Skip to content

[AMDGPU] Use subtarget call to determine number of VGPRs #157927

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

Merged
merged 1 commit into from
Sep 11, 2025
Merged

[AMDGPU] Use subtarget call to determine number of VGPRs #157927

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/SIISelLowering.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3452,7 +3452,7 @@ bool SITargetLowering::CanLowerReturn(

// We must use the stack if return would require unavailable registers.
unsigned MaxNumVGPRs = Subtarget->getMaxNumVGPRs(MF);
unsigned TotalNumVGPRs = AMDGPU::VGPR_32RegClass.getNumRegs();
unsigned TotalNumVGPRs = Subtarget->getAddressableNumArchVGPRs();
Copy link
Contributor

Choose a reason for hiding this comment

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

This use is curious, we should only be looking at the number of registers used for return in the calling convention

Copy link
Collaborator Author

Choose a reason for hiding this comment

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

I am not even saying it was correct. But using registers which do not exist is not correct for sure. It has just... floated to the top.

for (unsigned i = MaxNumVGPRs; i < TotalNumVGPRs; ++i)
if (CCInfo.isAllocated(AMDGPU::VGPR_32RegClass.getRegister(i)))
return false;
Expand Down
2 changes: 1 addition & 1 deletion llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,7 +86,7 @@ SIMachineFunctionInfo::SIMachineFunctionInfo(const Function &F,
// FIXME: MayNeedAGPRs is a misnomer for how this is used. MFMA selection
// should be separated from availability of AGPRs
if (MFMAVGPRForm ||
(ST.getMaxNumVGPRs(F) <= AMDGPU::VGPR_32RegClass.getNumRegs() &&
(ST.getMaxNumVGPRs(F) <= ST.getAddressableNumArchVGPRs() &&
Copy link
Contributor

Choose a reason for hiding this comment

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

I'm working on deleting this part

!mayUseAGPRs(F)))
MayNeedAGPRs = false; // We will select all MAI with VGPR operands.
}
Expand Down
9 changes: 6 additions & 3 deletions llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1399,13 +1399,16 @@ unsigned getTotalNumVGPRs(const MCSubtargetInfo *STI) {
return IsWave32 ? 1024 : 512;
}

unsigned getAddressableNumArchVGPRs(const MCSubtargetInfo *STI) { return 256; }
unsigned getAddressableNumArchVGPRs(const MCSubtargetInfo *STI) {
const auto &Features = STI->getFeatureBits();
if (Features.test(Feature1024AddressableVGPRs))
return Features.test(FeatureWavefrontSize32) ? 1024 : 512;
return 256;
}

unsigned getAddressableNumVGPRs(const MCSubtargetInfo *STI,
unsigned DynamicVGPRBlockSize) {
const auto &Features = STI->getFeatureBits();
if (Features.test(FeatureGFX1250Insts))
return Features.test(FeatureWavefrontSize32) ? 1024 : 512;
if (Features.test(FeatureGFX90AInsts))
return 512;

Expand Down