack162'de" = alloca float, align 4
  %210 = getelementptr float, float* %".unpack.unpack162'de", i64 0
  store float 0.000000e+00, float* %210, align 4
  %".unpack.unpack160'de" = alloca float, align 4
  %211 = getelementptr float, float* %".unpack.unpack160'de", i64 0
  store float 0.000000e+00, float* %211, align 4
  %".unpack.unpack157'de" = alloca float, align 4
  %212 = getelementptr float, float* %".unpack.unpack157'de", i64 0
  store float 0.000000e+00, float* %212, align 4
  %".unpack.unpack155'de" = alloca float, align 4
  %213 = getelementptr float, float* %".unpack.unpack155'de", i64 0
  store float 0.000000e+00, float* %213, align 4
  %".unpack.unpack'de" = alloca float, align 4
  %214 = getelementptr float, float* %".unpack.unpack'de", i64 0
  store float 0.000000e+00, float* %214, align 4
  %"'de" = alloca float, align 4
  %215 = getelementptr float, float* %"'de", i64 0
  store float 0.000000e+00, float* %215, align 4
  %"'de23" = alloca float, align 4
  %216 = getelementptr float, float* %"'de23", i64 0
  store float 0.000000e+00, float* %216, align 4
  %".'de" = alloca float, align 4
  %217 = getelementptr float, float* %".'de", i64 0
  store float 0.000000e+00, float* %217, align 4
  %.unpack.unpack_cache = alloca float, align 4
  store float 0.000000e+00, float* %.unpack.unpack_cache, align 4
  %.unpack.unpack160_cache = alloca float, align 4
  store float 0.000000e+00, float* %.unpack.unpack160_cache, align 4
  %"'de29" = alloca float, align 4
  %218 = getelementptr float, float* %"'de29", i64 0
  store float 0.000000e+00, float* %218, align 4
  %"'de30" = alloca float, align 4
  %219 = getelementptr float, float* %"'de30", i64 0
  store float 0.000000e+00, float* %219, align 4
  %"value_phi.i'de" = alloca float, align 4
  %220 = getelementptr float, float* %"value_phi.i'de", i64 0
  store float 0.000000e+00, float* %220, align 4
  %"'de38" = alloca float, align 4
  %221 = getelementptr float, float* %"'de38", i64 0
  store float 0.000000e+00, float* %221, align 4
  %"'de39" = alloca float, align 4
  %222 = getelementptr float, float* %"'de39", i64 0
  store float 0.000000e+00, float* %222, align 4
  %".185'de" = alloca float, align 4
  %223 = getelementptr float, float* %".185'de", i64 0
  store float 0.000000e+00, float* %223, align 4
  %.unpack.unpack155_cache = alloca float, align 4
  store float 0.000000e+00, float* %.unpack.unpack155_cache, align 4
  %.unpack.unpack162_cache = alloca float, align 4
  store float 0.000000e+00, float* %.unpack.unpack162_cache, align 4
  %"'de46" = alloca float, align 4
  %224 = getelementptr float, float* %"'de46", i64 0
  store float 0.000000e+00, float* %224, align 4
  %"'de47" = alloca float, align 4
  %225 = getelementptr float, float* %"'de47", i64 0
  store float 0.000000e+00, float* %225, align 4
  %"value_phi3.i'de" = alloca float, align 4
  %226 = getelementptr float, float* %"value_phi3.i'de", i64 0
  store float 0.000000e+00, float* %226, align 4
  %"'de60" = alloca float, align 4
  %227 = getelementptr float, float* %"'de60", i64 0
  store float 0.000000e+00, float* %227, align 4
  %"'de61" = alloca float, align 4
  %228 = getelementptr float, float* %"'de61", i64 0
  store float 0.000000e+00, float* %228, align 4
  %".186'de" = alloca float, align 4
  %229 = getelementptr float, float* %".186'de", i64 0
  store float 0.000000e+00, float* %229, align 4
  %.unpack.unpack157_cache = alloca float, align 4
  store float 0.000000e+00, float* %.unpack.unpack157_cache, align 4
  %.unpack.unpack164_cache = alloca float, align 4
  store float 0.000000e+00, float* %.unpack.unpack164_cache, align 4
  %"'de68" = alloca float, align 4
  %230 = getelementptr float, float* %"'de68", i64 0
  store float 0.000000e+00, float* %230, align 4
  %"'de69" = alloca float, align 4
  %231 = getelementptr float, float* %"'de69", i64 0
  store float 0.000000e+00, float* %231, align 4
  %"'de70" = alloca float, align 4
  %232 = getelementptr float, float* %"'de70", i64 0
  store float 0.000000e+00, float* %232, align 4
  %"'de112" = alloca float, align 4
  %233 = getelementptr float, float* %"'de112", i64 0
  store float 0.000000e+00, float* %233, align 4
  %"'de113" = alloca float, align 4
  %234 = getelementptr float, float* %"'de113", i64 0
  store float 0.000000e+00, float* %234, align 4
  %"'de114" = alloca float, align 4
  %235 = getelementptr float, float* %"'de114", i64 0
  store float 0.000000e+00, float* %235, align 4
  %"value_phi4.i'de" = alloca float, align 4
  %236 = getelementptr float, float* %"value_phi4.i'de", i64 0
  store float 0.000000e+00, float* %236, align 4
  %"'de115" = alloca float, align 4
  %237 = getelementptr float, float* %"'de115", i64 0
  store float 0.000000e+00, float* %237, align 4
  %"'de116" = alloca float, align 4
  %238 = getelementptr float, float* %"'de116", i64 0
  store float 0.000000e+00, float* %238, align 4
  %"'de124" = alloca float, align 4
  %239 = getelementptr float, float* %"'de124", i64 0
  store float 0.000000e+00, float* %239, align 4
  %.not170_cache = alloca i1, align 1
  store i1 false, i1* %.not170_cache, align 1
  %.not171_cache = alloca i1, align 1
  store i1 false, i1* %.not171_cache, align 1
  %.not171.1_cache = alloca i1, align 1
  store i1 false, i1* %.not171.1_cache, align 1
  %".0.i41.i.1'de" = alloca float, align 4
  %240 = getelementptr float, float* %".0.i41.i.1'de", i64 0
  store float 0.000000e+00, float* %240, align 4
  %"'de166" = alloca float, align 4
  %241 = getelementptr float, float* %"'de166", i64 0
  store float 0.000000e+00, float* %241, align 4
  %"'de169" = alloca float, align 4
  %242 = getelementptr float, float* %"'de169", i64 0
  store float 0.000000e+00, float* %242, align 4
  %"'de170" = alloca float, align 4
  %243 = getelementptr float, float* %"'de170", i64 0
  store float 0.000000e+00, float* %243, align 4

invertentry:                                      ; preds = %invertL33.i
  ret void

invertL33.i:                                      ; No predecessors!
  br label %invertentry

invertL50.i:                                      ; preds = %invertL109.i, %invertL99.i
  %244 = load float, float* %".unpack.unpack164'de", align 4
  store float 0.000000e+00, float* %".unpack.unpack164'de", align 4
  %"'ipc_unwrap" = bitcast i8 addrspace(1)* %".fca.0.extract47'ipev" to [1 x [3 x float]] addrspace(1)*
  %cond_unwrap = icmp eq i64 %.fca.2.0.extract49, -1
  br i1 %cond_unwrap, label %invertL50.i_phirc, label %invertL50.i_phirc13

invertL50.i_phirc:                                ; preds = %invertL50.i
  br label %invertL50.i_phimerge

invertL50.i_phirc13:                              ; preds = %invertL50.i
  %_unwrap = add nsw i64 %20, -1
  %_unwrap14 = srem i64 %_unwrap, %.fca.2.0.extract49
  br label %invertL50.i_phimerge

invertL50.i_phimerge:                             ; preds = %invertL50.i_phirc13, %invertL50.i_phirc
  %245 = phi i64 [ 0, %invertL50.i_phirc ], [ %_unwrap14, %invertL50.i_phirc13 ]
  %".unpack.elt163'ipg_unwrap" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc_unwrap", i64 %245, i64 0, i64 2
  %246 = atomicrmw fadd float addrspace(1)* %".unpack.elt163'ipg_unwrap", float %244 monotonic, align 4
  %247 = load float, float* %".unpack.unpack162'de", align 4
  store float 0.000000e+00, float* %".unpack.unpack162'de", align 4
  %".unpack.elt161'ipg_unwrap" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc_unwrap", i64 %245, i64 0, i64 1
  %248 = atomicrmw fadd float addrspace(1)* %".unpack.elt161'ipg_unwrap", float %247 monotonic, align 4
  %249 = load float, float* %".unpack.unpack160'de", align 4
  store float 0.000000e+00, float* %".unpack.unpack160'de", align 4
  %".unpack.elt159'ipg_unwrap" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc_unwrap", i64 %245, i64 0, i64 0
  %250 = atomicrmw fadd float addrspace(1)* %".unpack.elt159'ipg_unwrap", float %249 monotonic, align 4
  %251 = load float, float* %".unpack.unpack157'de", align 4
  store float 0.000000e+00, float* %".unpack.unpack157'de", align 4
  %_unwrap15 = sdiv i64 %20, %.fca.2.0.extract49
  %_unwrap16 = icmp sgt i64 %.fca.2.0.extract49, 0
  %_unwrap17 = mul i64 %_unwrap15, %.fca.2.0.extract49
  %_unwrap18 = icmp ne i64 %_unwrap17, %20
  %_unwrap19 = and i1 %_unwrap16, %_unwrap18
  %_unwrap20 = zext i1 %_unwrap19 to i64
  %_unwrap21 = add nsw i64 %_unwrap15, %_unwrap20
  %_unwrap22 = add nsw i64 %_unwrap21, -1
  %".unpack.elt156'ipg_unwrap" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc_unwrap", i64 %_unwrap22, i64 0, i64 2
  %252 = atomicrmw fadd float addrspace(1)* %".unpack.elt156'ipg_unwrap", float %251 monotonic, align 4
  %253 = load float, float* %".unpack.unpack155'de", align 4
  store float 0.000000e+00, float* %".unpack.unpack155'de", align 4
  %".unpack.elt154'ipg_unwrap" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc_unwrap", i64 %_unwrap22, i64 0, i64 1
  %254 = atomicrmw fadd float addrspace(1)* %".unpack.elt154'ipg_unwrap", float %253 monotonic, align 4
  %255 = load float, float* %".unpack.unpack'de", align 4
  store float 0.000000e+00, float* %".unpack.unpack'de", align 4
  %".unpack.elt'ipg_unwrap" = getelementptr inbounds [1 x [3 x float]], [1 x [3 x float]] addrspace(1)* %"'ipc_unwrap", i64 %_unwrap22, i64 0, i64 0
  %256 = atomicrmw fadd float addrspace(1)* %".unpack.elt'ipg_unwrap", float %255 monotonic, align 4
  br label %invertafter_srem.i

invertL99.i:                                      ; preds = %staging, %invertL106.i
  %257 = load float, float* %"'de", align 4
  %258 = fneg fast float %257
  store float 0.000000e+00, float* %"'de", align 4
  %259 = load float, float* %".unpack.unpack160'de", align 4
  %260 = fadd fast float %259, %257
  store float %260, float* %".unpack.unpack160'de", align 4
  %261 = load float, float* %".unpack.unpack'de", align 4
  %262 = fadd fast float %261, %258
  store float %262, float* %".unpack.unpack'de", align 4
  br label %invertL50.i

invertL106.i:                                     ; preds = %staging
  %263 = load float, float* %"'de23", align 4
  store float 0.000000e+00, float* %"'de23", align 4
  %264 = load float, float* %"'de", align 4
  %265 = fadd fast float %264, %263
  store float %265, float* %"'de", align 4
  br label %invertL99.i

invertL109.i:                                     ; preds = %invertL119.i
  %266 = load float, float* %".'de", align 4
  %267 = load float, float* %.unpack.unpack_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !914
  %268 = load float, float* %.unpack.unpack160_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !915
  %_unwrap24 = fsub float %267, %268
  %_unwrap25 = fsub float %268, %267
  %_unwrap26 = fadd float %.fca.0.0.0.extract, %_unwrap25
  %_unwrap27 = fcmp uge float %_unwrap24, %_unwrap26
  %diffe = select fast i1 %_unwrap27, float %266, float 0.000000e+00
  %269 = load float, float* %".'de", align 4
  %diffe28 = select fast i1 %_unwrap27, float 0.000000e+00, float %269
  store float 0.000000e+00, float* %".'de", align 4
  %270 = load float, float* %"'de29", align 4
  %271 = fadd fast float %270, %266
  %272 = select fast i1 %_unwrap27, float %271, float %270
  store float %272, float* %"'de29", align 4
  %273 = load float, float* %"'de30", align 4
  %274 = fadd fast float %273, %269
  %275 = select fast i1 %_unwrap27, float %273, float %274
  store float %275, float* %"'de30", align 4
  %276 = load float, float* %"'de29", align 4
  store float 0.000000e+00, float* %"'de29", align 4
  %277 = load float, float* %"'de30", align 4
  %278 = fadd fast float %277, %276
  store float %278, float* %"'de30", align 4
  %279 = load float, float* %"'de30", align 4
  %280 = fneg fast float %279
  store float 0.000000e+00, float* %"'de30", align 4
  %281 = load float, float* %".unpack.unpack160'de", align 4
  %282 = fadd fast float %281, %279
  store float %282, float* %".unpack.unpack160'de", align 4
  %283 = load float, float* %".unpack.unpack'de", align 4
  %284 = fadd fast float %283, %280
  store float %284, float* %".unpack.unpack'de", align 4
  br label %invertL50.i

invertL119.i:                                     ; preds = %invertL139.i, %invertL129.i
  %285 = load float, float* %"value_phi.i'de", align 4
  store float 0.000000e+00, float* %"value_phi.i'de", align 4
  %286 = load float, float* %.unpack.unpack_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !914
  %287 = load float, float* %.unpack.unpack160_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !915
  %_unwrap33 = fcmp uge float %286, %287
  %_unwrap34 = fsub float %287, %286
  %_unwrap35 = fsub float %286, %287
  %_unwrap36 = fadd float %.fca.0.0.0.extract, %_unwrap35
  %_unwrap37 = fcmp uge float %_unwrap34, %_unwrap36
  %anot1_ = xor i1 %_unwrap33, true
  %andVal0 = and i1 %_unwrap37, %anot1_
  %bnot1_ = xor i1 %_unwrap37, true
  %andVal1 = and i1 %bnot1_, %anot1_
  %288 = select fast i1 %_unwrap33, float %285, float 0.000000e+00
  %289 = load float, float* %".'de", align 4
  %290 = fadd fast float %289, %285
  %291 = select fast i1 %_unwrap33, float %290, float %289
  store float %291, float* %".'de", align 4
  %292 = select fast i1 %andVal0, float %285, float 0.000000e+00
  %293 = load float, float* %"'de23", align 4
  %294 = fadd fast float %293, %285
  %295 = select fast i1 %andVal0, float %294, float %293
  store float %295, float* %"'de23", align 4
  %296 = select fast i1 %andVal1, float %285, float 0.000000e+00
  %297 = load float, float* %"'de", align 4
  %298 = fadd fast float %297, %285
  %299 = select fast i1 %andVal1, float %298, float %297
  store float %299, float* %"'de", align 4
  br i1 %_unwrap33, label %invertL109.i, label %staging

invertL129.i:                                     ; preds = %staging59, %invertL136.i
  %300 = load float, float* %"'de38", align 4
  %301 = fneg fast float %300
  store float 0.000000e+00, float* %"'de38", align 4
  %302 = load float, float* %".unpack.unpack162'de", align 4
  %303 = fadd fast float %302, %300
  store float %303, float* %".unpack.unpack162'de", align 4
  %304 = load float, float* %".unpack.unpack155'de", align 4
  %305 = fadd fast float %304, %301
  store float %305, float* %".unpack.unpack155'de", align 4
  br label %invertL119.i

invertL136.i:                                     ; preds = %staging59
  %306 = load float, float* %"'de39", align 4
  store float 0.000000e+00, float* %"'de39", align 4
  %307 = load float, float* %"'de38", align 4
  %308 = fadd fast float %307, %306
  store float %308, float* %"'de38", align 4
  br label %invertL129.i

invertL139.i:                                     ; preds = %invertL149.i
  %309 = load float, float* %".185'de", align 4
  %310 = load float, float* %.unpack.unpack155_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !917
  %311 = load float, float* %.unpack.unpack162_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !918
  %_unwrap40 = fsub float %310, %311
  %_unwrap41 = fsub float %311, %310
  %_unwrap42 = fadd float %.fca.0.0.1.extract, %_unwrap41
  %_unwrap43 = fcmp uge float %_unwrap40, %_unwrap42
  %diffe44 = select fast i1 %_unwrap43, float %309, float 0.000000e+00
  %312 = load float, float* %".185'de", align 4
  %diffe45 = select fast i1 %_unwrap43, float 0.000000e+00, float %312
  store float 0.000000e+00, float* %".185'de", align 4
  %313 = load float, float* %"'de46", align 4
  %314 = fadd fast float %313, %309
  %315 = select fast i1 %_unwrap43, float %314, float %313
  store float %315, float* %"'de46", align 4
  %316 = load float, float* %"'de47", align 4
  %317 = fadd fast float %316, %312
  %318 = select fast i1 %_unwrap43, float %316, float %317
  store float %318, float* %"'de47", align 4
  %319 = load float, float* %"'de46", align 4
  store float 0.000000e+00, float* %"'de46", align 4
  %320 = load float, float* %"'de47", align 4
  %321 = fadd fast float %320, %319
  store float %321, float* %"'de47", align 4
  %322 = load float, float* %"'de47", align 4
  %323 = fneg fast float %322
  store float 0.000000e+00, float* %"'de47", align 4
  %324 = load float, float* %".unpack.unpack162'de", align 4
  %325 = fadd fast float %324, %322
  store float %325, float* %".unpack.unpack162'de", align 4
  %326 = load float, float* %".unpack.unpack155'de", align 4
  %327 = fadd fast float %326, %323
  store float %327, float* %".unpack.unpack155'de", align 4
  br label %invertL119.i

invertL149.i:                                     ; preds = %invertL169.i, %invertL159.i
  %328 = load float, float* %"value_phi3.i'de", align 4
  store float 0.000000e+00, float* %"value_phi3.i'de", align 4
  %329 = load float, float* %.unpack.unpack155_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !917
  %330 = load float, float* %.unpack.unpack162_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !918
  %_unwrap50 = fcmp uge float %329, %330
  %_unwrap51 = fsub float %330, %329
  %_unwrap52 = fsub float %329, %330
  %_unwrap53 = fadd float %.fca.0.0.1.extract, %_unwrap52
  %_unwrap54 = fcmp uge float %_unwrap51, %_unwrap53
  %anot1_55 = xor i1 %_unwrap50, true
  %andVal056 = and i1 %_unwrap54, %anot1_55
  %bnot1_57 = xor i1 %_unwrap54, true
  %andVal158 = and i1 %bnot1_57, %anot1_55
  %331 = select fast i1 %_unwrap50, float %328, float 0.000000e+00
  %332 = load float, float* %".185'de", align 4
  %333 = fadd fast float %332, %328
  %334 = select fast i1 %_unwrap50, float %333, float %332
  store float %334, float* %".185'de", align 4
  %335 = select fast i1 %andVal056, float %328, float 0.000000e+00
  %336 = load float, float* %"'de39", align 4
  %337 = fadd fast float %336, %328
  %338 = select fast i1 %andVal056, float %337, float %336
  store float %338, float* %"'de39", align 4
  %339 = select fast i1 %andVal158, float %328, float 0.000000e+00
  %340 = load float, float* %"'de38", align 4
  %341 = fadd fast float %340, %328
  %342 = select fast i1 %andVal158, float %341, float %340
  store float %342, float* %"'de38", align 4
  br i1 %_unwrap50, label %invertL139.i, label %staging59

invertL159.i:                                     ; preds = %staging123, %invertL166.i
  %343 = load float, float* %"'de60", align 4
  %344 = fneg fast float %343
  store float 0.000000e+00, float* %"'de60", align 4
  %345 = load float, float* %".unpack.unpack164'de", align 4
  %346 = fadd fast float %345, %343
  store float %346, float* %".unpack.unpack164'de", align 4
  %347 = load float, float* %".unpack.unpack157'de", align 4
  %348 = fadd fast float %347, %344
  store float %348, float* %".unpack.unpack157'de", align 4
  br label %invertL149.i

invertL166.i:                                     ; preds = %staging123
  %349 = load float, float* %"'de61", align 4
  store float 0.000000e+00, float* %"'de61", align 4
  %350 = load float, float* %"'de60", align 4
  %351 = fadd fast float %350, %349
  store float %351, float* %"'de60", align 4
  br label %invertL159.i

invertL169.i:                                     ; preds = %invertL179.i_phimerge_phimerge_phimerge
  %352 = load float, float* %".186'de", align 4
  %353 = load float, float* %.unpack.unpack157_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !916
  %354 = load float, float* %.unpack.unpack164_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !923
  %_unwrap62 = fsub float %353, %354
  %_unwrap63 = fsub float %354, %353
  %_unwrap64 = fadd float %.fca.0.0.2.extract, %_unwrap63
  %_unwrap65 = fcmp uge float %_unwrap62, %_unwrap64
  %diffe66 = select fast i1 %_unwrap65, float %352, float 0.000000e+00
  %355 = load float, float* %".186'de", align 4
  %diffe67 = select fast i1 %_unwrap65, float 0.000000e+00, float %355
  store float 0.000000e+00, float* %".186'de", align 4
  %356 = load float, float* %"'de68", align 4
  %357 = fadd fast float %356, %352
  %358 = select fast i1 %_unwrap65, float %357, float %356
  store float %358, float* %"'de68", align 4
  %359 = load float, float* %"'de69", align 4
  %360 = fadd fast float %359, %355
  %361 = select fast i1 %_unwrap65, float %359, float %360
  store float %361, float* %"'de69", align 4
  %362 = load float, float* %"'de68", align 4
  store float 0.000000e+00, float* %"'de68", align 4
  %363 = load float, float* %"'de69", align 4
  %364 = fadd fast float %363, %362
  store float %364, float* %"'de69", align 4
  %365 = load float, float* %"'de69", align 4
  %366 = fneg fast float %365
  store float 0.000000e+00, float* %"'de69", align 4
  %367 = load float, float* %".unpack.unpack164'de", align 4
  %368 = fadd fast float %367, %365
  store float %368, float* %".unpack.unpack164'de", align 4
  %369 = load float, float* %".unpack.unpack157'de", align 4
  %370 = fadd fast float %369, %366
  store float %370, float* %".unpack.unpack157'de", align 4
  br label %invertL149.i

invertL179.i:                                     ; preds = %invertL190.i
  %371 = load float, float* %"'de70", align 4
  store float 0.000000e+00, float* %"'de70", align 4
  %372 = load float, float* %.unpack.unpack_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !914
  %373 = load float, float* %.unpack.unpack160_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !915
  %_unwrap71 = fcmp uge float %372, %373
  %_unwrap72 = fsub float %373, %372
  %_unwrap73 = fsub float %372, %373
  %_unwrap74 = fadd float %.fca.0.0.0.extract, %_unwrap73
  %_unwrap75 = fcmp uge float %_unwrap72, %_unwrap74
  br i1 %_unwrap71, label %invertL179.i_phirc78, label %invertL179.i_phisplt

invertL179.i_phisplt:                             ; preds = %invertL179.i
  br i1 %_unwrap75, label %invertL179.i_phirc, label %invertL179.i_phirc77

invertL179.i_phirc:                               ; preds = %invertL179.i_phisplt
  %_unwrap76 = fsub float %_unwrap72, %.fca.0.0.0.extract
  br label %invertL179.i_phimerge

invertL179.i_phirc77:                             ; preds = %invertL179.i_phisplt
  br label %invertL179.i_phimerge

invertL179.i_phirc78:                             ; preds = %invertL179.i
  %_unwrap79 = fsub float %372, %373
  %_unwrap80 = fsub float %373, %372
  %_unwrap81 = fadd float %.fca.0.0.0.extract, %_unwrap80
  %_unwrap82 = fcmp uge float %_unwrap79, %_unwrap81
  %._unwrap = select i1 %_unwrap82, float %_unwrap81, float %_unwrap80
  br label %invertL179.i_phimerge

invertL179.i_phimerge:                            ; preds = %invertL179.i_phirc78, %invertL179.i_phirc77, %invertL179.i_phirc
  %374 = phi fast float [ %_unwrap76, %invertL179.i_phirc ], [ %_unwrap72, %invertL179.i_phirc77 ], [ %._unwrap, %invertL179.i_phirc78 ]
  %_unwrap83 = fmul float %374, %374
  %375 = load float, float* %.unpack.unpack155_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !917
  %376 = load float, float* %.unpack.unpack162_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !918
  %_unwrap84 = fcmp uge float %375, %376
  %_unwrap85 = fsub float %376, %375
  %_unwrap86 = fsub float %375, %376
  %_unwrap87 = fadd float %.fca.0.0.1.extract, %_unwrap86
  %_unwrap88 = fcmp uge float %_unwrap85, %_unwrap87
  br i1 %_unwrap84, label %invertL179.i_phimerge_phirc91, label %invertL179.i_phimerge_phisplt

invertL179.i_phimerge_phisplt:                    ; preds = %invertL179.i_phimerge
  br i1 %_unwrap88, label %invertL179.i_phimerge_phirc, label %invertL179.i_phimerge_phirc90

invertL179.i_phimerge_phirc:                      ; preds = %invertL179.i_phimerge_phisplt
  %_unwrap89 = fsub float %_unwrap85, %.fca.0.0.1.extract
  br label %invertL179.i_phimerge_phimerge

invertL179.i_phimerge_phirc90:                    ; preds = %invertL179.i_phimerge_phisplt
  br label %invertL179.i_phimerge_phimerge

invertL179.i_phimerge_phirc91:                    ; preds = %invertL179.i_phimerge
  %_unwrap92 = fsub float %375, %376
  %_unwrap93 = fsub float %376, %375
  %_unwrap94 = fadd float %.fca.0.0.1.extract, %_unwrap93
  %_unwrap95 = fcmp uge float %_unwrap92, %_unwrap94
  %.185_unwrap = select i1 %_unwrap95, float %_unwrap94, float %_unwrap93
  br label %invertL179.i_phimerge_phimerge

invertL179.i_phimerge_phimerge:                   ; preds = %invertL179.i_phimerge_phirc91, %invertL179.i_phimerge_phirc90, %invertL179.i_phimerge_phirc
  %377 = phi fast float [ %_unwrap89, %invertL179.i_phimerge_phirc ], [ %_unwrap85, %invertL179.i_phimerge_phirc90 ], [ %.185_unwrap, %invertL179.i_phimerge_phirc91 ]
  %_unwrap96 = fmul float %377, %377
  %_unwrap97 = fadd float %_unwrap83, %_unwrap96
  %378 = load float, float* %.unpack.unpack157_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !916
  %379 = load float, float* %.unpack.unpack164_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !923
  %_unwrap98 = fcmp uge float %378, %379
  %_unwrap99 = fsub float %379, %378
  %_unwrap100 = fsub float %378, %379
  %_unwrap101 = fadd float %.fca.0.0.2.extract, %_unwrap100
  %_unwrap102 = fcmp uge float %_unwrap99, %_unwrap101
  br i1 %_unwrap98, label %invertL179.i_phimerge_phimerge_phirc105, label %invertL179.i_phimerge_phimerge_phisplt

invertL179.i_phimerge_phimerge_phisplt:           ; preds = %invertL179.i_phimerge_phimerge
  br i1 %_unwrap102, label %invertL179.i_phimerge_phimerge_phirc, label %invertL179.i_phimerge_phimerge_phirc104

invertL179.i_phimerge_phimerge_phirc:             ; preds = %invertL179.i_phimerge_phimerge_phisplt
  %_unwrap103 = fsub float %_unwrap99, %.fca.0.0.2.extract
  br label %invertL179.i_phimerge_phimerge_phimerge

invertL179.i_phimerge_phimerge_phirc104:          ; preds = %invertL179.i_phimerge_phimerge_phisplt
  br label %invertL179.i_phimerge_phimerge_phimerge

invertL179.i_phimerge_phimerge_phirc105:          ; preds = %invertL179.i_phimerge_phimerge
  %_unwrap106 = fsub float %378, %379
  %_unwrap107 = fsub float %379, %378
  %_unwrap108 = fadd float %.fca.0.0.2.extract, %_unwrap107
  %_unwrap109 = fcmp uge float %_unwrap106, %_unwrap108
  %.186_unwrap = select i1 %_unwrap109, float %_unwrap108, float %_unwrap107
  br label %invertL179.i_phimerge_phimerge_phimerge

invertL179.i_phimerge_phimerge_phimerge:          ; preds = %invertL179.i_phimerge_phimerge_phirc105, %invertL179.i_phimerge_phimerge_phirc104, %invertL179.i_phimerge_phimerge_phirc
  %380 = phi fast float [ %_unwrap103, %invertL179.i_phimerge_phimerge_phirc ], [ %_unwrap99, %invertL179.i_phimerge_phimerge_phirc104 ], [ %.186_unwrap, %invertL179.i_phimerge_phimerge_phirc105 ]
  %_unwrap110 = fmul float %380, %380
  %_unwrap111 = fadd float %_unwrap97, %_unwrap110
  %381 = call fast float @__nv_sqrtf(float %_unwrap111), !dbg !943
  %382 = fmul fast float 5.000000e-01, %371
  %383 = fdiv fast float %382, %381
  %384 = fcmp fast oeq float %_unwrap111, 0.000000e+00
  %385 = select fast i1 %384, float 0.000000e+00, float %383
  %386 = load float, float* %"'de112", align 4
  %387 = fadd fast float %386, %383
  %388 = select fast i1 %384, float %386, float %387
  store float %388, float* %"'de112", align 4
  %389 = load float, float* %"'de112", align 4
  store float 0.000000e+00, float* %"'de112", align 4
  %390 = load float, float* %"'de113", align 4
  %391 = fadd fast float %390, %389
  store float %391, float* %"'de113", align 4
  %392 = load float, float* %"'de114", align 4
  %393 = fadd fast float %392, %389
  store float %393, float* %"'de114", align 4
  %394 = load float, float* %"'de114", align 4
  %m0diffevalue_phi4.i = fmul fast float %394, %380
  %m1diffevalue_phi4.i = fmul fast float %394, %380
  store float 0.000000e+00, float* %"'de114", align 4
  %395 = load float, float* %"value_phi4.i'de", align 4
  %396 = fadd fast float %395, %m0diffevalue_phi4.i
  store float %396, float* %"value_phi4.i'de", align 4
  %397 = load float, float* %"value_phi4.i'de", align 4
  %398 = fadd fast float %397, %m1diffevalue_phi4.i
  store float %398, float* %"value_phi4.i'de", align 4
  %399 = load float, float* %"'de113", align 4
  store float 0.000000e+00, float* %"'de113", align 4
  %400 = load float, float* %"'de115", align 4
  %401 = fadd fast float %400, %399
  store float %401, float* %"'de115", align 4
  %402 = load float, float* %"'de116", align 4
  %403 = fadd fast float %402, %399
  store float %403, float* %"'de116", align 4
  %404 = load float, float* %"'de116", align 4
  %m0diffevalue_phi3.i = fmul fast float %404, %377
  %m1diffevalue_phi3.i = fmul fast float %404, %377
  store float 0.000000e+00, float* %"'de116", align 4
  %405 = load float, float* %"value_phi3.i'de", align 4
  %406 = fadd fast float %405, %m0diffevalue_phi3.i
  store float %406, float* %"value_phi3.i'de", align 4
  %407 = load float, float* %"value_phi3.i'de", align 4
  %408 = fadd fast float %407, %m1diffevalue_phi3.i
  store float %408, float* %"value_phi3.i'de", align 4
  %409 = load float, float* %"'de115", align 4
  %m0diffevalue_phi.i = fmul fast float %409, %374
  %m1diffevalue_phi.i = fmul fast float %409, %374
  store float 0.000000e+00, float* %"'de115", align 4
  %410 = load float, float* %"value_phi.i'de", align 4
  %411 = fadd fast float %410, %m0diffevalue_phi.i
  store float %411, float* %"value_phi.i'de", align 4
  %412 = load float, float* %"value_phi.i'de", align 4
  %413 = fadd fast float %412, %m1diffevalue_phi.i
  store float %413, float* %"value_phi.i'de", align 4
  %414 = load float, float* %"value_phi4.i'de", align 4
  store float 0.000000e+00, float* %"value_phi4.i'de", align 4
  %anot1_119 = xor i1 %_unwrap98, true
  %andVal0120 = and i1 %_unwrap102, %anot1_119
  %bnot1_121 = xor i1 %_unwrap102, true
  %andVal1122 = and i1 %bnot1_121, %anot1_119
  %415 = select fast i1 %_unwrap98, float %414, float 0.000000e+00
  %416 = load float, float* %".186'de", align 4
  %417 = fadd fast float %416, %414
  %418 = select fast i1 %_unwrap98, float %417, float %416
  store float %418, float* %".186'de", align 4
  %419 = select fast i1 %andVal0120, float %414, float 0.000000e+00
  %420 = load float, float* %"'de61", align 4
  %421 = fadd fast float %420, %414
  %422 = select fast i1 %andVal0120, float %421, float %420
  store float %422, float* %"'de61", align 4
  %423 = select fast i1 %andVal1122, float %414, float 0.000000e+00
  %424 = load float, float* %"'de60", align 4
  %425 = fadd fast float %424, %414
  %426 = select fast i1 %andVal1122, float %425, float %424
  store float %426, float* %"'de60", align 4
  br i1 %_unwrap98, label %invertL169.i, label %staging123

invertL190.i:                                     ; No predecessors!
  br label %invertL179.i

invertL197.i:                                     ; No predecessors!
  %427 = load float, float* %"'de124", align 4
  %428 = load i1, i1* %.not170_cache, align 1, !invariant.group !954
  %429 = load float, float* %.unpack.unpack_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !914
  %430 = load float, float* %.unpack.unpack160_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !915
  %_unwrap125 = fcmp uge float %429, %430
  %_unwrap126 = fsub float %430, %429
  %_unwrap127 = fsub float %429, %430
  %_unwrap128 = fadd float %.fca.0.0.0.extract, %_unwrap127
  %_unwrap129 = fcmp uge float %_unwrap126, %_unwrap128
  br i1 %_unwrap125, label %invertL197.i_phirc132, label %invertL197.i_phisplt

invertL197.i_phisplt:                             ; preds = %invertL197.i
  br i1 %_unwrap129, label %invertL197.i_phirc, label %invertL197.i_phirc131

invertL197.i_phirc:                               ; preds = %invertL197.i_phisplt
  %_unwrap130 = fsub float %_unwrap126, %.fca.0.0.0.extract
  br label %invertL197.i_phimerge

invertL197.i_phirc131:                            ; preds = %invertL197.i_phisplt
  br label %invertL197.i_phimerge

invertL197.i_phirc132:                            ; preds = %invertL197.i
  %_unwrap133 = fsub float %429, %430
  %_unwrap134 = fsub float %430, %429
  %_unwrap135 = fadd float %.fca.0.0.0.extract, %_unwrap134
  %_unwrap136 = fcmp uge float %_unwrap133, %_unwrap135
  %._unwrap137 = select i1 %_unwrap136, float %_unwrap135, float %_unwrap134
  br label %invertL197.i_phimerge

invertL197.i_phimerge:                            ; preds = %invertL197.i_phirc132, %invertL197.i_phirc131, %invertL197.i_phirc
  %431 = phi fast float [ %_unwrap130, %invertL197.i_phirc ], [ %_unwrap126, %invertL197.i_phirc131 ], [ %._unwrap137, %invertL197.i_phirc132 ]
  %432 = call float @llvm.fabs.f32(float %431) #146, !dbg !948
  %433 = call float @llvm.nvvm.fabs.ftz.f(float %431) #147, !dbg !948
  %.0.i35.i_unwrap = select i1 %428, float %432, float %433
  %434 = load i1, i1* %.not171_cache, align 1, !invariant.group !958
  %435 = load float, float* %.unpack.unpack155_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !917
  %436 = load float, float* %.unpack.unpack162_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !918
  %_unwrap138 = fcmp uge float %435, %436
  %_unwrap139 = fsub float %436, %435
  %_unwrap140 = fsub float %435, %436
  %_unwrap141 = fadd float %.fca.0.0.1.extract, %_unwrap140
  %_unwrap142 = fcmp uge float %_unwrap139, %_unwrap141
  br i1 %_unwrap138, label %invertL197.i_phimerge_phirc145, label %invertL197.i_phimerge_phisplt

invertL197.i_phimerge_phisplt:                    ; preds = %invertL197.i_phimerge
  br i1 %_unwrap142, label %invertL197.i_phimerge_phirc, label %invertL197.i_phimerge_phirc144

invertL197.i_phimerge_phirc:                      ; preds = %invertL197.i_phimerge_phisplt
  %_unwrap143 = fsub float %_unwrap139, %.fca.0.0.1.extract
  br label %invertL197.i_phimerge_phimerge

invertL197.i_phimerge_phirc144:                   ; preds = %invertL197.i_phimerge_phisplt
  br label %invertL197.i_phimerge_phimerge

invertL197.i_phimerge_phirc145:                   ; preds = %invertL197.i_phimerge
  %_unwrap146 = fsub float %435, %436
  %_unwrap147 = fsub float %436, %435
  %_unwrap148 = fadd float %.fca.0.0.1.extract, %_unwrap147
  %_unwrap149 = fcmp uge float %_unwrap146, %_unwrap148
  %.185_unwrap150 = select i1 %_unwrap149, float %_unwrap148, float %_unwrap147
  br label %invertL197.i_phimerge_phimerge

invertL197.i_phimerge_phimerge:                   ; preds = %invertL197.i_phimerge_phirc145, %invertL197.i_phimerge_phirc144, %invertL197.i_phimerge_phirc
  %437 = phi fast float [ %_unwrap143, %invertL197.i_phimerge_phirc ], [ %_unwrap139, %invertL197.i_phimerge_phirc144 ], [ %.185_unwrap150, %invertL197.i_phimerge_phirc145 ]
  %438 = call float @llvm.fabs.f32(float %437) #146, !dbg !955
  %439 = call float @llvm.nvvm.fabs.ftz.f(float %437) #147, !dbg !955
  %.0.i41.i_unwrap = select i1 %434, float %438, float %439
  %.inv_unwrap = fcmp olt float %.0.i35.i_unwrap, %.0.i41.i_unwrap
  %_unwrap151 = select i1 %.inv_unwrap, float %.0.i41.i_unwrap, float %.0.i35.i_unwrap
  %440 = load i1, i1* %.not171.1_cache, align 1, !invariant.group !961
  %441 = load float, float* %.unpack.unpack157_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !916
  %442 = load float, float* %.unpack.unpack164_cache, align 4, !dbg !908, !tbaa !216, !alias.scope !909, !noalias !912, !invariant.group !923
  %_unwrap152 = fcmp uge float %441, %442
  %_unwrap153 = fsub float %442, %441
  %_unwrap154 = fsub float %441, %442
  %_unwrap155 = fadd float %.fca.0.0.2.extract, %_unwrap154
  %_unwrap156 = fcmp uge float %_unwrap153, %_unwrap155
  br i1 %_unwrap152, label %invertL197.i_phimerge_phimerge_phirc159, label %invertL197.i_phimerge_phimerge_phisplt

invertL197.i_phimerge_phimerge_phisplt:           ; preds = %invertL197.i_phimerge_phimerge
  br i1 %_unwrap156, label %invertL197.i_phimerge_phimerge_phirc, label %invertL197.i_phimerge_phimerge_phirc158

invertL197.i_phimerge_phimerge_phirc:             ; preds = %invertL197.i_phimerge_phimerge_phisplt
  %_unwrap157 = fsub float %_unwrap153, %.fca.0.0.2.extract
  br label %invertL197.i_phimerge_phimerge_phimerge

invertL197.i_phimerge_phimerge_phirc158:          ; preds = %invertL197.i_phimerge_phimerge_phisplt
  br label %invertL197.i_phimerge_phimerge_phimerge

invertL197.i_phimerge_phimerge_phirc159:          ; preds = %invertL197.i_phimerge_phimerge
  %_unwrap160 = fsub float %441, %442
  %_unwrap161 = fsub float %442, %441
  %_unwrap162 = fadd float %.fca.0.0.2.extract, %_unwrap161
  %_unwrap163 = fcmp uge float %_unwrap160, %_unwrap162
  %.186_unwrap164 = select i1 %_unwrap163, float %_unwrap162, float %_unwrap161
  br label %invertL197.i_phimerge_phimerge_phimerge

invertL197.i_phimerge_phimerge_phimerge:          ; preds = %invertL197.i_phimerge_phimerge_phirc159, %invertL197.i_phimerge_phimerge_phirc158, %invertL197.i_phimerge_phimerge_phirc
  %443 = phi fast float [ %_unwrap157, %invertL197.i_phimerge_phimerge_phirc ], [ %_unwrap153, %invertL197.i_phimerge_phimerge_phirc158 ], [ %.186_unwrap164, %invertL197.i_phimerge_phimerge_phirc159 ]
  %444 = call float @llvm.fabs.f32(float %443) #146, !dbg !955
  %445 = call float @llvm.nvvm.fabs.ftz.f(float %443) #147, !dbg !955
  %.0.i41.i.1_unwrap = select i1 %440, float %444, float %445
  %.inv.1_unwrap = fcmp olt float %_unwrap151, %.0.i41.i.1_unwrap
  %diffe.0.i41.i.1 = select fast i1 %.inv.1_unwrap, float %427, float 0.000000e+00
  %446 = load float, float* %"'de124", align 4
  %diffe165 = select fast i1 %.inv.1_unwrap, float 0.000000e+00, float %446
  store float 0.000000e+00, float* %"'de124", align 4
  %447 = load float, float* %".0.i41.i.1'de", align 4
  %448 = fadd fast float %447, %427
  %449 = select fast i1 %.inv.1_unwrap, float %448, float %447
  store float %449, float* %".0.i41.i.1'de", align 4
  %450 = load float, float* %"'de166", align 4
  %451 = fadd fast float %450, %446
  %452 = select fast i1 %.inv.1_unwrap, float %450, float %451
  store float %452, float* %"'de166", align 4
  %453 = load float, float* %".0.i41.i.1'de", align 4
  %diffe167 = select fast i1 %440, float %453, float 0.000000e+00
  %454 = load float, float* %".0.i41.i.1'de", align 4
  %diffe168 = select fast i1 %440, float 0.000000e+00, float %454
  store float 0.000000e+00, float* %".0.i41.i.1'de", align 4
  %455 = load float, float* %"'de169", align 4
  %456 = fadd fast float %455, %453
  %457 = select fast i1 %440, float %456, float %455
  store float %457, float* %"'de169", align 4
  %458 = load float, float* %"'de170", align 4
  %459 = fadd fast float %458, %454
  %460 = select fast i1 %440, float %458, float %459
  store float %460, float* %"'de170", align 4
  %461 = load float, float* %"'de169", align 4
  store float 0.000000e+00, float* %"'de169", align 4
  %462 = fcmp fast olt float %443, 0.000000e+00
  %463 = select fast i1 %462, float -1.000000e+00, float 1.000000e+00
  %464 = fmul fast float %463, %461
  %465 = load float, float* %"value_phi4.i'de", align 4
  %466 = fadd fast float %465, %464
  store float %466, float* %"value_phi4.i'de", align 4
  %467 = load float, float* %"'de170", align 4
  store float 0.000000e+00, float* %"'de170", align 4
  %468 = fcmp fast olt float %443, 0.000000e+00
  %469 = select fast i1 %468, float -1.000000e+00, float 1.000000e+00
  %470 = fmul fast float %469, %467
  %471 = load float, float* %"value_phi4.i'de", align 4
  %472 = fadd fast float %471, %470
  store float %472, float* %"value_phi4.i'de", align 4

invertL221.i:                                     ; No predecessors!

invertL223.i:                                     ; No predecessors!

invertL236.i:                                     ; No predecessors!

invertL245.i:                                     ; No predecessors!

invertL302.i:                                     ; No predecessors!

invertL523.i.us:                                  ; No predecessors!

invertL615.i.us:                                  ; No predecessors!

invertL523.i.us.1:                                ; No predecessors!

invertL523.i.us.2:                                ; No predecessors!

invertL524.i:                                     ; No predecessors!

invertL616.i:                                     ; No predecessors!

invertfail.i:                                     ; No predecessors!

invertpass.i:                                     ; No predecessors!

invertoksrem.i:                                   ; No predecessors!

invertafter_srem.i:                               ; preds = %invertL50.i_phimerge

invertjulia_gbsa_force_2_kernel__25091_inner.exit: ; preds = %julia_gbsa_force_2_kernel__25091_inner.exit

staging:                                          ; preds = %invertL119.i
  br i1 %_unwrap37, label %invertL106.i, label %invertL99.i

staging59:                                        ; preds = %invertL149.i
  br i1 %_unwrap54, label %invertL136.i, label %invertL129.i

staging123:                                       ; preds = %invertL179.i_phimerge_phimerge_phimerge
  br i1 %_unwrap102, label %invertL166.i, label %invertL159.i
}

attributes #0 = { nofree readnone "enzyme_inactive" "enzyme_shouldrecompute" "enzymejl_world"="33717" }
attributes #1 = { "enzymejl_world"="33717" }
attributes #2 = { argmemonly nofree nounwind willreturn "enzymejl_world"="33717" }
attributes #3 = { nounwind readnone speculatable "enzymejl_world"="33717" }
attributes #4 = { argmemonly nounwind readonly "enzymejl_world"="33717" }
attributes #5 = { noinline noreturn "enzymejl_world"="33717" "frame-pointer"="all" "probe-stack"="inline-asm" }
attributes #6 = { nofree nosync nounwind readnone speculatable willreturn "enzymejl_world"="33717" }
attributes #7 = { "disable-tail-calls"="false" "enzymejl_world"="33717" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #8 = { nofree nosync nounwind readnone willreturn "enzymejl_world"="33717" }
attributes #9 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="exp2f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.exp2.f32" "implements2"="exp2f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #10 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="sqrtf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.sqrt.f32" "implements2"="sqrtf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #11 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="sqrt" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.sqrt.f64" "implements2"="sqrt" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #12 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="isinff" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.isinf.f32" "implements2"="isinff" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #13 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="sinf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.sin.f32" "implements2"="sinf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #14 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="cosf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.cos.f32" "implements2"="cosf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #15 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="tanf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.tan.f32" "implements2"="tanf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #16 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="log2f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log2.f32" "implements2"="log2f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #17 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="expf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.exp.f32" "implements2"="expf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #18 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="exp10f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.exp10.f32" "implements2"="exp10f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #19 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="coshf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.cosh.f32" "implements2"="coshf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #20 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="sinhf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.sinh.f32" "implements2"="sinhf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #21 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="tanhf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.tanh.f32" "implements2"="tanhf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #22 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="atan2f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.atan2.f32" "implements2"="atan2f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #23 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="atanf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.atan.f32" "implements2"="atanf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #24 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="asinf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.asin.f32" "implements2"="asinf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #25 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="acosf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.acos.f32" "implements2"="acosf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #26 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="logf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log.f32" "implements2"="logf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #27 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="log10f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log10.f32" "implements2"="log10f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #28 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="log1pf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log1p.f32" "implements2"="log1pf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #29 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="acoshf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.acosh.f32" "implements2"="acoshf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #30 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="asinhf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.asinh.f32" "implements2"="asinhf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #31 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="atanhf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.atanh.f32" "implements2"="atanhf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #32 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="expm1f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.expm1.f32" "implements2"="expm1f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #33 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="hypotf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.hypot.f32" "implements2"="hypotf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #34 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="rhypotf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rhypot.f32" "implements2"="rhypotf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #35 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="norm3df" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.norm3d.f32" "implements2"="norm3df" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #36 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="rnorm3df" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rnorm3d.f32" "implements2"="rnorm3df" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #37 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="norm4df" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.norm4d.f32" "implements2"="norm4df" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #38 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="rnorm4df" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rnorm4d.f32" "implements2"="rnorm4df" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #39 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="normf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.norm.f32" "implements2"="normf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #40 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="rnormf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rnorm.f32" "implements2"="rnormf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #41 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="cbrtf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.cbrt.f32" "implements2"="cbrtf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #42 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="rcbrtf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rcbrt.f32" "implements2"="rcbrtf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #43 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="j0f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.j0.f32" "implements2"="j0f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #44 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="j1f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.j1.f32" "implements2"="j1f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #45 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="y0f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.y0.f32" "implements2"="y0f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #46 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="y1f" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.y1.f32" "implements2"="y1f" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #47 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="ynf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.yn.f32" "implements2"="ynf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #48 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="jnf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.jn.f32" "implements2"="jnf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #49 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="erff" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erf.f32" "implements2"="erff" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #50 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="erfinvf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfinv.f32" "implements2"="erfinvf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #51 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="erfcf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfc.f32" "implements2"="erfcf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #52 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="erfcxf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfcx.f32" "implements2"="erfcxf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #53 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="erfcinvf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfcinv.f32" "implements2"="erfcinvf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #54 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="normcdfinvf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.normcdfinv.f32" "implements2"="normcdfinvf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #55 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="normcdff" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.normcdf.f32" "implements2"="normcdff" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #56 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="lgammaf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.lgamma.f32" "implements2"="lgammaf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #57 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="ldexpf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.ldexp.f32" "implements2"="ldexpf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #58 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="scalbnf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.scalbn.f32" "implements2"="scalbnf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #59 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="frexpf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.frexp.f32" "implements2"="frexpf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #60 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="modff" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.modf.f32" "implements2"="modff" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #61 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="fmodf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.fmod.f32" "implements2"="fmodf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #62 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="remainderf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.remainder.f32" "implements2"="remainderf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #63 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="remquof" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.remquo.f32" "implements2"="remquof" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #64 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="powif" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.powi.f32" "implements2"="powif" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #65 = { nofree noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="powi" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.powi.f64" "implements2"="powi" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #66 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="powf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.pow.f32" "implements2"="powf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #67 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="tgammaf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.tgamma.f32" "implements2"="tgammaf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #68 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="roundf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.round.f32" "implements2"="roundf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #69 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="fdimf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.fdim.f32" "implements2"="fdimf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #70 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="ilogbf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.ilogb.f32" "implements2"="ilogbf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #71 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="logbf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.logb.f32" "implements2"="logbf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #72 = { argmemonly noinline nosync nounwind "disable-tail-calls"="false" "enzymejl_world"="33717" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #73 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="sin" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.sin.f64" "implements2"="sin" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #74 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="cos" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.cos.f64" "implements2"="cos" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #75 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="tan" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.tan.f64" "implements2"="tan" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #76 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="log" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log.f64" "implements2"="log" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #77 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="log2" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log2.f64" "implements2"="log2" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #78 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="log10" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log10.f64" "implements2"="log10" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #79 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="log1p" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.log1p.f64" "implements2"="log1p" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #80 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="exp" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.exp.f64" "implements2"="exp" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #81 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="exp2" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.exp2.f64" "implements2"="exp2" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #82 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="exp10" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.exp10.f64" "implements2"="exp10" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #83 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="expm1" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.expm1.f64" "implements2"="expm1" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #84 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="cosh" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.cosh.f64" "implements2"="cosh" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #85 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="sinh" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.sinh.f64" "implements2"="sinh" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #86 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="tanh" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.tanh.f64" "implements2"="tanh" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #87 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="atan2" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.atan2.f64" "implements2"="atan2" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #88 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="atan" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.atan.f64" "implements2"="atan" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #89 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="asin" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.asin.f64" "implements2"="asin" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #90 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="acos" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.acos.f64" "implements2"="acos" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #91 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="acosh" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.acosh.f64" "implements2"="acosh" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #92 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="asinh" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.asinh.f64" "implements2"="asinh" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #93 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="atanh" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.atanh.f64" "implements2"="atanh" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #94 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="hypot" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.hypot.f64" "implements2"="hypot" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #95 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="rhypot" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rhypot.f64" "implements2"="rhypot" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #96 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="norm3d" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.norm3d.f64" "implements2"="norm3d" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #97 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="rnorm3d" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rnorm3d.f64" "implements2"="rnorm3d" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #98 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="norm4d" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.norm4d.f64" "implements2"="norm4d" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #99 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="rnorm4d" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rnorm4d.f64" "implements2"="rnorm4d" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #100 = { argmemonly noinline nosync nounwind readonly "disable-tail-calls"="false" "enzyme_math"="norm" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.norm.f64" "implements2"="norm" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #101 = { argmemonly nofree noinline nosync nounwind readonly "disable-tail-calls"="false" "enzyme_math"="rnorm" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rnorm.f64" "implements2"="rnorm" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #102 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="cbrt" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.cbrt.f64" "implements2"="cbrt" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #103 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="rcbrt" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.rcbrt.f64" "implements2"="rcbrt" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #104 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="pow" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.pow.f64" "implements2"="pow" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #105 = { noinline nounwind "disable-tail-calls"="false" "enzymejl_world"="33717" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #106 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="j0" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.j0.f64" "implements2"="j0" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #107 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="j1" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.j1.f64" "implements2"="j1" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #108 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="y0" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.y0.f64" "implements2"="y0" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #109 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="y1" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.y1.f64" "implements2"="y1" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #110 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="yn" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.yn.f64" "implements2"="yn" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #111 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="jn" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.jn.f64" "implements2"="jn" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #112 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="erf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erf.f64" "implements2"="erf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #113 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="erfinv" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfinv.f64" "implements2"="erfinv" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #114 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="erfcinv" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfcinv.f64" "implements2"="erfcinv" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #115 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="normcdfinv" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.normcdfinv.f64" "implements2"="normcdfinv" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #116 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="erfc" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfc.f64" "implements2"="erfc" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #117 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="erfcx" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.erfcx.f64" "implements2"="erfcx" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #118 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="normcdf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.normcdf.f64" "implements2"="normcdf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #119 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="tgamma" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.tgamma.f64" "implements2"="tgamma" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #120 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="lgamma" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.lgamma.f64" "implements2"="lgamma" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #121 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzymejl_world"="33717" "frame-pointer"="all" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #122 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="ldexp" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.ldexp.f64" "implements2"="ldexp" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #123 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="scalbn" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.scalbn.f64" "implements2"="scalbn" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #124 = { argmemonly nofree noinline nosync nounwind willreturn writeonly "disable-tail-calls"="false" "enzyme_math"="frexp" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.frexp.f64" "implements2"="frexp" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #125 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="modf" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.modf.f64" "implements2"="modf" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #126 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="fmod" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.fmod.f64" "implements2"="fmod" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #127 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="remainder" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.remainder.f64" "implements2"="remainder" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #128 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="remquo" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.remquo.f64" "implements2"="remquo" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #129 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="round" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.round.f64" "implements2"="round" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #130 = { noinline nounwind "disable-tail-calls"="false" "enzyme_math"="ilogb" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.ilogb.f64" "implements2"="ilogb" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #131 = { nofree noinline nosync nounwind readnone willreturn "disable-tail-calls"="false" "enzyme_math"="logb" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.logb.f64" "implements2"="logb" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #132 = { noinline nosync nounwind readnone "disable-tail-calls"="false" "enzyme_math"="fdim" "enzymejl_world"="33717" "frame-pointer"="all" "implements"="llvm.fdim.f64" "implements2"="fdim" "less-precise-fpmad"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
attributes #133 = { nofree "enzyme_inactive" "enzymejl_world"="33717" "probe-stack"="inline-asm" }
attributes #134 = { readnone "enzymejl_world"="33717" }
attributes #135 = { nounwind "enzymejl_world"="33717" }
attributes #136 = { argmemonly nofree nosync nounwind willreturn "enzymejl_world"="33717" }
attributes #137 = { "enzymejl_world"="33717" "probe-stack"="inline-asm" }
attributes #138 = { mustprogress willreturn "enzymejl_world"="33717" "probe-stack"="inline-asm" }
attributes #139 = { nounwind }
attributes #140 = { readnone willreturn }
attributes #141 = { nounwind readnone }
attributes #142 = { nounwind readnone willreturn }
attributes #143 = { nofree nosync nounwind readnone willreturn }
attributes #144 = { willreturn }
attributes #145 = { noreturn "frame-pointer"="all" "probe-stack"="inline-asm" }
attributes #146 = { mustprogress willreturn }
attributes #147 = { mustprogress nounwind willreturn }
attributes #148 = { mustprogress noreturn willreturn "frame-pointer"="all" "probe-stack"="inline-asm" }

!llvm.module.flags = !{!0, !1}
!llvm.dbg.cu = !{!2, !4, !6, !7, !8, !9, !10, !11, !12, !15, !16, !17, !19, !20, !21, !22, !23, !24, !25, !26, !27, !28, !29, !30, !31, !32, !33, !34, !35, !36, !37, !38, !39, !40, !41, !42, !44}
!llvm.ident = !{!45}
!nvvmir.version = !{!46}
!nvvm-reflect-ftz = !{!47}

!0 = !{i32 2, !"Dwarf Version", i32 4}
!1 = !{i32 2, !"Debug Info Version", i32 3}
!2 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!3 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/quirks.jl", directory: ".")
!4 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !5, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!5 = !DIFile(filename: "/home/jgreener/.julia/dev/Molly/src/interactions/implicit_solvent.jl", directory: ".")
!6 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!7 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!8 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!9 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!10 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!11 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !3, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, nameTableKind: None)
!12 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!13 = !DIFile(filename: "/home/jgreener/.julia/packages/GPUCompiler/YU9ZW/src/runtime.jl", directory: ".")
!14 = !{}
!15 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!16 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!17 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !18, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!18 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/runtime.jl", directory: ".")
!19 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!20 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!21 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !18, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!22 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!23 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!24 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !18, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!25 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!26 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !18, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!27 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!28 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!29 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!30 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!31 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!32 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!33 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!34 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!35 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!36 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!37 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!38 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!39 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!40 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!41 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !13, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!42 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !43, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!43 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/intrinsics/memory_dynamic.jl", directory: ".")
!44 = distinct !DICompileUnit(language: DW_LANG_Julia, file: !18, producer: "julia", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !14, nameTableKind: None)
!45 = !{!"clang version 3.8.0 (tags/RELEASE_380/final)"}
!46 = !{i32 2, i32 0}
!47 = !{i32 1}
!48 = distinct !DISubprogram(name: "#throw_boundserror", linkageName: "julia_#throw_boundserror_25134", scope: null, file: !3, line: 40, type: !49, scopeLine: 40, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !14)
!49 = !DISubroutineType(types: !14)
!50 = !DILocation(line: 40, scope: !51, inlinedAt: !53)
!51 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !52, file: !52, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !14)
!52 = !DIFile(filename: "/home/jgreener/.julia/packages/LLVM/TLGyi/src/interop/base.jl", directory: ".")
!53 = !DILocation(line: 38, scope: !54, inlinedAt: !56)
!54 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !55, file: !55, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !14)
!55 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/intrinsics/output.jl", directory: ".")
!56 = !DILocation(line: 38, scope: !57, inlinedAt: !58)
!57 = distinct !DISubprogram(name: "_cuprintf;", linkageName: "_cuprintf", scope: !55, file: !55, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !14)
!58 = !DILocation(line: 173, scope: !54, inlinedAt: !59)
!59 = !DILocation(line: 0, scope: !60, inlinedAt: !62)
!60 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !61, file: !61, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !14)
!61 = !DIFile(filename: "none", directory: ".")
!62 = !DILocation(line: 0, scope: !63, inlinedAt: !64)
!63 = distinct !DISubprogram(name: "_cuprint;", linkageName: "_cuprint", scope: !61, file: !61, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !7, retainedNodes: !14)
!64 = !DILocation(line: 222, scope: !54, inlinedAt: !65)
!65 = !DILocation(line: 3, scope: !48)
!66 = !DILocation(line: 4, scope: !48)
!67 = !{i32 32878, i32 32882, i32 32927, i32 32972}
!68 = !{!69}
!69 = distinct !{!69, !70, !"na_addr13"}
!70 = distinct !{!70, !"addr13"}
!71 = distinct !{!71, !72}
!72 = !{!"llvm.loop.unroll.count", i32 1}
!73 = !{i32 21046}
!74 = !{i32 9032}
!75 = !{i32 9109}
!76 = !{i32 0, i32 33}
!77 = !{i32 156190, i32 156194, i32 156263, i32 156311, i32 156359, i32 156407, i32 156455, i32 156503, i32 156551, i32 156599, i32 156647, i32 156695, i32 156743, i32 156791, i32 156839, i32 156887, i32 156935}
!78 = distinct !{!78, !72}
!79 = !{i32 153924, i32 153928, i32 153999, i32 154041, i32 154083, i32 154125, i32 154167, i32 154209, i32 154251, i32 154293, i32 154335, i32 154377, i32 154419}
!80 = !{i64 0, i64 65}
!81 = !{i32 154965, i32 154969, i32 155028, i32 155075, i32 155122, i32 155169, i32 155216, i32 155263, i32 155310, i32 155357, i32 155404, i32 155451, i32 155498, i32 155545, i32 155592, i32 155639}
!82 = !{i32 152726, i32 152730, i32 152801, i32 152843, i32 152885, i32 152927, i32 152969, i32 153011, i32 153053, i32 153095, i32 153137, i32 153179, i32 153221}
!83 = !{i32 150120}
!84 = !{i32 149741}
!85 = distinct !{!85, !86}
!86 = !{!"llvm.loop.unroll.disable"}
!87 = distinct !{!87, !86}
!88 = distinct !{!88, !86}
!89 = distinct !{!89, !86}
!90 = !{i32 285845}
!91 = distinct !DISubprogram(name: "report_exception", linkageName: "julia_report_exception_21845", scope: null, file: !18, line: 51, type: !49, scopeLine: 51, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !17, retainedNodes: !14)
!92 = !DILocation(line: 40, scope: !93, inlinedAt: !94)
!93 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !52, file: !52, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !17, retainedNodes: !14)
!94 = !DILocation(line: 38, scope: !95, inlinedAt: !96)
!95 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !55, file: !55, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !17, retainedNodes: !14)
!96 = !DILocation(line: 38, scope: !97, inlinedAt: !98)
!97 = distinct !DISubprogram(name: "_cuprintf;", linkageName: "_cuprintf", scope: !55, file: !55, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !17, retainedNodes: !14)
!98 = !DILocation(line: 52, scope: !91)
!99 = !DILocation(line: 56, scope: !91)
!100 = distinct !DISubprogram(name: "signal_exception", linkageName: "julia_signal_exception_21886", scope: null, file: !18, line: 37, type: !49, scopeLine: 37, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!101 = !DILocation(line: 40, scope: !102, inlinedAt: !103)
!102 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !52, file: !52, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!103 = !DILocation(line: 0, scope: !104, inlinedAt: !105)
!104 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !61, file: !61, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!105 = !DILocation(line: 0, scope: !106, inlinedAt: !107)
!106 = distinct !DISubprogram(name: "kernel_state;", linkageName: "kernel_state", scope: !61, file: !61, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!107 = !DILocation(line: 35, scope: !108, inlinedAt: !109)
!108 = distinct !DISubprogram(name: "exception_flag;", linkageName: "exception_flag", scope: !18, file: !18, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!109 = !DILocation(line: 38, scope: !100)
!110 = !DILocation(line: 39, scope: !100)
!111 = !DILocation(line: 118, scope: !112, inlinedAt: !114)
!112 = distinct !DISubprogram(name: "unsafe_store!;", linkageName: "unsafe_store!", scope: !113, file: !113, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!113 = !DIFile(filename: "pointer.jl", directory: ".")
!114 = !DILocation(line: 118, scope: !112, inlinedAt: !115)
!115 = !DILocation(line: 40, scope: !100)
!116 = !{!117, !117, i64 0}
!117 = !{!"jtbaa_data", !118, i64 0}
!118 = !{!"jtbaa", !119, i64 0}
!119 = !{!"jtbaa"}
!120 = !DILocation(line: 121, scope: !121, inlinedAt: !123)
!121 = distinct !DISubprogram(name: "threadfence_system;", linkageName: "threadfence_system", scope: !122, file: !122, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!122 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/intrinsics/synchronization.jl", directory: ".")
!123 = !DILocation(line: 41, scope: !100)
!124 = !DILocation(line: 40, scope: !102, inlinedAt: !125)
!125 = !DILocation(line: 38, scope: !126, inlinedAt: !127)
!126 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !55, file: !55, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!127 = !DILocation(line: 38, scope: !128, inlinedAt: !129)
!128 = distinct !DISubprogram(name: "_cuprintf;", linkageName: "_cuprintf", scope: !55, file: !55, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !26, retainedNodes: !14)
!129 = !DILocation(line: 43, scope: !100)
!130 = !DILocation(line: 48, scope: !100)
!131 = distinct !DISubprogram(name: "gbsa_force_2_kernel!", linkageName: "julia_gbsa_force_2_kernel!_25091", scope: null, file: !5, line: 1013, type: !49, scopeLine: 1013, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!132 = !DILocation(line: 0, scope: !131)
!133 = !DILocation(line: 88, scope: !134, inlinedAt: !136)
!134 = distinct !DISubprogram(name: "*;", linkageName: "*", scope: !135, file: !135, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!135 = !DIFile(filename: "int.jl", directory: ".")
!136 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !139)
!137 = distinct !DISubprogram(name: "literal_pow;", linkageName: "literal_pow", scope: !138, file: !138, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!138 = !DIFile(filename: "intfuncs.jl", directory: ".")
!139 = distinct !DILocation(line: 1024, scope: !131, inlinedAt: !140)
!140 = distinct !DILocation(line: 0, scope: !131)
!141 = !DILocation(line: 40, scope: !142, inlinedAt: !143)
!142 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !52, file: !52, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!143 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !146)
!144 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!145 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/intrinsics/indexing.jl", directory: ".")
!146 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !148)
!147 = distinct !DISubprogram(name: "_index;", linkageName: "_index", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!148 = distinct !DILocation(line: 56, scope: !149, inlinedAt: !150)
!149 = distinct !DISubprogram(name: "blockIdx_x;", linkageName: "blockIdx_x", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!150 = distinct !DILocation(line: 78, scope: !151, inlinedAt: !152)
!151 = distinct !DISubprogram(name: "#blockIdx;", linkageName: "#blockIdx", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!152 = distinct !DILocation(line: 1025, scope: !131, inlinedAt: !140)
!153 = !{i32 0, i32 2147483646}
!154 = !DILocation(line: 86, scope: !155, inlinedAt: !156)
!155 = distinct !DISubprogram(name: "-;", linkageName: "-", scope: !135, file: !135, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!156 = distinct !DILocation(line: 1042, scope: !155, inlinedAt: !152)
!157 = !DILocation(line: 40, scope: !142, inlinedAt: !158)
!158 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !159)
!159 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !160)
!160 = distinct !DILocation(line: 51, scope: !161, inlinedAt: !162)
!161 = distinct !DISubprogram(name: "blockDim_x;", linkageName: "blockDim_x", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!162 = distinct !DILocation(line: 85, scope: !163, inlinedAt: !152)
!163 = distinct !DISubprogram(name: "#blockDim;", linkageName: "#blockDim", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!164 = !{i32 1, i32 1024}
!165 = !DILocation(line: 523, scope: !166, inlinedAt: !167)
!166 = distinct !DISubprogram(name: "rem;", linkageName: "rem", scope: !135, file: !135, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!167 = distinct !DILocation(line: 1040, scope: !134, inlinedAt: !152)
!168 = !DILocation(line: 88, scope: !134, inlinedAt: !169)
!169 = distinct !DILocation(line: 1042, scope: !134, inlinedAt: !152)
!170 = !DILocation(line: 40, scope: !142, inlinedAt: !171)
!171 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !172)
!172 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !173)
!173 = distinct !DILocation(line: 46, scope: !174, inlinedAt: !175)
!174 = distinct !DISubprogram(name: "threadIdx_x;", linkageName: "threadIdx_x", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!175 = distinct !DILocation(line: 92, scope: !176, inlinedAt: !152)
!176 = distinct !DISubprogram(name: "#threadIdx;", linkageName: "#threadIdx", scope: !145, file: !145, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!177 = !{i32 0, i32 1023}
!178 = !DILocation(line: 87, scope: !179, inlinedAt: !173)
!179 = distinct !DISubprogram(name: "+;", linkageName: "+", scope: !135, file: !135, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!180 = !DILocation(line: 523, scope: !166, inlinedAt: !181)
!181 = distinct !DILocation(line: 1040, scope: !179, inlinedAt: !152)
!182 = !DILocation(line: 87, scope: !179, inlinedAt: !183)
!183 = distinct !DILocation(line: 1042, scope: !179, inlinedAt: !152)
!184 = !DILocation(line: 488, scope: !185, inlinedAt: !186)
!185 = distinct !DISubprogram(name: "<=;", linkageName: "<=", scope: !135, file: !135, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!186 = distinct !DILocation(line: 1027, scope: !131, inlinedAt: !140)
!187 = !DILocation(line: 1027, scope: !131, inlinedAt: !140)
!188 = !DILocation(line: 295, scope: !189, inlinedAt: !190)
!189 = distinct !DISubprogram(name: "div;", linkageName: "div", scope: !135, file: !135, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!190 = distinct !DILocation(line: 308, scope: !191, inlinedAt: !193)
!191 = distinct !DISubprogram(name: "div;", linkageName: "div", scope: !192, file: !192, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!192 = !DIFile(filename: "div.jl", directory: ".")
!193 = distinct !DILocation(line: 363, scope: !191, inlinedAt: !194)
!194 = distinct !DILocation(line: 320, scope: !195, inlinedAt: !196)
!195 = distinct !DISubprogram(name: "cld;", linkageName: "cld", scope: !192, file: !192, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!196 = distinct !DILocation(line: 1028, scope: !131, inlinedAt: !140)
!197 = !DILocation(line: 86, scope: !155, inlinedAt: !198)
!198 = distinct !DILocation(line: 40, scope: !142, inlinedAt: !199)
!199 = distinct !DILocation(line: 9, scope: !200, inlinedAt: !202)
!200 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !201, file: !201, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!201 = !DIFile(filename: "/home/jgreener/.julia/packages/LLVM/TLGyi/src/interop/pointer.jl", directory: ".")
!202 = distinct !DILocation(line: 9, scope: !203, inlinedAt: !204)
!203 = distinct !DISubprogram(name: "pointerref;", linkageName: "pointerref", scope: !201, file: !201, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!204 = distinct !DILocation(line: 85, scope: !205, inlinedAt: !206)
!205 = distinct !DISubprogram(name: "unsafe_load;", linkageName: "unsafe_load", scope: !201, file: !201, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!206 = distinct !DILocation(line: 52, scope: !207, inlinedAt: !209)
!207 = distinct !DISubprogram(name: "unsafe_cached_load;", linkageName: "unsafe_cached_load", scope: !208, file: !208, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!208 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/pointer.jl", directory: ".")
!209 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !212)
!210 = distinct !DISubprogram(name: "#const_arrayref;", linkageName: "#const_arrayref", scope: !211, file: !211, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!211 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/array.jl", directory: ".")
!212 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !214)
!213 = distinct !DISubprogram(name: "getindex;", linkageName: "getindex", scope: !211, file: !211, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!214 = distinct !DILocation(line: 1031, scope: !131, inlinedAt: !140)
!215 = !DILocation(line: 40, scope: !142, inlinedAt: !199)
!216 = !{!217, !217, i64 0, i64 0}
!217 = !{!"custom_tbaa_addrspace(1)", !218, i64 0}
!218 = !{!"custom_tbaa"}
!219 = !DILocation(line: 535, scope: !220, inlinedAt: !222)
!220 = distinct !DISubprogram(name: "<;", linkageName: "<", scope: !221, file: !221, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!221 = !DIFile(filename: "float.jl", directory: ".")
!222 = distinct !DILocation(line: 292, scope: !223, inlinedAt: !225)
!223 = distinct !DISubprogram(name: "vector_1D;", linkageName: "vector_1D", scope: !224, file: !224, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!224 = !DIFile(filename: "/home/jgreener/.julia/dev/Molly/src/spatial.jl", directory: ".")
!225 = distinct !DILocation(line: 313, scope: !226, inlinedAt: !227)
!226 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !224, file: !224, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!227 = distinct !DILocation(line: 311, scope: !228, inlinedAt: !214)
!228 = distinct !DISubprogram(name: "vector;", linkageName: "vector", scope: !224, file: !224, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!229 = !DILocation(line: 292, scope: !223, inlinedAt: !225)
!230 = !DILocation(line: 409, scope: !231, inlinedAt: !232)
!231 = distinct !DISubprogram(name: "-;", linkageName: "-", scope: !221, file: !221, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!232 = distinct !DILocation(line: 293, scope: !223, inlinedAt: !225)
!233 = !DILocation(line: 408, scope: !234, inlinedAt: !232)
!234 = distinct !DISubprogram(name: "+;", linkageName: "+", scope: !221, file: !221, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!235 = !DILocation(line: 535, scope: !220, inlinedAt: !232)
!236 = !DILocation(line: 293, scope: !223, inlinedAt: !225)
!237 = !DILocation(line: 409, scope: !231, inlinedAt: !238)
!238 = distinct !DILocation(line: 295, scope: !223, inlinedAt: !225)
!239 = !DILocation(line: 408, scope: !234, inlinedAt: !238)
!240 = !DILocation(line: 535, scope: !220, inlinedAt: !238)
!241 = !DILocation(line: 295, scope: !223, inlinedAt: !225)
!242 = !DILocation(line: 410, scope: !243, inlinedAt: !244)
!243 = distinct !DISubprogram(name: "*;", linkageName: "*", scope: !221, file: !221, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!244 = distinct !DILocation(line: 189, scope: !245, inlinedAt: !247)
!245 = distinct !DISubprogram(name: "abs2;", linkageName: "abs2", scope: !246, file: !246, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!246 = !DIFile(filename: "number.jl", directory: ".")
!247 = distinct !DILocation(line: 459, scope: !248, inlinedAt: !250)
!248 = distinct !DISubprogram(name: "norm_sqr;", linkageName: "norm_sqr", scope: !249, file: !249, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!249 = !DIFile(filename: "/cache/build/default-amdci5-5/julialang/julia-release-1-dot-9/usr/share/julia/stdlib/v1.9/LinearAlgebra/src/generic.jl", directory: ".")
!250 = distinct !DILocation(line: 266, scope: !251, inlinedAt: !253)
!251 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!252 = !DIFile(filename: "/home/jgreener/.julia/packages/StaticArrays/jO3cC/src/linalg.jl", directory: ".")
!253 = distinct !DILocation(line: 256, scope: !254, inlinedAt: !255)
!254 = distinct !DISubprogram(name: "_norm;", linkageName: "_norm", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!255 = distinct !DILocation(line: 255, scope: !256, inlinedAt: !257)
!256 = distinct !DISubprogram(name: "norm;", linkageName: "norm", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!257 = distinct !DILocation(line: 1032, scope: !131, inlinedAt: !140)
!258 = !DILocation(line: 408, scope: !234, inlinedAt: !250)
!259 = !DILocation(line: 223, scope: !260, inlinedAt: !250)
!260 = distinct !DISubprogram(name: "#sqrt;", linkageName: "#sqrt", scope: !261, file: !261, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!261 = !DIFile(filename: "/home/jgreener/.julia/dev/CUDA/src/device/intrinsics/math.jl", directory: ".")
!262 = !DILocation(line: 535, scope: !220, inlinedAt: !263)
!263 = distinct !DILocation(line: 268, scope: !251, inlinedAt: !253)
!264 = !DILocation(line: 268, scope: !251, inlinedAt: !253)
!265 = !DILocation(line: 189, scope: !266, inlinedAt: !263)
!266 = distinct !DISubprogram(name: "#isfinite;", linkageName: "#isfinite", scope: !261, file: !261, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!267 = !DILocation(line: 215, scope: !268, inlinedAt: !269)
!268 = distinct !DISubprogram(name: "#abs;", linkageName: "#abs", scope: !261, file: !261, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!269 = distinct !DILocation(line: 228, scope: !270, inlinedAt: !271)
!270 = distinct !DISubprogram(name: "maxabs_nested;", linkageName: "maxabs_nested", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!271 = distinct !DILocation(line: 232, scope: !270, inlinedAt: !272)
!272 = distinct !DILocation(line: 248, scope: !251, inlinedAt: !273)
!273 = distinct !DILocation(line: 240, scope: !274, inlinedAt: !275)
!274 = distinct !DISubprogram(name: "_norm_scaled;", linkageName: "_norm_scaled", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!275 = distinct !DILocation(line: 269, scope: !251, inlinedAt: !253)
!276 = !DILocation(line: 215, scope: !268, inlinedAt: !277)
!277 = distinct !DILocation(line: 228, scope: !270, inlinedAt: !278)
!278 = distinct !DILocation(line: 234, scope: !270, inlinedAt: !272)
!279 = !DILocation(line: 575, scope: !280, inlinedAt: !282)
!280 = distinct !DISubprogram(name: "ifelse;", linkageName: "ifelse", scope: !281, file: !281, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!281 = !DIFile(filename: "essentials.jl", directory: ".")
!282 = distinct !DILocation(line: 243, scope: !283, inlinedAt: !278)
!283 = distinct !DISubprogram(name: "max_fast;", linkageName: "max_fast", scope: !284, file: !284, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!284 = !DIFile(filename: "fastmath.jl", directory: ".")
!285 = !DILocation(line: 533, scope: !286, inlinedAt: !287)
!286 = distinct !DISubprogram(name: "==;", linkageName: "==", scope: !221, file: !221, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!287 = distinct !DILocation(line: 42, scope: !288, inlinedAt: !289)
!288 = distinct !DISubprogram(name: "iszero;", linkageName: "iszero", scope: !246, file: !246, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!289 = distinct !DILocation(line: 250, scope: !251, inlinedAt: !273)
!290 = !DILocation(line: 250, scope: !251, inlinedAt: !273)
!291 = !DILocation(line: 215, scope: !268, inlinedAt: !292)
!292 = distinct !DILocation(line: 639, scope: !293, inlinedAt: !294)
!293 = distinct !DISubprogram(name: "norm;", linkageName: "norm", scope: !249, file: !249, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!294 = distinct !DILocation(line: 639, scope: !293, inlinedAt: !295)
!295 = distinct !DILocation(line: 222, scope: !296, inlinedAt: !289)
!296 = distinct !DISubprogram(name: "_init_zero;", linkageName: "_init_zero", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!297 = !DILocation(line: 891, scope: !298, inlinedAt: !300)
!298 = distinct !DISubprogram(name: "iterate;", linkageName: "iterate", scope: !299, file: !299, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!299 = !DIFile(filename: "range.jl", directory: ".")
!300 = distinct !DILocation(line: 235, scope: !270, inlinedAt: !272)
!301 = !DILocation(line: 411, scope: !302, inlinedAt: !303)
!302 = distinct !DISubprogram(name: "/;", linkageName: "/", scope: !221, file: !221, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!303 = distinct !DILocation(line: 251, scope: !251, inlinedAt: !273)
!304 = !DILocation(line: 410, scope: !243, inlinedAt: !305)
!305 = distinct !DILocation(line: 189, scope: !245, inlinedAt: !306)
!306 = distinct !DILocation(line: 459, scope: !248, inlinedAt: !303)
!307 = !DILocation(line: 408, scope: !234, inlinedAt: !303)
!308 = !DILocation(line: 223, scope: !260, inlinedAt: !303)
!309 = !DILocation(line: 410, scope: !243, inlinedAt: !303)
!310 = !DILocation(line: 533, scope: !286, inlinedAt: !311)
!311 = distinct !DILocation(line: 42, scope: !288, inlinedAt: !312)
!312 = distinct !DILocation(line: 1034, scope: !131, inlinedAt: !140)
!313 = !DILocation(line: 1034, scope: !131, inlinedAt: !140)
!314 = !DILocation(line: 88, scope: !134, inlinedAt: !315)
!315 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !317)
!316 = distinct !DISubprogram(name: "pointerref_ldg;", linkageName: "pointerref_ldg", scope: !208, file: !208, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!317 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !318)
!318 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !319)
!319 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !320)
!320 = distinct !DILocation(line: 1035, scope: !131, inlinedAt: !140)
!321 = !DILocation(line: 114, scope: !322, inlinedAt: !315)
!322 = distinct !DISubprogram(name: "+;", linkageName: "+", scope: !201, file: !201, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!323 = !DILocation(line: 40, scope: !142, inlinedAt: !324)
!324 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !325)
!325 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !327)
!326 = distinct !DISubprogram(name: "_typed_llvmcall;", linkageName: "_typed_llvmcall", scope: !201, file: !201, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!327 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !328)
!328 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !317)
!329 = !DILocation(line: 408, scope: !234, inlinedAt: !330)
!330 = distinct !DILocation(line: 1036, scope: !131, inlinedAt: !140)
!331 = !DILocation(line: 535, scope: !220, inlinedAt: !332)
!332 = distinct !DILocation(line: 1037, scope: !131, inlinedAt: !140)
!333 = !DILocation(line: 1037, scope: !131, inlinedAt: !140)
!334 = !DILocation(line: 409, scope: !231, inlinedAt: !335)
!335 = distinct !DILocation(line: 1038, scope: !131, inlinedAt: !140)
!336 = !DILocation(line: 215, scope: !268, inlinedAt: !335)
!337 = !DILocation(line: 292, scope: !338, inlinedAt: !335)
!338 = distinct !DISubprogram(name: "#max;", linkageName: "#max", scope: !261, file: !261, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!339 = !DILocation(line: 411, scope: !302, inlinedAt: !340)
!340 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !335)
!341 = distinct !DISubprogram(name: "inv;", linkageName: "inv", scope: !246, file: !246, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!342 = !DILocation(line: 411, scope: !302, inlinedAt: !343)
!343 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !344)
!344 = distinct !DILocation(line: 1039, scope: !131, inlinedAt: !140)
!345 = !DILocation(line: 411, scope: !302, inlinedAt: !346)
!346 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !347)
!347 = distinct !DILocation(line: 1040, scope: !131, inlinedAt: !140)
!348 = !DILocation(line: 410, scope: !243, inlinedAt: !349)
!349 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !350)
!350 = distinct !DILocation(line: 1041, scope: !131, inlinedAt: !140)
!351 = !DILocation(line: 410, scope: !243, inlinedAt: !352)
!352 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !353)
!353 = distinct !DILocation(line: 1042, scope: !131, inlinedAt: !140)
!354 = !DILocation(line: 410, scope: !243, inlinedAt: !353)
!355 = !DILocation(line: 408, scope: !234, inlinedAt: !356)
!356 = distinct !DILocation(line: 410, scope: !357, inlinedAt: !353)
!357 = distinct !DISubprogram(name: "+;", linkageName: "+", scope: !358, file: !358, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!358 = !DIFile(filename: "promotion.jl", directory: ".")
!359 = !DILocation(line: 409, scope: !231, inlinedAt: !353)
!360 = !DILocation(line: 411, scope: !302, inlinedAt: !361)
!361 = distinct !DILocation(line: 413, scope: !362, inlinedAt: !353)
!362 = distinct !DISubprogram(name: "/;", linkageName: "/", scope: !358, file: !358, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!363 = !DILocation(line: 411, scope: !302, inlinedAt: !353)
!364 = !DILocation(line: 105, scope: !365, inlinedAt: !353)
!365 = distinct !DISubprogram(name: "#log;", linkageName: "#log", scope: !261, file: !261, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!366 = !DILocation(line: 408, scope: !234, inlinedAt: !353)
!367 = !DILocation(line: 40, scope: !142, inlinedAt: !368)
!368 = distinct !DILocation(line: 9, scope: !200, inlinedAt: !369)
!369 = distinct !DILocation(line: 9, scope: !203, inlinedAt: !370)
!370 = distinct !DILocation(line: 85, scope: !205, inlinedAt: !371)
!371 = distinct !DILocation(line: 91, scope: !372, inlinedAt: !373)
!372 = distinct !DISubprogram(name: "arrayref_bits;", linkageName: "arrayref_bits", scope: !211, file: !211, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!373 = distinct !DILocation(line: 85, scope: !374, inlinedAt: !375)
!374 = distinct !DISubprogram(name: "#arrayref;", linkageName: "#arrayref", scope: !211, file: !211, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!375 = distinct !DILocation(line: 164, scope: !213, inlinedAt: !376)
!376 = distinct !DILocation(line: 1043, scope: !131, inlinedAt: !140)
!377 = !DILocation(line: 114, scope: !322, inlinedAt: !378)
!378 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !379)
!379 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !380)
!380 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !381)
!381 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !376)
!382 = !DILocation(line: 40, scope: !142, inlinedAt: !383)
!383 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !384)
!384 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !385)
!385 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !386)
!386 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !379)
!387 = !DILocation(line: 410, scope: !243, inlinedAt: !388)
!388 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !376)
!389 = !DILocation(line: 410, scope: !243, inlinedAt: !390)
!390 = distinct !DILocation(line: 578, scope: !391, inlinedAt: !376)
!391 = distinct !DISubprogram(name: "*;", linkageName: "*", scope: !392, file: !392, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!392 = !DIFile(filename: "operators.jl", directory: ".")
!393 = !DILocation(line: 575, scope: !280, inlinedAt: !394)
!394 = distinct !DILocation(line: 510, scope: !395, inlinedAt: !396)
!395 = distinct !DISubprogram(name: "max;", linkageName: "max", scope: !358, file: !358, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!396 = distinct !DILocation(line: 444, scope: !397, inlinedAt: !398)
!397 = distinct !DISubprogram(name: "OneTo;", linkageName: "OneTo", scope: !299, file: !299, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!398 = distinct !DILocation(line: 457, scope: !397, inlinedAt: !399)
!399 = distinct !DILocation(line: 459, scope: !400, inlinedAt: !401)
!400 = distinct !DISubprogram(name: "oneto;", linkageName: "oneto", scope: !299, file: !299, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!401 = distinct !DILocation(line: 274, scope: !402, inlinedAt: !404)
!402 = distinct !DISubprogram(name: "map;", linkageName: "map", scope: !403, file: !403, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!403 = !DIFile(filename: "tuple.jl", directory: ".")
!404 = distinct !DILocation(line: 98, scope: !405, inlinedAt: !407)
!405 = distinct !DISubprogram(name: "axes;", linkageName: "axes", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!406 = !DIFile(filename: "abstractarray.jl", directory: ".")
!407 = distinct !DILocation(line: 203, scope: !408, inlinedAt: !409)
!408 = distinct !DISubprogram(name: "axes;", linkageName: "axes", scope: !211, file: !211, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!409 = distinct !DILocation(line: 2933, scope: !410, inlinedAt: !411)
!410 = distinct !DISubprogram(name: "_sub2ind;", linkageName: "_sub2ind", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!411 = distinct !DILocation(line: 1333, scope: !412, inlinedAt: !413)
!412 = distinct !DISubprogram(name: "_to_linear_index;", linkageName: "_to_linear_index", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!413 = distinct !DILocation(line: 1327, scope: !414, inlinedAt: !415)
!414 = distinct !DISubprogram(name: "_getindex;", linkageName: "_getindex", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!415 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !417)
!416 = distinct !DISubprogram(name: "getindex;", linkageName: "getindex", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!417 = distinct !DILocation(line: 1044, scope: !131, inlinedAt: !140)
!418 = !DILocation(line: 88, scope: !134, inlinedAt: !419)
!419 = distinct !DILocation(line: 2965, scope: !420, inlinedAt: !421)
!420 = distinct !DISubprogram(name: "_sub2ind_recurse;", linkageName: "_sub2ind_recurse", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!421 = distinct !DILocation(line: 2965, scope: !420, inlinedAt: !422)
!422 = distinct !DILocation(line: 2949, scope: !410, inlinedAt: !409)
!423 = !DILocation(line: 87, scope: !179, inlinedAt: !419)
!424 = !DILocation(line: 88, scope: !134, inlinedAt: !425)
!425 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !426)
!426 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !427)
!427 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !428)
!428 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !413)
!429 = !DILocation(line: 114, scope: !322, inlinedAt: !425)
!430 = !DILocation(line: 40, scope: !142, inlinedAt: !431)
!431 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !432)
!432 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !433)
!433 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !434)
!434 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !426)
!435 = !DILocation(line: 409, scope: !231, inlinedAt: !417)
!436 = !DILocation(line: 410, scope: !243, inlinedAt: !437)
!437 = distinct !DILocation(line: 578, scope: !391, inlinedAt: !417)
!438 = !DILocation(line: 406, scope: !231, inlinedAt: !439)
!439 = distinct !DILocation(line: 75, scope: !440, inlinedAt: !442)
!440 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !441, file: !441, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!441 = !DIFile(filename: "/home/jgreener/.julia/packages/StaticArrays/jO3cC/src/mapreduce.jl", directory: ".")
!442 = distinct !DILocation(line: 40, scope: !443, inlinedAt: !444)
!443 = distinct !DISubprogram(name: "_map;", linkageName: "_map", scope: !441, file: !441, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!444 = distinct !DILocation(line: 31, scope: !445, inlinedAt: !446)
!445 = distinct !DISubprogram(name: "map;", linkageName: "map", scope: !441, file: !441, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!446 = distinct !DILocation(line: 8, scope: !447, inlinedAt: !448)
!447 = distinct !DISubprogram(name: "-;", linkageName: "-", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!448 = distinct !DILocation(line: 1045, scope: !131, inlinedAt: !140)
!449 = !DILocation(line: 410, scope: !243, inlinedAt: !450)
!450 = distinct !DILocation(line: 22, scope: !451, inlinedAt: !452)
!451 = distinct !DISubprogram(name: "#267;", linkageName: "#267", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!452 = distinct !DILocation(line: 75, scope: !440, inlinedAt: !453)
!453 = distinct !DILocation(line: 40, scope: !443, inlinedAt: !454)
!454 = distinct !DILocation(line: 31, scope: !445, inlinedAt: !455)
!455 = distinct !DILocation(line: 22, scope: !456, inlinedAt: !448)
!456 = distinct !DISubprogram(name: "*;", linkageName: "*", scope: !252, file: !252, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!457 = !DILocation(line: 1049, scope: !131, inlinedAt: !140)
!458 = !DILocation(line: 444, scope: !397, inlinedAt: !459)
!459 = distinct !DILocation(line: 457, scope: !397, inlinedAt: !460)
!460 = distinct !DILocation(line: 459, scope: !400, inlinedAt: !461)
!461 = distinct !DILocation(line: 274, scope: !402, inlinedAt: !462)
!462 = distinct !DILocation(line: 98, scope: !405, inlinedAt: !463)
!463 = distinct !DILocation(line: 486, scope: !464, inlinedAt: !466)
!464 = distinct !DISubprogram(name: "LinearIndices;", linkageName: "LinearIndices", scope: !465, file: !465, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!465 = !DIFile(filename: "indices.jl", directory: ".")
!466 = distinct !DILocation(line: 102, scope: !467, inlinedAt: !469)
!467 = distinct !DISubprogram(name: "pointer;", linkageName: "pointer", scope: !468, file: !468, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!468 = !DIFile(filename: "/home/jgreener/.julia/packages/Atomix/F9VIX/src/references.jl", directory: ".")
!469 = distinct !DILocation(line: 30, scope: !470, inlinedAt: !472)
!470 = distinct !DISubprogram(name: "modify!;", linkageName: "modify!", scope: !471, file: !471, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!471 = !DIFile(filename: "/home/jgreener/.julia/packages/Atomix/F9VIX/src/core.jl", directory: ".")
!472 = distinct !DILocation(line: 1051, scope: !131, inlinedAt: !140)
!473 = !{!474, !474, i64 0}
!474 = !{!"jtbaa_stack", !118, i64 0}
!475 = !{!476}
!476 = !{!"jnoalias_stack", !477}
!477 = !{!"jnoalias"}
!478 = !{!69, !479, !480, !481, !482}
!479 = !{!"jnoalias_gcframe", !477}
!480 = !{!"jnoalias_data", !477}
!481 = !{!"jnoalias_typemd", !477}
!482 = !{!"jnoalias_const", !477}
!483 = !DILocation(line: 707, scope: !484, inlinedAt: !485)
!484 = distinct !DISubprogram(name: "checkbounds;", linkageName: "checkbounds", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!485 = distinct !DILocation(line: 1326, scope: !414, inlinedAt: !486)
!486 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !466)
!487 = !DILocation(line: 709, scope: !484, inlinedAt: !485)
!488 = !DILocation(line: 88, scope: !134, inlinedAt: !489)
!489 = distinct !DILocation(line: 1247, scope: !490, inlinedAt: !491)
!490 = distinct !DISubprogram(name: "_memory_offset;", linkageName: "_memory_offset", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!491 = distinct !DILocation(line: 52, scope: !492, inlinedAt: !493)
!492 = distinct !DISubprogram(name: "pointer;", linkageName: "pointer", scope: !211, file: !211, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!493 = distinct !DILocation(line: 103, scope: !467, inlinedAt: !469)
!494 = !DILocation(line: 114, scope: !322, inlinedAt: !491)
!495 = !DILocation(line: 40, scope: !142, inlinedAt: !496)
!496 = distinct !DILocation(line: 255, scope: !497, inlinedAt: !499)
!497 = distinct !DISubprogram(name: "macro expansion;", linkageName: "macro expansion", scope: !498, file: !498, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!498 = !DIFile(filename: "/home/jgreener/.julia/packages/UnsafeAtomicsLLVM/xcnBP/src/atomics.jl", directory: ".")
!499 = distinct !DILocation(line: 255, scope: !500, inlinedAt: !501)
!500 = distinct !DISubprogram(name: "llvm_atomic_op;", linkageName: "llvm_atomic_op", scope: !498, file: !498, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!501 = distinct !DILocation(line: 359, scope: !502, inlinedAt: !503)
!502 = distinct !DISubprogram(name: "atomic_pointermodify;", linkageName: "atomic_pointermodify", scope: !498, file: !498, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!503 = distinct !DILocation(line: 20, scope: !504, inlinedAt: !506)
!504 = distinct !DISubprogram(name: "modify!;", linkageName: "modify!", scope: !505, file: !505, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!505 = !DIFile(filename: "/home/jgreener/.julia/packages/UnsafeAtomicsLLVM/xcnBP/src/internal.jl", directory: ".")
!506 = distinct !DILocation(line: 33, scope: !470, inlinedAt: !472)
!507 = !DILocation(line: 444, scope: !397, inlinedAt: !508)
!508 = distinct !DILocation(line: 457, scope: !397, inlinedAt: !509)
!509 = distinct !DILocation(line: 459, scope: !400, inlinedAt: !510)
!510 = distinct !DILocation(line: 274, scope: !402, inlinedAt: !511)
!511 = distinct !DILocation(line: 98, scope: !405, inlinedAt: !512)
!512 = distinct !DILocation(line: 486, scope: !464, inlinedAt: !513)
!513 = distinct !DILocation(line: 102, scope: !467, inlinedAt: !514)
!514 = distinct !DILocation(line: 30, scope: !470, inlinedAt: !515)
!515 = distinct !DILocation(line: 1052, scope: !131, inlinedAt: !140)
!516 = !DILocation(line: 707, scope: !484, inlinedAt: !517)
!517 = distinct !DILocation(line: 1326, scope: !414, inlinedAt: !518)
!518 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !513)
!519 = !DILocation(line: 709, scope: !484, inlinedAt: !517)
!520 = !DILocation(line: 406, scope: !231, inlinedAt: !515)
!521 = !DILocation(line: 88, scope: !134, inlinedAt: !522)
!522 = distinct !DILocation(line: 1247, scope: !490, inlinedAt: !523)
!523 = distinct !DILocation(line: 52, scope: !492, inlinedAt: !524)
!524 = distinct !DILocation(line: 103, scope: !467, inlinedAt: !514)
!525 = !DILocation(line: 114, scope: !322, inlinedAt: !523)
!526 = !DILocation(line: 40, scope: !142, inlinedAt: !527)
!527 = distinct !DILocation(line: 255, scope: !497, inlinedAt: !528)
!528 = distinct !DILocation(line: 255, scope: !500, inlinedAt: !529)
!529 = distinct !DILocation(line: 359, scope: !502, inlinedAt: !530)
!530 = distinct !DILocation(line: 20, scope: !504, inlinedAt: !531)
!531 = distinct !DILocation(line: 33, scope: !470, inlinedAt: !515)
!532 = !DILocation(line: 488, scope: !185, inlinedAt: !533)
!533 = distinct !DILocation(line: 768, scope: !534, inlinedAt: !535)
!534 = distinct !DISubprogram(name: "checkindex;", linkageName: "checkindex", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!535 = distinct !DILocation(line: 735, scope: !536, inlinedAt: !537)
!536 = distinct !DISubprogram(name: "checkbounds_indices;", linkageName: "checkbounds_indices", scope: !406, file: !406, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!537 = distinct !DILocation(line: 688, scope: !484, inlinedAt: !538)
!538 = distinct !DILocation(line: 709, scope: !484, inlinedAt: !485)
!539 = !DILocation(line: 1058, scope: !131, inlinedAt: !140)
!540 = !DILocation(line: 83, scope: !541, inlinedAt: !542)
!541 = distinct !DISubprogram(name: "<;", linkageName: "<", scope: !135, file: !135, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!542 = distinct !DILocation(line: 369, scope: !543, inlinedAt: !544)
!543 = distinct !DISubprogram(name: ">;", linkageName: ">", scope: !392, file: !392, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!544 = distinct !DILocation(line: 364, scope: !191, inlinedAt: !194)
!545 = !DILocation(line: 88, scope: !134, inlinedAt: !544)
!546 = !DILocation(line: 499, scope: !547, inlinedAt: !548)
!547 = distinct !DISubprogram(name: "==;", linkageName: "==", scope: !358, file: !358, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!548 = distinct !DILocation(line: 269, scope: !549, inlinedAt: !544)
!549 = distinct !DISubprogram(name: "!=;", linkageName: "!=", scope: !392, file: !392, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!550 = !DILocation(line: 38, scope: !551, inlinedAt: !544)
!551 = distinct !DISubprogram(name: "&;", linkageName: "&", scope: !552, file: !552, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!552 = !DIFile(filename: "bool.jl", directory: ".")
!553 = !DILocation(line: 711, scope: !554, inlinedAt: !556)
!554 = distinct !DISubprogram(name: "toInt64;", linkageName: "toInt64", scope: !555, file: !555, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!555 = !DIFile(filename: "boot.jl", directory: ".")
!556 = distinct !DILocation(line: 784, scope: !557, inlinedAt: !558)
!557 = distinct !DISubprogram(name: "Int64;", linkageName: "Int64", scope: !555, file: !555, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!558 = distinct !DILocation(line: 7, scope: !559, inlinedAt: !560)
!559 = distinct !DISubprogram(name: "convert;", linkageName: "convert", scope: !246, file: !246, type: !49, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!560 = distinct !DILocation(line: 520, scope: !166, inlinedAt: !561)
!561 = distinct !DILocation(line: 1040, scope: !179, inlinedAt: !544)
!562 = !DILocation(line: 87, scope: !179, inlinedAt: !563)
!563 = distinct !DILocation(line: 1042, scope: !179, inlinedAt: !544)
!564 = !DILocation(line: 296, scope: !166, inlinedAt: !565)
!565 = distinct !DILocation(line: 1029, scope: !131, inlinedAt: !140)
!566 = !DILocation(line: 86, scope: !155, inlinedAt: !565)
!567 = !DILocation(line: 87, scope: !179, inlinedAt: !565)
!568 = !DILocation(line: 499, scope: !547, inlinedAt: !569)
!569 = distinct !DILocation(line: 269, scope: !549, inlinedAt: !570)
!570 = distinct !DILocation(line: 1030, scope: !131, inlinedAt: !140)
!571 = !DILocation(line: 1030, scope: !131, inlinedAt: !140)
!572 = distinct !DISubprogram(name: "gbsa_force_2_kernel!", linkageName: "julia_gbsa_force_2_kernel!_25091", scope: null, file: !5, line: 1013, type: !49, scopeLine: 1013, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!573 = !DILocation(line: 0, scope: !572)
!574 = !DILocation(line: 88, scope: !134, inlinedAt: !575)
!575 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !576)
!576 = distinct !DILocation(line: 1024, scope: !572, inlinedAt: !577)
!577 = distinct !DILocation(line: 0, scope: !572)
!578 = !DILocation(line: 40, scope: !142, inlinedAt: !579)
!579 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !580)
!580 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !581)
!581 = distinct !DILocation(line: 56, scope: !149, inlinedAt: !582)
!582 = distinct !DILocation(line: 78, scope: !151, inlinedAt: !583)
!583 = distinct !DILocation(line: 1025, scope: !572, inlinedAt: !577)
!584 = !DILocation(line: 86, scope: !155, inlinedAt: !585)
!585 = distinct !DILocation(line: 1042, scope: !155, inlinedAt: !583)
!586 = !DILocation(line: 40, scope: !142, inlinedAt: !587)
!587 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !588)
!588 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !589)
!589 = distinct !DILocation(line: 51, scope: !161, inlinedAt: !590)
!590 = distinct !DILocation(line: 85, scope: !163, inlinedAt: !583)
!591 = !DILocation(line: 523, scope: !166, inlinedAt: !592)
!592 = distinct !DILocation(line: 1040, scope: !134, inlinedAt: !583)
!593 = !DILocation(line: 88, scope: !134, inlinedAt: !594)
!594 = distinct !DILocation(line: 1042, scope: !134, inlinedAt: !583)
!595 = !DILocation(line: 40, scope: !142, inlinedAt: !596)
!596 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !597)
!597 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !598)
!598 = distinct !DILocation(line: 46, scope: !174, inlinedAt: !599)
!599 = distinct !DILocation(line: 92, scope: !176, inlinedAt: !583)
!600 = !DILocation(line: 87, scope: !179, inlinedAt: !598)
!601 = !DILocation(line: 523, scope: !166, inlinedAt: !602)
!602 = distinct !DILocation(line: 1040, scope: !179, inlinedAt: !583)
!603 = !DILocation(line: 87, scope: !179, inlinedAt: !604)
!604 = distinct !DILocation(line: 1042, scope: !179, inlinedAt: !583)
!605 = !DILocation(line: 488, scope: !185, inlinedAt: !606)
!606 = distinct !DILocation(line: 1027, scope: !572, inlinedAt: !577)
!607 = !DILocation(line: 1027, scope: !572, inlinedAt: !577)
!608 = !DILocation(line: 295, scope: !189, inlinedAt: !609)
!609 = distinct !DILocation(line: 308, scope: !191, inlinedAt: !610)
!610 = distinct !DILocation(line: 363, scope: !191, inlinedAt: !611)
!611 = distinct !DILocation(line: 320, scope: !195, inlinedAt: !612)
!612 = distinct !DILocation(line: 1028, scope: !572, inlinedAt: !577)
!613 = !DILocation(line: 86, scope: !155, inlinedAt: !614)
!614 = distinct !DILocation(line: 40, scope: !142, inlinedAt: !615)
!615 = distinct !DILocation(line: 9, scope: !200, inlinedAt: !616)
!616 = distinct !DILocation(line: 9, scope: !203, inlinedAt: !617)
!617 = distinct !DILocation(line: 85, scope: !205, inlinedAt: !618)
!618 = distinct !DILocation(line: 52, scope: !207, inlinedAt: !619)
!619 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !620)
!620 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !621)
!621 = distinct !DILocation(line: 1031, scope: !572, inlinedAt: !577)
!622 = !DILocation(line: 40, scope: !142, inlinedAt: !615)
!623 = !DILocation(line: 535, scope: !220, inlinedAt: !624)
!624 = distinct !DILocation(line: 292, scope: !223, inlinedAt: !625)
!625 = distinct !DILocation(line: 313, scope: !226, inlinedAt: !626)
!626 = distinct !DILocation(line: 311, scope: !228, inlinedAt: !621)
!627 = !DILocation(line: 292, scope: !223, inlinedAt: !625)
!628 = !DILocation(line: 409, scope: !231, inlinedAt: !629)
!629 = distinct !DILocation(line: 293, scope: !223, inlinedAt: !625)
!630 = !DILocation(line: 408, scope: !234, inlinedAt: !629)
!631 = !DILocation(line: 535, scope: !220, inlinedAt: !629)
!632 = !DILocation(line: 293, scope: !223, inlinedAt: !625)
!633 = !DILocation(line: 409, scope: !231, inlinedAt: !634)
!634 = distinct !DILocation(line: 295, scope: !223, inlinedAt: !625)
!635 = !DILocation(line: 408, scope: !234, inlinedAt: !634)
!636 = !DILocation(line: 535, scope: !220, inlinedAt: !634)
!637 = !DILocation(line: 295, scope: !223, inlinedAt: !625)
!638 = !DILocation(line: 410, scope: !243, inlinedAt: !639)
!639 = distinct !DILocation(line: 189, scope: !245, inlinedAt: !640)
!640 = distinct !DILocation(line: 459, scope: !248, inlinedAt: !641)
!641 = distinct !DILocation(line: 266, scope: !251, inlinedAt: !642)
!642 = distinct !DILocation(line: 256, scope: !254, inlinedAt: !643)
!643 = distinct !DILocation(line: 255, scope: !256, inlinedAt: !644)
!644 = distinct !DILocation(line: 1032, scope: !572, inlinedAt: !577)
!645 = !DILocation(line: 408, scope: !234, inlinedAt: !641)
!646 = !DILocation(line: 223, scope: !260, inlinedAt: !641)
!647 = !DILocation(line: 535, scope: !220, inlinedAt: !648)
!648 = distinct !DILocation(line: 268, scope: !251, inlinedAt: !642)
!649 = !DILocation(line: 268, scope: !251, inlinedAt: !642)
!650 = !DILocation(line: 189, scope: !266, inlinedAt: !648)
!651 = !DILocation(line: 215, scope: !268, inlinedAt: !652)
!652 = distinct !DILocation(line: 228, scope: !270, inlinedAt: !653)
!653 = distinct !DILocation(line: 232, scope: !270, inlinedAt: !654)
!654 = distinct !DILocation(line: 248, scope: !251, inlinedAt: !655)
!655 = distinct !DILocation(line: 240, scope: !274, inlinedAt: !656)
!656 = distinct !DILocation(line: 269, scope: !251, inlinedAt: !642)
!657 = !DILocation(line: 215, scope: !268, inlinedAt: !658)
!658 = distinct !DILocation(line: 228, scope: !270, inlinedAt: !659)
!659 = distinct !DILocation(line: 234, scope: !270, inlinedAt: !654)
!660 = !DILocation(line: 575, scope: !280, inlinedAt: !661)
!661 = distinct !DILocation(line: 243, scope: !283, inlinedAt: !659)
!662 = !DILocation(line: 533, scope: !286, inlinedAt: !663)
!663 = distinct !DILocation(line: 42, scope: !288, inlinedAt: !664)
!664 = distinct !DILocation(line: 250, scope: !251, inlinedAt: !655)
!665 = !DILocation(line: 250, scope: !251, inlinedAt: !655)
!666 = !DILocation(line: 215, scope: !268, inlinedAt: !667)
!667 = distinct !DILocation(line: 639, scope: !293, inlinedAt: !668)
!668 = distinct !DILocation(line: 639, scope: !293, inlinedAt: !669)
!669 = distinct !DILocation(line: 222, scope: !296, inlinedAt: !664)
!670 = !DILocation(line: 891, scope: !298, inlinedAt: !671)
!671 = distinct !DILocation(line: 235, scope: !270, inlinedAt: !654)
!672 = !DILocation(line: 411, scope: !302, inlinedAt: !673)
!673 = distinct !DILocation(line: 251, scope: !251, inlinedAt: !655)
!674 = !DILocation(line: 410, scope: !243, inlinedAt: !675)
!675 = distinct !DILocation(line: 189, scope: !245, inlinedAt: !676)
!676 = distinct !DILocation(line: 459, scope: !248, inlinedAt: !673)
!677 = !DILocation(line: 408, scope: !234, inlinedAt: !673)
!678 = !DILocation(line: 223, scope: !260, inlinedAt: !673)
!679 = !DILocation(line: 410, scope: !243, inlinedAt: !673)
!680 = !DILocation(line: 533, scope: !286, inlinedAt: !681)
!681 = distinct !DILocation(line: 42, scope: !288, inlinedAt: !682)
!682 = distinct !DILocation(line: 1034, scope: !572, inlinedAt: !577)
!683 = !DILocation(line: 1034, scope: !572, inlinedAt: !577)
!684 = !DILocation(line: 88, scope: !134, inlinedAt: !685)
!685 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !686)
!686 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !687)
!687 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !688)
!688 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !689)
!689 = distinct !DILocation(line: 1035, scope: !572, inlinedAt: !577)
!690 = !DILocation(line: 114, scope: !322, inlinedAt: !685)
!691 = !DILocation(line: 40, scope: !142, inlinedAt: !692)
!692 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !693)
!693 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !694)
!694 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !695)
!695 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !686)
!696 = !DILocation(line: 408, scope: !234, inlinedAt: !697)
!697 = distinct !DILocation(line: 1036, scope: !572, inlinedAt: !577)
!698 = !DILocation(line: 535, scope: !220, inlinedAt: !699)
!699 = distinct !DILocation(line: 1037, scope: !572, inlinedAt: !577)
!700 = !DILocation(line: 1037, scope: !572, inlinedAt: !577)
!701 = !DILocation(line: 409, scope: !231, inlinedAt: !702)
!702 = distinct !DILocation(line: 1038, scope: !572, inlinedAt: !577)
!703 = !DILocation(line: 215, scope: !268, inlinedAt: !702)
!704 = !DILocation(line: 292, scope: !338, inlinedAt: !702)
!705 = !DILocation(line: 411, scope: !302, inlinedAt: !706)
!706 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !702)
!707 = !DILocation(line: 411, scope: !302, inlinedAt: !708)
!708 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !709)
!709 = distinct !DILocation(line: 1039, scope: !572, inlinedAt: !577)
!710 = !DILocation(line: 411, scope: !302, inlinedAt: !711)
!711 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !712)
!712 = distinct !DILocation(line: 1040, scope: !572, inlinedAt: !577)
!713 = !DILocation(line: 410, scope: !243, inlinedAt: !714)
!714 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !715)
!715 = distinct !DILocation(line: 1041, scope: !572, inlinedAt: !577)
!716 = !DILocation(line: 410, scope: !243, inlinedAt: !717)
!717 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !718)
!718 = distinct !DILocation(line: 1042, scope: !572, inlinedAt: !577)
!719 = !DILocation(line: 410, scope: !243, inlinedAt: !718)
!720 = !DILocation(line: 408, scope: !234, inlinedAt: !721)
!721 = distinct !DILocation(line: 410, scope: !357, inlinedAt: !718)
!722 = !DILocation(line: 409, scope: !231, inlinedAt: !718)
!723 = !DILocation(line: 411, scope: !302, inlinedAt: !724)
!724 = distinct !DILocation(line: 413, scope: !362, inlinedAt: !718)
!725 = !DILocation(line: 411, scope: !302, inlinedAt: !718)
!726 = !DILocation(line: 105, scope: !365, inlinedAt: !718)
!727 = !DILocation(line: 408, scope: !234, inlinedAt: !718)
!728 = !DILocation(line: 40, scope: !142, inlinedAt: !729)
!729 = distinct !DILocation(line: 9, scope: !200, inlinedAt: !730)
!730 = distinct !DILocation(line: 9, scope: !203, inlinedAt: !731)
!731 = distinct !DILocation(line: 85, scope: !205, inlinedAt: !732)
!732 = distinct !DILocation(line: 91, scope: !372, inlinedAt: !733)
!733 = distinct !DILocation(line: 85, scope: !374, inlinedAt: !734)
!734 = distinct !DILocation(line: 164, scope: !213, inlinedAt: !735)
!735 = distinct !DILocation(line: 1043, scope: !572, inlinedAt: !577)
!736 = !DILocation(line: 114, scope: !322, inlinedAt: !737)
!737 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !738)
!738 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !739)
!739 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !740)
!740 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !735)
!741 = !DILocation(line: 40, scope: !142, inlinedAt: !742)
!742 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !743)
!743 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !744)
!744 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !745)
!745 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !738)
!746 = !DILocation(line: 410, scope: !243, inlinedAt: !747)
!747 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !735)
!748 = !DILocation(line: 410, scope: !243, inlinedAt: !749)
!749 = distinct !DILocation(line: 578, scope: !391, inlinedAt: !735)
!750 = !DILocation(line: 575, scope: !280, inlinedAt: !751)
!751 = distinct !DILocation(line: 510, scope: !395, inlinedAt: !752)
!752 = distinct !DILocation(line: 444, scope: !397, inlinedAt: !753)
!753 = distinct !DILocation(line: 457, scope: !397, inlinedAt: !754)
!754 = distinct !DILocation(line: 459, scope: !400, inlinedAt: !755)
!755 = distinct !DILocation(line: 274, scope: !402, inlinedAt: !756)
!756 = distinct !DILocation(line: 98, scope: !405, inlinedAt: !757)
!757 = distinct !DILocation(line: 203, scope: !408, inlinedAt: !758)
!758 = distinct !DILocation(line: 2933, scope: !410, inlinedAt: !759)
!759 = distinct !DILocation(line: 1333, scope: !412, inlinedAt: !760)
!760 = distinct !DILocation(line: 1327, scope: !414, inlinedAt: !761)
!761 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !762)
!762 = distinct !DILocation(line: 1044, scope: !572, inlinedAt: !577)
!763 = !DILocation(line: 88, scope: !134, inlinedAt: !764)
!764 = distinct !DILocation(line: 2965, scope: !420, inlinedAt: !765)
!765 = distinct !DILocation(line: 2965, scope: !420, inlinedAt: !766)
!766 = distinct !DILocation(line: 2949, scope: !410, inlinedAt: !758)
!767 = !DILocation(line: 87, scope: !179, inlinedAt: !764)
!768 = !DILocation(line: 88, scope: !134, inlinedAt: !769)
!769 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !770)
!770 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !771)
!771 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !772)
!772 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !760)
!773 = !DILocation(line: 114, scope: !322, inlinedAt: !769)
!774 = !DILocation(line: 40, scope: !142, inlinedAt: !775)
!775 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !776)
!776 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !777)
!777 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !778)
!778 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !770)
!779 = !DILocation(line: 409, scope: !231, inlinedAt: !762)
!780 = !DILocation(line: 410, scope: !243, inlinedAt: !781)
!781 = distinct !DILocation(line: 578, scope: !391, inlinedAt: !762)
!782 = !DILocation(line: 406, scope: !231, inlinedAt: !783)
!783 = distinct !DILocation(line: 75, scope: !440, inlinedAt: !784)
!784 = distinct !DILocation(line: 40, scope: !443, inlinedAt: !785)
!785 = distinct !DILocation(line: 31, scope: !445, inlinedAt: !786)
!786 = distinct !DILocation(line: 8, scope: !447, inlinedAt: !787)
!787 = distinct !DILocation(line: 1045, scope: !572, inlinedAt: !577)
!788 = !DILocation(line: 410, scope: !243, inlinedAt: !789)
!789 = distinct !DILocation(line: 22, scope: !451, inlinedAt: !790)
!790 = distinct !DILocation(line: 75, scope: !440, inlinedAt: !791)
!791 = distinct !DILocation(line: 40, scope: !443, inlinedAt: !792)
!792 = distinct !DILocation(line: 31, scope: !445, inlinedAt: !793)
!793 = distinct !DILocation(line: 22, scope: !456, inlinedAt: !787)
!794 = !DILocation(line: 1049, scope: !572, inlinedAt: !577)
!795 = !DILocation(line: 88, scope: !134, inlinedAt: !796)
!796 = distinct !DILocation(line: 1247, scope: !490, inlinedAt: !797)
!797 = distinct !DILocation(line: 52, scope: !492, inlinedAt: !798)
!798 = distinct !DILocation(line: 103, scope: !467, inlinedAt: !799)
!799 = distinct !DILocation(line: 30, scope: !470, inlinedAt: !800)
!800 = distinct !DILocation(line: 1051, scope: !572, inlinedAt: !577)
!801 = !DILocation(line: 114, scope: !322, inlinedAt: !797)
!802 = !DILocation(line: 40, scope: !142, inlinedAt: !803)
!803 = distinct !DILocation(line: 255, scope: !497, inlinedAt: !804)
!804 = distinct !DILocation(line: 255, scope: !500, inlinedAt: !805)
!805 = distinct !DILocation(line: 359, scope: !502, inlinedAt: !806)
!806 = distinct !DILocation(line: 20, scope: !504, inlinedAt: !807)
!807 = distinct !DILocation(line: 33, scope: !470, inlinedAt: !800)
!808 = !DILocation(line: 709, scope: !484, inlinedAt: !809)
!809 = distinct !DILocation(line: 1326, scope: !414, inlinedAt: !810)
!810 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !811)
!811 = distinct !DILocation(line: 102, scope: !467, inlinedAt: !812)
!812 = distinct !DILocation(line: 30, scope: !470, inlinedAt: !813)
!813 = distinct !DILocation(line: 1052, scope: !572, inlinedAt: !577)
!814 = !DILocation(line: 406, scope: !231, inlinedAt: !813)
!815 = !DILocation(line: 88, scope: !134, inlinedAt: !816)
!816 = distinct !DILocation(line: 1247, scope: !490, inlinedAt: !817)
!817 = distinct !DILocation(line: 52, scope: !492, inlinedAt: !818)
!818 = distinct !DILocation(line: 103, scope: !467, inlinedAt: !812)
!819 = !DILocation(line: 114, scope: !322, inlinedAt: !817)
!820 = !DILocation(line: 40, scope: !142, inlinedAt: !821)
!821 = distinct !DILocation(line: 255, scope: !497, inlinedAt: !822)
!822 = distinct !DILocation(line: 255, scope: !500, inlinedAt: !823)
!823 = distinct !DILocation(line: 359, scope: !502, inlinedAt: !824)
!824 = distinct !DILocation(line: 20, scope: !504, inlinedAt: !825)
!825 = distinct !DILocation(line: 33, scope: !470, inlinedAt: !813)
!826 = !DILocation(line: 488, scope: !185, inlinedAt: !827)
!827 = distinct !DILocation(line: 768, scope: !534, inlinedAt: !828)
!828 = distinct !DILocation(line: 735, scope: !536, inlinedAt: !829)
!829 = distinct !DILocation(line: 688, scope: !484, inlinedAt: !830)
!830 = distinct !DILocation(line: 709, scope: !484, inlinedAt: !831)
!831 = distinct !DILocation(line: 1326, scope: !414, inlinedAt: !832)
!832 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !833)
!833 = distinct !DILocation(line: 102, scope: !467, inlinedAt: !799)
!834 = !DILocation(line: 709, scope: !484, inlinedAt: !831)
!835 = !DILocation(line: 1058, scope: !572, inlinedAt: !577)
!836 = !DILocation(line: 83, scope: !541, inlinedAt: !837)
!837 = distinct !DILocation(line: 369, scope: !543, inlinedAt: !838)
!838 = distinct !DILocation(line: 364, scope: !191, inlinedAt: !611)
!839 = !DILocation(line: 88, scope: !134, inlinedAt: !838)
!840 = !DILocation(line: 499, scope: !547, inlinedAt: !841)
!841 = distinct !DILocation(line: 269, scope: !549, inlinedAt: !838)
!842 = !DILocation(line: 38, scope: !551, inlinedAt: !838)
!843 = !DILocation(line: 711, scope: !554, inlinedAt: !844)
!844 = distinct !DILocation(line: 784, scope: !557, inlinedAt: !845)
!845 = distinct !DILocation(line: 7, scope: !559, inlinedAt: !846)
!846 = distinct !DILocation(line: 520, scope: !166, inlinedAt: !847)
!847 = distinct !DILocation(line: 1040, scope: !179, inlinedAt: !838)
!848 = !DILocation(line: 87, scope: !179, inlinedAt: !849)
!849 = distinct !DILocation(line: 1042, scope: !179, inlinedAt: !838)
!850 = !DILocation(line: 296, scope: !166, inlinedAt: !851)
!851 = distinct !DILocation(line: 1029, scope: !572, inlinedAt: !577)
!852 = !DILocation(line: 86, scope: !155, inlinedAt: !851)
!853 = !DILocation(line: 87, scope: !179, inlinedAt: !851)
!854 = !DILocation(line: 499, scope: !547, inlinedAt: !855)
!855 = distinct !DILocation(line: 269, scope: !549, inlinedAt: !856)
!856 = distinct !DILocation(line: 1030, scope: !572, inlinedAt: !577)
!857 = !DILocation(line: 1030, scope: !572, inlinedAt: !577)
!858 = distinct !DISubprogram(name: "gbsa_force_2_kernel!", linkageName: "julia_gbsa_force_2_kernel!_25091", scope: null, file: !5, line: 1013, type: !49, scopeLine: 1013, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !4, retainedNodes: !14)
!859 = !DILocation(line: 0, scope: !858)
!860 = !DILocation(line: 88, scope: !134, inlinedAt: !861)
!861 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !862)
!862 = distinct !DILocation(line: 1024, scope: !858, inlinedAt: !863)
!863 = distinct !DILocation(line: 0, scope: !858)
!864 = !DILocation(line: 40, scope: !142, inlinedAt: !865)
!865 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !866)
!866 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !867)
!867 = distinct !DILocation(line: 56, scope: !149, inlinedAt: !868)
!868 = distinct !DILocation(line: 78, scope: !151, inlinedAt: !869)
!869 = distinct !DILocation(line: 1025, scope: !858, inlinedAt: !863)
!870 = !DILocation(line: 86, scope: !155, inlinedAt: !871)
!871 = distinct !DILocation(line: 1042, scope: !155, inlinedAt: !869)
!872 = !DILocation(line: 40, scope: !142, inlinedAt: !873)
!873 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !874)
!874 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !875)
!875 = distinct !DILocation(line: 51, scope: !161, inlinedAt: !876)
!876 = distinct !DILocation(line: 85, scope: !163, inlinedAt: !869)
!877 = !DILocation(line: 523, scope: !166, inlinedAt: !878)
!878 = distinct !DILocation(line: 1040, scope: !134, inlinedAt: !869)
!879 = !DILocation(line: 88, scope: !134, inlinedAt: !880)
!880 = distinct !DILocation(line: 1042, scope: !134, inlinedAt: !869)
!881 = !DILocation(line: 40, scope: !142, inlinedAt: !882)
!882 = distinct !DILocation(line: 6, scope: !144, inlinedAt: !883)
!883 = distinct !DILocation(line: 6, scope: !147, inlinedAt: !884)
!884 = distinct !DILocation(line: 46, scope: !174, inlinedAt: !885)
!885 = distinct !DILocation(line: 92, scope: !176, inlinedAt: !869)
!886 = !DILocation(line: 87, scope: !179, inlinedAt: !884)
!887 = !DILocation(line: 523, scope: !166, inlinedAt: !888)
!888 = distinct !DILocation(line: 1040, scope: !179, inlinedAt: !869)
!889 = !DILocation(line: 87, scope: !179, inlinedAt: !890)
!890 = distinct !DILocation(line: 1042, scope: !179, inlinedAt: !869)
!891 = !DILocation(line: 488, scope: !185, inlinedAt: !892)
!892 = distinct !DILocation(line: 1027, scope: !858, inlinedAt: !863)
!893 = !DILocation(line: 1027, scope: !858, inlinedAt: !863)
!894 = !DILocation(line: 295, scope: !189, inlinedAt: !895)
!895 = distinct !DILocation(line: 308, scope: !191, inlinedAt: !896)
!896 = distinct !DILocation(line: 363, scope: !191, inlinedAt: !897)
!897 = distinct !DILocation(line: 320, scope: !195, inlinedAt: !898)
!898 = distinct !DILocation(line: 1028, scope: !858, inlinedAt: !863)
!899 = !DILocation(line: 86, scope: !155, inlinedAt: !900)
!900 = distinct !DILocation(line: 40, scope: !142, inlinedAt: !901)
!901 = distinct !DILocation(line: 9, scope: !200, inlinedAt: !902)
!902 = distinct !DILocation(line: 9, scope: !203, inlinedAt: !903)
!903 = distinct !DILocation(line: 85, scope: !205, inlinedAt: !904)
!904 = distinct !DILocation(line: 52, scope: !207, inlinedAt: !905)
!905 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !906)
!906 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !907)
!907 = distinct !DILocation(line: 1031, scope: !858, inlinedAt: !863)
!908 = !DILocation(line: 40, scope: !142, inlinedAt: !901)
!909 = !{!910}
!910 = distinct !{!910, !911, !"primal"}
!911 = distinct !{!911, !" diff: %.fca.0.extract47"}
!912 = !{!913}
!913 = distinct !{!913, !911, !"shadow_0"}
!914 = distinct !{}
!915 = distinct !{}
!916 = distinct !{}
!917 = distinct !{}
!918 = distinct !{}
!919 = !DILocation(line: 535, scope: !220, inlinedAt: !920)
!920 = distinct !DILocation(line: 292, scope: !223, inlinedAt: !921)
!921 = distinct !DILocation(line: 313, scope: !226, inlinedAt: !922)
!922 = distinct !DILocation(line: 311, scope: !228, inlinedAt: !907)
!923 = distinct !{}
!924 = !DILocation(line: 292, scope: !223, inlinedAt: !921)
!925 = !DILocation(line: 409, scope: !231, inlinedAt: !926)
!926 = distinct !DILocation(line: 293, scope: !223, inlinedAt: !921)
!927 = !DILocation(line: 408, scope: !234, inlinedAt: !926)
!928 = !DILocation(line: 535, scope: !220, inlinedAt: !926)
!929 = !DILocation(line: 293, scope: !223, inlinedAt: !921)
!930 = !DILocation(line: 409, scope: !231, inlinedAt: !931)
!931 = distinct !DILocation(line: 295, scope: !223, inlinedAt: !921)
!932 = !DILocation(line: 408, scope: !234, inlinedAt: !931)
!933 = !DILocation(line: 535, scope: !220, inlinedAt: !931)
!934 = !DILocation(line: 295, scope: !223, inlinedAt: !921)
!935 = !DILocation(line: 410, scope: !243, inlinedAt: !936)
!936 = distinct !DILocation(line: 189, scope: !245, inlinedAt: !937)
!937 = distinct !DILocation(line: 459, scope: !248, inlinedAt: !938)
!938 = distinct !DILocation(line: 266, scope: !251, inlinedAt: !939)
!939 = distinct !DILocation(line: 256, scope: !254, inlinedAt: !940)
!940 = distinct !DILocation(line: 255, scope: !256, inlinedAt: !941)
!941 = distinct !DILocation(line: 1032, scope: !858, inlinedAt: !863)
!942 = !DILocation(line: 408, scope: !234, inlinedAt: !938)
!943 = !DILocation(line: 223, scope: !260, inlinedAt: !938)
!944 = !DILocation(line: 535, scope: !220, inlinedAt: !945)
!945 = distinct !DILocation(line: 268, scope: !251, inlinedAt: !939)
!946 = !DILocation(line: 268, scope: !251, inlinedAt: !939)
!947 = !DILocation(line: 189, scope: !266, inlinedAt: !945)
!948 = !DILocation(line: 215, scope: !268, inlinedAt: !949)
!949 = distinct !DILocation(line: 228, scope: !270, inlinedAt: !950)
!950 = distinct !DILocation(line: 232, scope: !270, inlinedAt: !951)
!951 = distinct !DILocation(line: 248, scope: !251, inlinedAt: !952)
!952 = distinct !DILocation(line: 240, scope: !274, inlinedAt: !953)
!953 = distinct !DILocation(line: 269, scope: !251, inlinedAt: !939)
!954 = distinct !{}
!955 = !DILocation(line: 215, scope: !268, inlinedAt: !956)
!956 = distinct !DILocation(line: 228, scope: !270, inlinedAt: !957)
!957 = distinct !DILocation(line: 234, scope: !270, inlinedAt: !951)
!958 = distinct !{}
!959 = !DILocation(line: 575, scope: !280, inlinedAt: !960)
!960 = distinct !DILocation(line: 243, scope: !283, inlinedAt: !957)
!961 = distinct !{}
!962 = !DILocation(line: 533, scope: !286, inlinedAt: !963)
!963 = distinct !DILocation(line: 42, scope: !288, inlinedAt: !964)
!964 = distinct !DILocation(line: 250, scope: !251, inlinedAt: !952)
!965 = !DILocation(line: 250, scope: !251, inlinedAt: !952)
!966 = !DILocation(line: 215, scope: !268, inlinedAt: !967)
!967 = distinct !DILocation(line: 639, scope: !293, inlinedAt: !968)
!968 = distinct !DILocation(line: 639, scope: !293, inlinedAt: !969)
!969 = distinct !DILocation(line: 222, scope: !296, inlinedAt: !964)
!970 = !DILocation(line: 891, scope: !298, inlinedAt: !971)
!971 = distinct !DILocation(line: 235, scope: !270, inlinedAt: !951)
!972 = !DILocation(line: 411, scope: !302, inlinedAt: !973)
!973 = distinct !DILocation(line: 251, scope: !251, inlinedAt: !952)
!974 = !DILocation(line: 410, scope: !243, inlinedAt: !975)
!975 = distinct !DILocation(line: 189, scope: !245, inlinedAt: !976)
!976 = distinct !DILocation(line: 459, scope: !248, inlinedAt: !973)
!977 = !DILocation(line: 408, scope: !234, inlinedAt: !973)
!978 = !DILocation(line: 223, scope: !260, inlinedAt: !973)
!979 = !DILocation(line: 410, scope: !243, inlinedAt: !973)
!980 = !DILocation(line: 533, scope: !286, inlinedAt: !981)
!981 = distinct !DILocation(line: 42, scope: !288, inlinedAt: !982)
!982 = distinct !DILocation(line: 1034, scope: !858, inlinedAt: !863)
!983 = !DILocation(line: 1034, scope: !858, inlinedAt: !863)
!984 = !DILocation(line: 88, scope: !134, inlinedAt: !985)
!985 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !986)
!986 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !987)
!987 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !988)
!988 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !989)
!989 = distinct !DILocation(line: 1035, scope: !858, inlinedAt: !863)
!990 = !DILocation(line: 114, scope: !322, inlinedAt: !985)
!991 = !DILocation(line: 40, scope: !142, inlinedAt: !992)
!992 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !993)
!993 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !994)
!994 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !995)
!995 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !986)
!996 = !DILocation(line: 408, scope: !234, inlinedAt: !997)
!997 = distinct !DILocation(line: 1036, scope: !858, inlinedAt: !863)
!998 = !DILocation(line: 535, scope: !220, inlinedAt: !999)
!999 = distinct !DILocation(line: 1037, scope: !858, inlinedAt: !863)
!1000 = !DILocation(line: 1037, scope: !858, inlinedAt: !863)
!1001 = !DILocation(line: 409, scope: !231, inlinedAt: !1002)
!1002 = distinct !DILocation(line: 1038, scope: !858, inlinedAt: !863)
!1003 = !DILocation(line: 215, scope: !268, inlinedAt: !1002)
!1004 = !DILocation(line: 292, scope: !338, inlinedAt: !1002)
!1005 = !DILocation(line: 411, scope: !302, inlinedAt: !1006)
!1006 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !1002)
!1007 = !DILocation(line: 411, scope: !302, inlinedAt: !1008)
!1008 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !1009)
!1009 = distinct !DILocation(line: 1039, scope: !858, inlinedAt: !863)
!1010 = !DILocation(line: 411, scope: !302, inlinedAt: !1011)
!1011 = distinct !DILocation(line: 255, scope: !341, inlinedAt: !1012)
!1012 = distinct !DILocation(line: 1040, scope: !858, inlinedAt: !863)
!1013 = !DILocation(line: 410, scope: !243, inlinedAt: !1014)
!1014 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !1015)
!1015 = distinct !DILocation(line: 1041, scope: !858, inlinedAt: !863)
!1016 = !DILocation(line: 410, scope: !243, inlinedAt: !1017)
!1017 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !1018)
!1018 = distinct !DILocation(line: 1042, scope: !858, inlinedAt: !863)
!1019 = !DILocation(line: 410, scope: !243, inlinedAt: !1018)
!1020 = !DILocation(line: 408, scope: !234, inlinedAt: !1021)
!1021 = distinct !DILocation(line: 410, scope: !357, inlinedAt: !1018)
!1022 = !DILocation(line: 409, scope: !231, inlinedAt: !1018)
!1023 = !DILocation(line: 411, scope: !302, inlinedAt: !1024)
!1024 = distinct !DILocation(line: 413, scope: !362, inlinedAt: !1018)
!1025 = !DILocation(line: 411, scope: !302, inlinedAt: !1018)
!1026 = !DILocation(line: 105, scope: !365, inlinedAt: !1018)
!1027 = !DILocation(line: 408, scope: !234, inlinedAt: !1018)
!1028 = !DILocation(line: 40, scope: !142, inlinedAt: !1029)
!1029 = distinct !DILocation(line: 9, scope: !200, inlinedAt: !1030)
!1030 = distinct !DILocation(line: 9, scope: !203, inlinedAt: !1031)
!1031 = distinct !DILocation(line: 85, scope: !205, inlinedAt: !1032)
!1032 = distinct !DILocation(line: 91, scope: !372, inlinedAt: !1033)
!1033 = distinct !DILocation(line: 85, scope: !374, inlinedAt: !1034)
!1034 = distinct !DILocation(line: 164, scope: !213, inlinedAt: !1035)
!1035 = distinct !DILocation(line: 1043, scope: !858, inlinedAt: !863)
!1036 = !DILocation(line: 114, scope: !322, inlinedAt: !1037)
!1037 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !1038)
!1038 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !1039)
!1039 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !1040)
!1040 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !1035)
!1041 = !DILocation(line: 40, scope: !142, inlinedAt: !1042)
!1042 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !1043)
!1043 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !1044)
!1044 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !1045)
!1045 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !1038)
!1046 = !DILocation(line: 410, scope: !243, inlinedAt: !1047)
!1047 = distinct !DILocation(line: 319, scope: !137, inlinedAt: !1035)
!1048 = !DILocation(line: 410, scope: !243, inlinedAt: !1049)
!1049 = distinct !DILocation(line: 578, scope: !391, inlinedAt: !1035)
!1050 = !DILocation(line: 575, scope: !280, inlinedAt: !1051)
!1051 = distinct !DILocation(line: 510, scope: !395, inlinedAt: !1052)
!1052 = distinct !DILocation(line: 444, scope: !397, inlinedAt: !1053)
!1053 = distinct !DILocation(line: 457, scope: !397, inlinedAt: !1054)
!1054 = distinct !DILocation(line: 459, scope: !400, inlinedAt: !1055)
!1055 = distinct !DILocation(line: 274, scope: !402, inlinedAt: !1056)
!1056 = distinct !DILocation(line: 98, scope: !405, inlinedAt: !1057)
!1057 = distinct !DILocation(line: 203, scope: !408, inlinedAt: !1058)
!1058 = distinct !DILocation(line: 2933, scope: !410, inlinedAt: !1059)
!1059 = distinct !DILocation(line: 1333, scope: !412, inlinedAt: !1060)
!1060 = distinct !DILocation(line: 1327, scope: !414, inlinedAt: !1061)
!1061 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !1062)
!1062 = distinct !DILocation(line: 1044, scope: !858, inlinedAt: !863)
!1063 = !DILocation(line: 88, scope: !134, inlinedAt: !1064)
!1064 = distinct !DILocation(line: 2965, scope: !420, inlinedAt: !1065)
!1065 = distinct !DILocation(line: 2965, scope: !420, inlinedAt: !1066)
!1066 = distinct !DILocation(line: 2949, scope: !410, inlinedAt: !1058)
!1067 = !DILocation(line: 87, scope: !179, inlinedAt: !1064)
!1068 = !DILocation(line: 88, scope: !134, inlinedAt: !1069)
!1069 = distinct !DILocation(line: 39, scope: !316, inlinedAt: !1070)
!1070 = distinct !DILocation(line: 48, scope: !207, inlinedAt: !1071)
!1071 = distinct !DILocation(line: 156, scope: !210, inlinedAt: !1072)
!1072 = distinct !DILocation(line: 204, scope: !213, inlinedAt: !1060)
!1073 = !DILocation(line: 114, scope: !322, inlinedAt: !1069)
!1074 = !DILocation(line: 40, scope: !142, inlinedAt: !1075)
!1075 = distinct !DILocation(line: 149, scope: !200, inlinedAt: !1076)
!1076 = distinct !DILocation(line: 149, scope: !326, inlinedAt: !1077)
!1077 = distinct !DILocation(line: 300, scope: !200, inlinedAt: !1078)
!1078 = distinct !DILocation(line: 40, scope: !316, inlinedAt: !1070)
!1079 = !DILocation(line: 409, scope: !231, inlinedAt: !1062)
!1080 = !DILocation(line: 410, scope: !243, inlinedAt: !1081)
!1081 = distinct !DILocation(line: 578, scope: !391, inlinedAt: !1062)
!1082 = !DILocation(line: 406, scope: !231, inlinedAt: !1083)
!1083 = distinct !DILocation(line: 75, scope: !440, inlinedAt: !1084)
!1084 = distinct !DILocation(line: 40, scope: !443, inlinedAt: !1085)
!1085 = distinct !DILocation(line: 31, scope: !445, inlinedAt: !1086)
!1086 = distinct !DILocation(line: 8, scope: !447, inlinedAt: !1087)
!1087 = distinct !DILocation(line: 1045, scope: !858, inlinedAt: !863)
!1088 = !DILocation(line: 410, scope: !243, inlinedAt: !1089)
!1089 = distinct !DILocation(line: 22, scope: !451, inlinedAt: !1090)
!1090 = distinct !DILocation(line: 75, scope: !440, inlinedAt: !1091)
!1091 = distinct !DILocation(line: 40, scope: !443, inlinedAt: !1092)
!1092 = distinct !DILocation(line: 31, scope: !445, inlinedAt: !1093)
!1093 = distinct !DILocation(line: 22, scope: !456, inlinedAt: !1087)
!1094 = !DILocation(line: 1049, scope: !858, inlinedAt: !863)
!1095 = !DILocation(line: 88, scope: !134, inlinedAt: !1096)
!1096 = distinct !DILocation(line: 1247, scope: !490, inlinedAt: !1097)
!1097 = distinct !DILocation(line: 52, scope: !492, inlinedAt: !1098)
!1098 = distinct !DILocation(line: 103, scope: !467, inlinedAt: !1099)
!1099 = distinct !DILocation(line: 30, scope: !470, inlinedAt: !1100)
!1100 = distinct !DILocation(line: 1051, scope: !858, inlinedAt: !863)
!1101 = !DILocation(line: 114, scope: !322, inlinedAt: !1097)
!1102 = !DILocation(line: 40, scope: !142, inlinedAt: !1103)
!1103 = distinct !DILocation(line: 255, scope: !497, inlinedAt: !1104)
!1104 = distinct !DILocation(line: 255, scope: !500, inlinedAt: !1105)
!1105 = distinct !DILocation(line: 359, scope: !502, inlinedAt: !1106)
!1106 = distinct !DILocation(line: 20, scope: !504, inlinedAt: !1107)
!1107 = distinct !DILocation(line: 33, scope: !470, inlinedAt: !1100)
!1108 = !DILocation(line: 709, scope: !484, inlinedAt: !1109)
!1109 = distinct !DILocation(line: 1326, scope: !414, inlinedAt: !1110)
!1110 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !1111)
!1111 = distinct !DILocation(line: 102, scope: !467, inlinedAt: !1112)
!1112 = distinct !DILocation(line: 30, scope: !470, inlinedAt: !1113)
!1113 = distinct !DILocation(line: 1052, scope: !858, inlinedAt: !863)
!1114 = !DILocation(line: 406, scope: !231, inlinedAt: !1113)
!1115 = !DILocation(line: 88, scope: !134, inlinedAt: !1116)
!1116 = distinct !DILocation(line: 1247, scope: !490, inlinedAt: !1117)
!1117 = distinct !DILocation(line: 52, scope: !492, inlinedAt: !1118)
!1118 = distinct !DILocation(line: 103, scope: !467, inlinedAt: !1112)
!1119 = !DILocation(line: 114, scope: !322, inlinedAt: !1117)
!1120 = !DILocation(line: 40, scope: !142, inlinedAt: !1121)
!1121 = distinct !DILocation(line: 255, scope: !497, inlinedAt: !1122)
!1122 = distinct !DILocation(line: 255, scope: !500, inlinedAt: !1123)
!1123 = distinct !DILocation(line: 359, scope: !502, inlinedAt: !1124)
!1124 = distinct !DILocation(line: 20, scope: !504, inlinedAt: !1125)
!1125 = distinct !DILocation(line: 33, scope: !470, inlinedAt: !1113)
!1126 = !DILocation(line: 488, scope: !185, inlinedAt: !1127)
!1127 = distinct !DILocation(line: 768, scope: !534, inlinedAt: !1128)
!1128 = distinct !DILocation(line: 735, scope: !536, inlinedAt: !1129)
!1129 = distinct !DILocation(line: 688, scope: !484, inlinedAt: !1130)
!1130 = distinct !DILocation(line: 709, scope: !484, inlinedAt: !1131)
!1131 = distinct !DILocation(line: 1326, scope: !414, inlinedAt: !1132)
!1132 = distinct !DILocation(line: 1294, scope: !416, inlinedAt: !1133)
!1133 = distinct !DILocation(line: 102, scope: !467, inlinedAt: !1099)
!1134 = !DILocation(line: 709, scope: !484, inlinedAt: !1131)
!1135 = !DILocation(line: 1058, scope: !858, inlinedAt: !863)
!1136 = !DILocation(line: 83, scope: !541, inlinedAt: !1137)
!1137 = distinct !DILocation(line: 369, scope: !543, inlinedAt: !1138)
!1138 = distinct !DILocation(line: 364, scope: !191, inlinedAt: !897)
!1139 = !DILocation(line: 88, scope: !134, inlinedAt: !1138)
!1140 = !DILocation(line: 499, scope: !547, inlinedAt: !1141)
!1141 = distinct !DILocation(line: 269, scope: !549, inlinedAt: !1138)
!1142 = !DILocation(line: 38, scope: !551, inlinedAt: !1138)
!1143 = !DILocation(line: 711, scope: !554, inlinedAt: !1144)
!1144 = distinct !DILocation(line: 784, scope: !557, inlinedAt: !1145)
!1145 = distinct !DILocation(line: 7, scope: !559, inlinedAt: !1146)
!1146 = distinct !DILocation(line: 520, scope: !166, inlinedAt: !1147)
!1147 = distinct !DILocation(line: 1040, scope: !179, inlinedAt: !1138)
!1148 = !DILocation(line: 87, scope: !179, inlinedAt: !1149)
!1149 = distinct !DILocation(line: 1042, scope: !179, inlinedAt: !1138)
!1150 = !DILocation(line: 296, scope: !166, inlinedAt: !1151)
!1151 = distinct !DILocation(line: 1029, scope: !858, inlinedAt: !863)
!1152 = !DILocation(line: 86, scope: !155, inlinedAt: !1151)
!1153 = !DILocation(line: 87, scope: !179, inlinedAt: !1151)
!1154 = !DILocation(line: 499, scope: !547, inlinedAt: !1155)
!1155 = distinct !DILocation(line: 269, scope: !549, inlinedAt: !1156)
!1156 = distinct !DILocation(line: 1030, scope: !858, inlinedAt: !863)
!1157 = !DILocation(line: 1030, scope: !858, inlinedAt: !863)

No augmented forward pass found for __nvvm_reflect
declare i32 @__nvvm_reflect(i8*) local_unnamed_addr #7



Stacktrace:
  [1] julia_error(cstr::Cstring, val::Ptr{LLVM.API.LLVMOpaqueValue}, errtype::Enzyme.API.ErrorType, data::Ptr{Nothing})
    @ Enzyme.Compiler ~/.julia/dev/Enzyme/src/compiler.jl:5242
  [2] EnzymeCreatePrimalAndGradient(logic::Enzyme.Logic, todiff::LLVM.Function, retType::Enzyme.API.CDIFFE_TYPE, constant_args::Vector{Enzyme.API.CDIFFE_TYPE}, TA::Enzyme.TypeAnalysis, returnValue::Bool, dretUsed::Bool, mode::Enzyme.API.CDerivativeMode, width::Int64, additionalArg::Ptr{Nothing}, forceAnonymousTape::Bool, typeInfo::Enzyme.FnTypeInfo, uncacheable_args::Vector{Bool}, augmented::Ptr{Nothing}, atomicAdd::Bool)
    @ Enzyme.API ~/.julia/dev/Enzyme/src/api.jl:124
  [3] enzyme!(job::GPUCompiler.CompilerJob{Enzyme.Compiler.EnzymeTarget, Enzyme.Compiler.EnzymeCompilerParams}, mod::LLVM.Module, primalf::LLVM.Function, TT::Type, mode::Enzyme.API.CDerivativeMode, width::Int64, parallel::Bool, actualRetType::Type, wrap::Bool, modifiedBetween::NTuple{13, Bool}, returnPrimal::Bool, jlrules::Vector{String}, expectedTapeType::Type)
    @ Enzyme.Compiler ~/.julia/dev/Enzyme/src/compiler.jl:6710
  [4] codegen(output::Symbol, job::GPUCompiler.CompilerJob{Enzyme.Compiler.EnzymeTarget, Enzyme.Compiler.EnzymeCompilerParams}; libraries::Bool, deferred_codegen::Bool, optimize::Bool, ctx::LLVM.ThreadSafeContext, strip::Bool, validate::Bool, only_entry::Bool, parent_job::GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams})
    @ Enzyme.Compiler ~/.julia/dev/Enzyme/src/compiler.jl:7951
  [5] codegen
    @ ~/.julia/dev/Enzyme/src/compiler.jl:7577 [inlined]
  [6] (::GPUCompiler.var"#116#119"{LLVM.ThreadSafeContext, GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, GPUCompiler.CompilerJob{Enzyme.Compiler.EnzymeTarget, Enzyme.Compiler.EnzymeCompilerParams}})()
    @ GPUCompiler ~/.julia/packages/GPUCompiler/YU9ZW/src/driver.jl:289
  [7] get!(default::GPUCompiler.var"#116#119"{LLVM.ThreadSafeContext, GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, GPUCompiler.CompilerJob{Enzyme.Compiler.EnzymeTarget, Enzyme.Compiler.EnzymeCompilerParams}}, h::Dict{GPUCompiler.CompilerJob, String}, key::GPUCompiler.CompilerJob{Enzyme.Compiler.EnzymeTarget, Enzyme.Compiler.EnzymeCompilerParams})
    @ Base ./dict.jl:468
  [8] macro expansion
    @ ~/.julia/packages/GPUCompiler/YU9ZW/src/driver.jl:288 [inlined]
  [9] emit_llvm(job::GPUCompiler.CompilerJob; libraries::Bool, deferred_codegen::Bool, optimize::Bool, cleanup::Bool, only_entry::Bool, validate::Bool, ctx::LLVM.ThreadSafeContext)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/YU9ZW/src/utils.jl:83
 [10] emit_llvm
    @ ~/.julia/packages/GPUCompiler/YU9ZW/src/utils.jl:77 [inlined]
 [11] codegen(output::Symbol, job::GPUCompiler.CompilerJob; libraries::Bool, deferred_codegen::Bool, optimize::Bool, cleanup::Bool, strip::Bool, validate::Bool, only_entry::Bool, parent_job::Nothing, ctx::LLVM.ThreadSafeContext)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/YU9ZW/src/driver.jl:120
 [12] compile(target::Symbol, job::GPUCompiler.CompilerJob; libraries::Bool, deferred_codegen::Bool, optimize::Bool, cleanup::Bool, strip::Bool, validate::Bool, only_entry::Bool, ctx::LLVM.ThreadSafeContext)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/YU9ZW/src/driver.jl:90
 [13] compile
    @ ~/.julia/packages/GPUCompiler/YU9ZW/src/driver.jl:81 [inlined]
 [14] compile(job::GPUCompiler.CompilerJob, ctx::LLVM.ThreadSafeContext)
    @ CUDA ~/.julia/dev/CUDA/src/compiler/compilation.jl:105
 [15] #203
    @ ~/.julia/dev/CUDA/src/compiler/compilation.jl:100 [inlined]
 [16] LLVM.ThreadSafeContext(f::CUDA.var"#203#204"{GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}})
    @ LLVM ~/.julia/packages/LLVM/TLGyi/src/executionengine/ts_module.jl:14
 [17] JuliaContext(f::CUDA.var"#203#204"{GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}})
    @ GPUCompiler ~/.julia/packages/GPUCompiler/YU9ZW/src/driver.jl:35
 [18] compile
    @ ~/.julia/dev/CUDA/src/compiler/compilation.jl:99 [inlined]
 [19] actual_compilation(cache::Dict{UInt64, Any}, key::UInt64, cfg::GPUCompiler.CompilerConfig{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, ft::Type, tt::Type, compiler::typeof(CUDA.compile), linker::typeof(CUDA.link))
    @ GPUCompiler ~/.julia/packages/GPUCompiler/YU9ZW/src/cache.jl:53
 [20] cached_compilation(cache::Dict{UInt64, Any}, cfg::GPUCompiler.CompilerConfig{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams}, ft::Type, tt::Type, compiler::Function, linker::Function)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/YU9ZW/src/cache.jl:37
 [21] macro expansion
    @ ~/.julia/dev/CUDA/src/compiler/execution.jl:310 [inlined]
 [22] macro expansion
    @ ./lock.jl:267 [inlined]
 [23] cufunction(f::typeof(Molly.grad_gbsa_force_2_kernel!), tt::Type{Tuple{CuDeviceMatrix{Float32, 1}, CuDeviceMatrix{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{SVector{3, Float32}, 1}, CuDeviceVector{SVector{3, Float32}, 1}, CubicBoundary{Float32}, Float32, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceVector{Float32, 1}, CuDeviceMatrix{Float32, 1}, CuDeviceMatrix{Float32, 1}, Val{3}, Val{}}}; kwargs::Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
    @ CUDA ~/.julia/dev/CUDA/src/compiler/execution.jl:306
 [24] cufunction
    @ ~/.julia/dev/CUDA/src/compiler/execution.jl:303 [inlined]
 [25] macro expansion
    @ ~/.julia/dev/CUDA/src/compiler/execution.jl:104 [inlined]
 [26] macro expansion
    @ ~/.julia/dev/CUDA/src/utilities.jl:25 [inlined]
 [27] (::Molly.var"#gbsa_force_2_gpu_pullback#451"{3, Float32, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Unitful.FreeUnits{(), NoDims, nothing}})(d_fs_mat::CuArray{Float32, 2, CUDA.Mem.DeviceBuffer})
    @ Molly ~/.julia/dev/Molly/src/chain_rules.jl:903
 [28] ZBack
    @ ~/.julia/dev/Zygote/src/compiler/chainrules.jl:211 [inlined]
 [29] Pullback
    @ ~/.julia/dev/Molly/src/interactions/implicit_solvent.jl:914 [inlined]
 [30] (::Zygote.Pullback{Tuple{typeof(Molly.forces_gbsa), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}, Tuple{Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:kappa, Zygote.Context{false}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, Float32}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:factor_solvent, Zygote.Context{false}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, Float32}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:factor_solute, Zygote.Context{false}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, Float32}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:force_units, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}}}, Zygote.Pullback{Tuple{typeof(*), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.var"#4153#back#1387"{Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcasted), typeof(*), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.var"#2863#back#683"{Zygote.var"#map_back#677"{typeof(Base.Broadcast.broadcastable), 1, Tuple{Tuple{}}, Tuple{Val{0}}, Tuple{}}}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcastable), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}, Tuple{}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.var"#2156#back#289"{Zygote.var"#287#288"{Tuple{NTuple{4, Nothing}, Tuple{}}, Molly.var"#983#back#694"{Zygote.var"#bc_fwd_back#1401"{CuArray{ForwardDiff.Dual{Nothing, Float32, 2}, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Base.RefValue{Unitful.FreeUnits{(), NoDims, nothing}}}, Val{2}}}}}, Zygote.var"#2156#back#289"{Zygote.var"#287#288"{Tuple{Tuple{Nothing, Nothing}, Tuple{}}, Zygote.var"#combine_styles_pullback#1169"{Tuple{Nothing, Nothing, Nothing}}}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcastable), Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.Pullback{Tuple{Type{Ref}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.Pullback{Tuple{Type{Base.RefValue}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.Pullback{Tuple{Type{Base.RefValue{Unitful.FreeUnits{(), NoDims, nothing}}}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.ZBack{ChainRules.var"#fieldtype_pullback#420"}, Zygote.Pullback{Tuple{typeof(convert), Type{Unitful.FreeUnits{(), NoDims, nothing}}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{}}, Zygote.var"#2197#back#309"{Zygote.Jnew{Base.RefValue{Unitful.FreeUnits{(), NoDims, nothing}}, Base.RefValue{Any}, false}}}}}}}}}}}}}}}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.materialize), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}, Tuple{}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:dist_cutoff, Zygote.Context{false}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, Float32}}, Zygote.Pullback{Tuple{typeof(unit), Type{Float32}}, Tuple{Zygote.var"#1973#back#190"{Zygote.var"#186#189"{Zygote.Context{false}, GlobalRef, Unitful.FreeUnits{(), NoDims, nothing}}}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:coords, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:force_units, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}}}, Zygote.Pullback{Tuple{typeof(*), CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.var"#4153#back#1387"{Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcasted), typeof(*), CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.var"#2863#back#683"{Zygote.var"#map_back#677"{typeof(Base.Broadcast.broadcastable), 1, Tuple{Tuple{}}, Tuple{Val{0}}, Tuple{}}}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcastable), CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}}, Tuple{}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.var"#2156#back#289"{Zygote.var"#287#288"{Tuple{NTuple{4, Nothing}, Tuple{}}, Molly.var"#983#back#694"{Molly.var"#bc_fwd_back#474"{3, Float32, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, Base.RefValue{Unitful.FreeUnits{(), NoDims, nothing}}, CuArray{SVector{3, ForwardDiff.Dual{Nothing, Float32, 4}}, 1, CUDA.Mem.DeviceBuffer}}}}}, Zygote.var"#2156#back#289"{Zygote.var"#287#288"{Tuple{Tuple{Nothing, Nothing}, Tuple{}}, Zygote.var"#combine_styles_pullback#1169"{Tuple{Nothing, Nothing, Nothing}}}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcastable), Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.Pullback{Tuple{Type{Ref}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.Pullback{Tuple{Type{Base.RefValue}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.Pullback{Tuple{Type{Base.RefValue{Unitful.FreeUnits{(), NoDims, nothing}}}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{Zygote.ZBack{ChainRules.var"#fieldtype_pullback#420"}, Zygote.Pullback{Tuple{typeof(convert), Type{Unitful.FreeUnits{(), NoDims, nothing}}, Unitful.FreeUnits{(), NoDims, nothing}}, Tuple{}}, Zygote.var"#2197#back#309"{Zygote.Jnew{Base.RefValue{Unitful.FreeUnits{(), NoDims, nothing}}, Base.RefValue{Any}, false}}}}}}}}}}}}}}}, Zygote.var"#back#242"{Zygote.var"#2016#back#209"{Zygote.var"#back#207"{2, 2, Zygote.Context{false}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:offset_radii, Zygote.Context{false}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}, Zygote.ZBack{Molly.var"#gbsa_force_2_gpu_pullback#451"{3, Float32, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Unitful.FreeUnits{(), NoDims, nothing}}}, Zygote.var"#2016#back#209"{Zygote.var"#back#207"{2, 1, Zygote.Context{false}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:scaled_offset_radii, Zygote.Context{false}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}, Zygote.Pullback{Tuple{typeof(eltype), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}, Tuple{Zygote.ZBack{ChainRules.var"#eltype_pullback#384"}, Zygote.ZBack{ChainRules.var"#typeof_pullback#45"}}}, Zygote.var"#2016#back#209"{Zygote.var"#back#207"{2, 2, Zygote.Context{false}, Int64}}, Zygote.ZBack{Molly.var"#gbsa_force_1_gpu_pullback#450"{3, Float32, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, Float32, Float32, Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Unitful.FreeUnits{(), NoDims, nothing}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:dist_cutoff, Zygote.Context{false}, ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, Float32}}, Zygote.var"#2016#back#209"{Zygote.var"#back#207"{2, 1, Zygote.Context{false}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}}}, Zygote.var"#back#241"{Zygote.var"#2016#back#209"{Zygote.var"#back#207"{2, 1, Zygote.Context{false}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}}}}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.materialize), CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}}, Tuple{}}, Zygote.ZBack{Molly.var"#reinterpret_pullback#425"{3, Float32}}, Zygote.Pullback{Tuple{typeof(vec), CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}}, Tuple{Zygote.var"#2781#back#620"{Zygote.var"#614#618"{CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, Tuple{Int64}}}, Zygote.ZBack{ChainRules.var"#length_pullback#746"}}}, Zygote.var"#3752#back#1184"{Zygote.var"#1178#1182"{Tuple{CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:coords, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Zygote.var"#1909#back#157"{Zygote.var"#153#156"}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:boundary, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CubicBoundary{Float32}}}, Zygote.var"#3752#back#1184"{Zygote.var"#1178#1182"{Tuple{CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:boundary, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CubicBoundary{Float32}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:force_units, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}}}}})(Δ::CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [31] Pullback
    @ ~/.julia/dev/Molly/src/interactions/implicit_solvent.jl:1073 [inlined]
 [32] (::Zygote.Pullback{Tuple{Molly.var"##forces#130", Int64, typeof(forces), ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, NeighborList{CuArray{Tuple{Int32, Int32, Bool}, 1, CUDA.Mem.DeviceBuffer}}}, Any})(Δ::CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [33] Pullback
    @ ~/.julia/dev/Molly/src/interactions/implicit_solvent.jl:1061 [inlined]
 [34] (::Zygote.Pullback{Tuple{typeof(Core.kwcall), NamedTuple{(:n_threads,), Tuple{Int64}}, typeof(forces), ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, NeighborList{CuArray{Tuple{Int32, Int32, Bool}, 1, CUDA.Mem.DeviceBuffer}}}, Any})(Δ::CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [35] Pullback
    @ ~/.julia/dev/Molly/src/force.jl:352 [inlined]
 [36] (::Zygote.Pullback{Tuple{Molly.var"##forces#86", Int64, typeof(forces), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, NeighborList{CuArray{Tuple{Int32, Int32, Bool}, 1, CUDA.Mem.DeviceBuffer}}}, Any})(Δ::CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [37] Pullback
    @ ~/.julia/dev/Molly/src/force.jl:320 [inlined]
 [38] (::Zygote.Pullback{Tuple{typeof(Core.kwcall), NamedTuple{(:n_threads,), Tuple{Int64}}, typeof(forces), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, NeighborList{CuArray{Tuple{Int32, Int32, Bool}, 1, CUDA.Mem.DeviceBuffer}}}, Any})(Δ::CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [39] Pullback
    @ ~/.julia/dev/Molly/src/force.jl:41 [inlined]
 [40] (::Zygote.Pullback{Tuple{Molly.var"##accelerations#68", Int64, typeof(accelerations), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, NeighborList{CuArray{Tuple{Int32, Int32, Bool}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{Zygote.Pullback{Tuple{typeof(Base.Broadcast.materialize), CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}}, Tuple{}}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcasted), typeof(/), CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}, Tuple{Zygote.var"#2863#back#683"{Zygote.var"#map_back#677"{typeof(Base.Broadcast.broadcastable), 1, Tuple{Tuple{}}, Tuple{Val{0}}, Tuple{}}}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcastable), CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}}, Tuple{}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.var"#2156#back#289"{Zygote.var"#287#288"{Tuple{NTuple{4, Nothing}, Tuple{}}, Molly.var"#1015#back#706"{Molly.var"#bc_fwd_back#474"{3, Float32, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, ForwardDiff.Dual{Nothing, Float32, 4}}, 1, CUDA.Mem.DeviceBuffer}}}}}, Zygote.var"#2156#back#289"{Zygote.var"#287#288"{Tuple{Tuple{Nothing, Nothing}, Tuple{}}, Zygote.var"#combine_styles_pullback#1169"{Tuple{Nothing, Nothing, Nothing}}}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.Pullback{Tuple{typeof(Base.Broadcast.broadcastable), CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}, Tuple{}}}}, Zygote.Pullback{Tuple{typeof(masses), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}}, Tuple{Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:masses, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}}}}, Zygote.Pullback{Tuple{Type{NamedTuple{(:n_threads,)}}, Tuple{Int64}}, Tuple{Zygote.Pullback{Tuple{Type{NamedTuple{(:n_threads,), Tuple{Int64}}}, Tuple{Int64}}, Tuple{Zygote.var"#2207#back#311"{Zygote.Jnew{NamedTuple{(:n_threads,), Tuple{Int64}}, Nothing, true}}}}}}, Zygote.Pullback{Tuple{typeof(Core.kwcall), NamedTuple{(:n_threads,), Tuple{Int64}}, typeof(forces), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, NeighborList{CuArray{Tuple{Int32, Int32, Bool}, 1, CUDA.Mem.DeviceBuffer}}}, Any}, Zygote.var"#2000#back#200"{typeof(identity)}}})(Δ::CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [41] Pullback
    @ ~/.julia/dev/Molly/src/force.jl:40 [inlined]
 [42] (::Zygote.Pullback{Tuple{typeof(Core.kwcall), NamedTuple{(:n_threads,), Tuple{Int64}}, typeof(accelerations), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, NeighborList{CuArray{Tuple{Int32, Int32, Bool}, 1, CUDA.Mem.DeviceBuffer}}}, Any})(Δ::CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [43] Pullback
    @ ~/.julia/dev/Molly/src/simulators.jl:332 [inlined]
 [44] (::Zygote.Pullback{Tuple{Molly.var"##simulate!#167", Int64, Xoshiro, typeof(simulate!), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Langevin{Float32, Float32, Float32, Float32}, Int64}, Any})(Δ::Nothing)
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [45] Pullback
    @ ~/.julia/dev/Molly/src/simulators.jl:321 [inlined]
 [46] (::Zygote.Pullback{Tuple{typeof(Core.kwcall), NamedTuple{(:n_threads, :rng), Tuple{Int64, Xoshiro}}, typeof(simulate!), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Langevin{Float32, Float32, Float32, Float32}, Int64}, Any})(Δ::Nothing)
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [47] Pullback
    @ ~/dms/training/grads_14.jl:189 [inlined]
 [48] (::Zygote.Pullback{Tuple{typeof(run_sim_chunk), CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{AtomData}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Int64, UInt32}, Tuple{Zygote.ZBack{ChainRules.var"#AbstractRNG_pullback#2155"{Tuple{UInt32}}}, Zygote.var"#1973#back#190"{Zygote.var"#186#189"{Zygote.Context{false}, GlobalRef, Int64}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:velocities, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Zygote.var"#1973#back#190"{Zygote.var"#186#189"{Zygote.Context{false}, GlobalRef, Langevin{Float32, Float32, Float32, Float32}}}, Zygote.Pullback{Tuple{Type{NamedTuple{(:n_threads, :rng)}}, Tuple{Int64, Xoshiro}}, Tuple{Zygote.Pullback{Tuple{Type{NamedTuple{(:n_threads, :rng), Tuple{Int64, Xoshiro}}}, Tuple{Int64, Xoshiro}}, Tuple{Zygote.var"#2207#back#311"{Zygote.Jnew{NamedTuple{(:n_threads, :rng), Tuple{Int64, Xoshiro}}, Nothing, true}}}}}}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:coords, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Zygote.Pullback{Tuple{typeof(Core.kwcall), NamedTuple{(:atoms, :pairwise_inters, :specific_inter_lists, :general_inters, :coords, :velocities, :boundary, :neighbor_finder, :force_units, :energy_units), Tuple{CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Type{System}}, Any}, Zygote.Pullback{Tuple{Type{NamedTuple{(:atoms, :pairwise_inters, :specific_inter_lists, :general_inters, :coords, :velocities, :boundary, :neighbor_finder, :force_units, :energy_units)}}, Tuple{CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{Zygote.Pullback{Tuple{Type{NamedTuple{(:atoms, :pairwise_inters, :specific_inter_lists, :general_inters, :coords, :velocities, :boundary, :neighbor_finder, :force_units, :energy_units), Tuple{CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}}, Tuple{CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{Zygote.var"#2207#back#311"{Zygote.Jnew{NamedTuple{(:atoms, :pairwise_inters, :specific_inter_lists, :general_inters, :coords, :velocities, :boundary, :neighbor_finder, :force_units, :energy_units), Tuple{CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Nothing, true}}}}}}, Zygote.Pullback{Tuple{typeof(Core.kwcall), NamedTuple{(:n_threads, :rng), Tuple{Int64, Xoshiro}}, typeof(simulate!), System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{Any}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Langevin{Float32, Float32, Float32, Float32}, Int64}, Any}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:neighbor_finder, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{AtomData}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}}}, Zygote.var"#2000#back#200"{typeof(identity)}, Zygote.var"#2167#back#299"{Zygote.var"#back#298"{:boundary, Zygote.Context{false}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{AtomData}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, CubicBoundary{Float32}}}, Zygote.var"#2000#back#200"{typeof(identity)}}})(Δ::Tuple{CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, Nothing})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [49] (::Zygote.var"#pullback_checkpointed#120"{Zygote.Context{false}, typeof(run_sim_chunk), Tuple{CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{AtomData}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Int64, UInt32}})(Δy::Tuple{CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, Nothing})
    @ Zygote ~/.julia/dev/Zygote/src/lib/grad.jl:25
 [50] Pullback
    @ ~/dms/training/grads_14.jl:238 [inlined]
 [51] (::Zygote.Pullback{Tuple{typeof(loss), Dict{String, Float32}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{AtomData}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Int64}, Any})(Δ::Float32)
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface2.jl:0
 [52] (::Zygote.var"#75#76"{Zygote.Pullback{Tuple{typeof(loss), Dict{String, Float32}, System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{AtomData}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, Int64}, Any}})(Δ::Float32)
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface.jl:45
 [53] gradient(::Function, ::Dict{String, Float32}, ::Vararg{Any})
    @ Zygote ~/.julia/dev/Zygote/src/compiler/interface.jl:97
 [54] print_grads(io::Base.DevNull, params_dic::Dict{String, Float32}, s_ref::System{3, true, Float32, CuArray{Atom{Float32, Float32, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}, Vector{AtomData}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{LennardJones{false, DistanceCutoff{Float32, Float32, Float32}, Float32, Int64, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}, Coulomb{DistanceCutoff{Float32, Float32, Float32}, Float32, Float32, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}}}, Tuple{InteractionList2Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicBond{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList3Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{HarmonicAngle{Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}, InteractionList4Atoms{CuArray{Int32, 1, CUDA.Mem.DeviceBuffer}, CuArray{PeriodicTorsion{6, Float32, Float32}, 1, CUDA.Mem.DeviceBuffer}}}, Tuple{ImplicitSolventGBN2{Float32, Float32, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Float32, Float32, Float32, CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, CuArray{Float32, 2, CUDA.Mem.DeviceBuffer}, SubArray{Float32, 2, CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}, Tuple{CuArray{Int64, 2, CUDA.Mem.DeviceBuffer}}, false}}}, Tuple{}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CuArray{SVector{3, Float32}, 1, CUDA.Mem.DeviceBuffer}, CubicBoundary{Float32}, DistanceNeighborFinder{CuArray{Bool, 2, CUDA.Mem.DeviceBuffer}, Float32}, Tuple{}, Unitful.FreeUnits{(), NoDims, nothing}, Unitful.FreeUnits{(), NoDims, nothing}, Float32}, n_steps::Int64)
    @ Main ~/dms/training/grads_14.jl:318