Skip to content

[AMDGPU] Fix s_singleuse_vdst not detecting exec mask changes #89401

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.

Sign up for GitHub

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

[AMDGPU] Fix s_singleuse_vdst not detecting exec mask changes #89401

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
2 changes: 1 addition & 1 deletion llvm/lib/Target/AMDGPU/AMDGPUInsertSingleUseVDST.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,7 +112,7 @@ class AMDGPUInsertSingleUseVDST : public MachineFunctionPass {

// Reset uses count when a register is no longer live.
for (const MCRegUnit &Unit : RegUnits)
RegisterUseCount.erase(Unit);
RegisterUseCount[Unit] = 0;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's really hard to understand the effect of this change because of other problems in this loop-over-the-operands.

  1. The modifiesRegister check on line 104 is odd because it checks if any operand of the current instruction modifies Reg, not just the current operand. Can it just check Operand.isDef() instead?

  2. Really the loop should be split into two processing defs first and then uses. I think you had a downstream patch to do that.

Could you post a patch to fix either (1) or (2) first, and then we can revisit this one?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Okay, I do have a downstream patch for that. I'll post that first.

}
if (AllProducerOperandsAreSingleUse && SIInstrInfo::isVALU(MI)) {
// TODO: Replace with candidate logging for instruction grouping
Expand Down
3 changes: 0 additions & 3 deletions llvm/test/CodeGen/AMDGPU/insert-singleuse-vdst.mir
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -398,7 +398,6 @@ body: |
; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: liveins: $sgpr0_sgpr1
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: S_SINGLEUSE_VDST 1
; CHECK-NEXT: $vgpr0 = V_MOV_B32_e32 0, implicit $exec
; CHECK-NEXT: $exec = COPY $sgpr0_sgpr1
; CHECK-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr0, implicit $exec
Expand All @@ -424,7 +423,6 @@ body: |
; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: liveins: $sgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: S_SINGLEUSE_VDST 1
; CHECK-NEXT: $vgpr0 = V_MOV_B32_e32 0, implicit $exec
; CHECK-NEXT: $exec_lo = COPY $sgpr0
; CHECK-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr0, implicit $exec
Expand All @@ -450,7 +448,6 @@ body: |
; CHECK-NEXT: successors: %bb.1(0x80000000)
; CHECK-NEXT: liveins: $sgpr0
; CHECK-NEXT: {{ $}}
; CHECK-NEXT: S_SINGLEUSE_VDST 1
; CHECK-NEXT: $vgpr0 = V_MOV_B32_e32 0, implicit $exec
; CHECK-NEXT: $exec_hi = COPY $sgpr0
; CHECK-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr0, implicit $exec
Expand Down