[PATCH] D143759: [AMDGPU] Implement whole wave register spill
Matt Arsenault via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Mon Jun 19 06:56:59 PDT 2023
arsenm added inline comments.
================
Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12934-12940
+ // Reserve the SGPR(s) to save/restore EXEC for WWM spill/copy handling.
+ unsigned MaxNumSGPRs = ST.getMaxNumSGPRs(MF);
+ Register SReg = ST.isWave32()
+ ? AMDGPU::SGPR_32RegClass.getRegister(MaxNumSGPRs - 1)
+ : TRI->getAlignedHighSGPRForRC(MF, /*Align=*/2,
+ &AMDGPU::SGPR_64RegClass);
+ Info->setSGPRForEXECCopy(SReg);
----------------
arsenm wrote:
> Move this up with the other reserved register logic
Actually, why is this not in getReservedRegs?
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D143759/new/
https://reviews.llvm.org/D143759
More information about the llvm-commits
mailing list