Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[AArch64] Fix scheduling model issue #96394 #97047

Merged
merged 3 commits into from
Jul 1, 2024
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions llvm/lib/Target/AArch64/AArch64SchedPredNeoverse.td
Original file line number Diff line number Diff line change
Expand Up @@ -60,8 +60,9 @@ def NeoverseZeroMove : MCSchedPredicate<
// MOV Wd, #0
// MOV Xd, #0
CheckAll<[CheckOpcode<[MOVZWi, MOVZXi]>,
CheckAll<[CheckImmOperand<1, 0>,
CheckImmOperand<2, 0>]>]>,
CheckIsImmOperand<1>,
CheckImmOperand<1, 0>,
CheckImmOperand<2, 0>]>,
// MOV Wd, WZR
// MOV Xd, XZR
// MOV Wd, Wn
Expand Down
52 changes: 52 additions & 0 deletions llvm/test/CodeGen/AArch64/misched-move-imm.mir
Original file line number Diff line number Diff line change
@@ -0,0 +1,52 @@
# RUN: llc -run-pass=machine-scheduler -mtriple=aarch64-linux-gnu -mcpu=neoverse-v2 %s -o /dev/null 2>&1
# Just ensure this doesn't crash. Ensures in the neoverse-v2
# scheduling model we don't attempt to treat the first input
# operand of MOVZXi as an immediate operand.

--- |
target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128-Fn32"

declare void @foo2(<2 x float>) #0

define void @foo1() #0 {
call void @foo2(<2 x float> <float 2.500000e-01, float 7.500000e-01>)
ret void
}

attributes #0 = { "target-cpu"="neoverse-v2" }

...
---
name: foo1
alignment: 16
tracksRegLiveness: true
registers:
- { id: 0, class: gpr64 }
- { id: 1, class: gpr64 }
- { id: 2, class: gpr64common }
- { id: 3, class: gpr64common }
- { id: 4, class: fpr64 }
frameInfo:
maxAlignment: 1
adjustsStack: true
hasCalls: true
maxCallFrameSize: 0
constants:
- id: 0
value: '<2 x i32> <i32 1048576000, i32 1061158912>'
alignment: 8
machineFunctionInfo: {}
body: |
bb.0 (%ir-block.0):
ADJCALLSTACKDOWN 0, 0, implicit-def dead $sp, implicit $sp
%2:gpr64common = MOVZXi target-flags(aarch64-g0, aarch64-nc) %const.0, 0
%2:gpr64common = MOVKXi %2, target-flags(aarch64-g1, aarch64-nc) %const.0, 16
%2:gpr64common = MOVKXi %2, target-flags(aarch64-g2, aarch64-nc) %const.0, 32
%2:gpr64common = MOVKXi %2, target-flags(aarch64-g3) %const.0, 48
%4:fpr64 = LDRDui %2, 0 :: (load (s64) from constant-pool)
$d0 = COPY %4
BL @foo2, csr_aarch64_aapcs, implicit-def dead $lr, implicit $sp, implicit killed $d0, implicit-def $sp
ADJCALLSTACKUP 0, 0, implicit-def dead $sp, implicit $sp
RET_ReallyLR

...
Loading