[PATCH] D124195: [AMDGPU] Separate out custom SGPR spills to VGPR during PEI
Christudasan Devadasan via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Fri Sep 23 05:37:38 PDT 2022
cdevadas added inline comments.
================
Comment at: llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.h:469
unsigned NumVGPRSpillLanes = 0;
+ unsigned NumVGPRCustomSpillLanes = 0;
SmallVector<Register, 2> SpillVGPRs;
----------------
arsenm wrote:
> Isn't this count implied by SGPRToVGPRCustomSpills.size()? I'd like to avoid multiplying the number of unserialized fields
Currently, the function `SIMachineFunctionInfo::allocateSGPRSpillToVGPR` needs this variable to choose the num spills between the SILowerSGPRSpills pass and the custom spills later during FrameLowering. I'm planning to move these functions entirely out of SIMachineFunctionInfo and can avoid these variables entirely.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D124195/new/
https://reviews.llvm.org/D124195
More information about the llvm-commits
mailing list