[PATCH] D123525: [AMDGPU] On gfx908, reserve VGPR for AGPR copy based on register budget.
Stanislav Mekhanoshin via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Mon Apr 11 11:22:50 PDT 2022
rampitec requested changes to this revision.
rampitec added inline comments.
This revision now requires changes to proceed.
================
Comment at: llvm/lib/Target/AMDGPU/SIInstrInfo.cpp:628
// or spill with the slot.
while (RegNo-- && RS.FindUnusedReg(&AMDGPU::VGPR_32RegClass)) {
Register Tmp2 = RS.scavengeRegister(&AMDGPU::VGPR_32RegClass, 0);
----------------
It will never get replaced if RegNo == 0.
================
Comment at: llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.h:520
+ VGPRForAGPRCopy =
+ (HighestAvailableVGPR < 16)
+ ? AMDGPU::VGPR_32RegClass.getRegister(32)
----------------
I do not know real world situation with such low budget.
================
Comment at: llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.h:521
+ (HighestAvailableVGPR < 16)
+ ? AMDGPU::VGPR_32RegClass.getRegister(32)
+ : AMDGPU::VGPR_32RegClass.getRegister(HighestAvailableVGPR);
----------------
And then 32 is over the budget. This is a bug.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D123525/new/
https://reviews.llvm.org/D123525
More information about the llvm-commits
mailing list