unsupported call through a literal pointer (call to log1pf) on Julia 1.6.5

[guyvdbroeck]

In an attempt to circumvent #1350 I run into another issue with CUDA 3.8.0 on Julia 1.6.5 which does not occur on 1.7:

julia> using CUDA                                                                                                              
[ Info: Precompiling CUDA [052768ef-5323-5732-b1bb-66c8b64840ba]                                                                                                                                                                                              
WARNING: could not import Printf.ini_hex into BFloat16s
WARNING: could not import Printf.ini_HEX into BFloat16s                                                                        
                                                                                                                               
julia> f(x) = log1p(x)
f (generic function with 1 method)
                                                               
julia> g(x) = @fastmath log1p(x)                           
g (generic function with 1 method)                             
                                                               
julia> map(f, cu([0.1,0.2]))                                                                                                   
2-element CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}:
 0.09531018                                                    
 0.18232156 

julia> map(g, cu([0.1,0.2]))
ERROR: InvalidIRError: compiling kernel broadcast_kernel(CUDA.CuKernelContext, CuDeviceVector{Float32, 1}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}}, typeof(g), Tuple{Base.Broadcast.Extruded{CuDeviceVector{Float32, 1}, Tuple{Bool}, Tup
le{Int64}}}}, Int64) resulted in invalid LLVM IR
Reason: unsupported call through a literal pointer (call to log1pf)
Stacktrace:
 [1] log1p_fast
   @ ./fastmath.jl:295
 [2] g
   @ ./REPL[5]:1
 [3] _broadcast_getindex_evalf
   @ ./broadcast.jl:648
 [4] _broadcast_getindex
   @ ./broadcast.jl:621
 [5] getindex
   @ ./broadcast.jl:575
 [6] broadcast_kernel
   @ ~/.julia/packages/GPUArrays/umZob/src/host/broadcast.jl:59
Stacktrace:
  [1] check_ir(job::GPUCompiler.CompilerJob{GPUCompiler.PTXCompilerTarget, CUDA.CUDACompilerParams, GPUCompiler.FunctionSpec{GPUArrays.var"#broadcast_kernel#17", Tuple{CUDA.CuKernelContext, CuDeviceVector{Float32, 1}, Base.Broadcast.Broadcasted{Nothing, 
Tuple{Base.OneTo{Int64}}, typeof(g), Tuple{Base.Broadcast.Extruded{CuDeviceVector{Float32, 1}, Tuple{Bool}, Tuple{Int64}}}}, Int64}}}, args::LLVM.Module)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/1Ajz2/src/validation.jl:111
  [2] macro expansion
    @ ~/.julia/packages/GPUCompiler/1Ajz2/src/driver.jl:326 [inlined]
  [3] macro expansion
    @ ~/.julia/packages/TimerOutputs/5tW2E/src/TimerOutput.jl:252 [inlined]
  [4] macro expansion
    @ ~/.julia/packages/GPUCompiler/1Ajz2/src/driver.jl:324 [inlined]
  [5] emit_asm(job::GPUCompiler.CompilerJob, ir::LLVM.Module; strip::Bool, validate::Bool, format::LLVM.API.LLVMCodeGenFileType)
    @ GPUCompiler ~/.julia/packages/GPUCompiler/1Ajz2/src/utils.jl:64
  [6] cufunction_compile(job::GPUCompiler.CompilerJob)
    @ CUDA ~/.julia/packages/CUDA/bki2w/src/compiler/execution.jl:326
  [7] cached_compilation(cache::Dict{UInt64, Any}, job::GPUCompiler.CompilerJob, compiler::typeof(CUDA.cufunction_compile), linker::typeof(CUDA.cufunction_link))
    @ GPUCompiler ~/.julia/packages/GPUCompiler/1Ajz2/src/cache.jl:90
  [8] cufunction(f::GPUArrays.var"#broadcast_kernel#17", tt::Type{Tuple{CUDA.CuKernelContext, CuDeviceVector{Float32, 1}, Base.Broadcast.Broadcasted{Nothing, Tuple{Base.OneTo{Int64}}, typeof(g), Tuple{Base.Broadcast.Extruded{CuDeviceVector{Float32, 1}, T
uple{Bool}, Tuple{Int64}}}}, Int64}}; name::Nothing, kwargs::Base.Iterators.Pairs{Union{}, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
    @ CUDA ~/.julia/packages/CUDA/bki2w/src/compiler/execution.jl:297
  [9] cufunction
    @ ~/.julia/packages/CUDA/bki2w/src/compiler/execution.jl:291 [inlined]
 [10] macro expansion
    @ ~/.julia/packages/CUDA/bki2w/src/compiler/execution.jl:102 [inlined]
 [11] #launch_heuristic#268
    @ ~/.julia/packages/CUDA/bki2w/src/gpuarrays.jl:17 [inlined]
 [12] copyto!
    @ ~/.julia/packages/GPUArrays/umZob/src/host/broadcast.jl:65 [inlined]
 [13] copyto!
    @ ./broadcast.jl:936 [inlined]
 [14] copy
    @ ~/.julia/packages/GPUArrays/umZob/src/host/broadcast.jl:47 [inlined]
 [15] materialize(bc::Base.Broadcast.Broadcasted{CUDA.CuArrayStyle{1}, Nothing, typeof(g), Tuple{CuArray{Float32, 1, CUDA.Mem.DeviceBuffer}}})
    @ Base.Broadcast ./broadcast.jl:883
 [16] map(::Function, ::CuArray{Float32, 1, CUDA.Mem.DeviceBuffer})
    @ GPUArrays ~/.julia/packages/GPUArrays/umZob/src/host/broadcast.jl:90
 [17] top-level scope
    @ REPL[7]:1
 [18] top-level scope
    @ ~/.julia/packages/CUDA/bki2w/src/initialization.jl:52