[PATCH] D136813: [AMDGPU] Add pseudo wavemode to optimize strict_wqm

[Carl Ritson via Phabricator via llvm-commits]

critson created this revision.
critson added reviewers: piotr, rampitec, nhaehnle, foad.
Herald added subscribers: kosarev, kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, jvesely, kzhuravl, arsenm.
Herald added a project: All.
critson requested review of this revision.
Herald added subscribers: llvm-commits, wdng.
Herald added a project: LLVM.

Strict WQM does not require a WQM transistion if it occurs within
an existing WQM section.
This occurs heavily in GFX11 pixel shaders with LDS_PARAM_LOAD.
Which leads to unnecessary EXEC mask manipulation.

To avoid these transitions, detect WQM -> Strict WQM -> WQM
and substitute new ENTER_PSEUDO_WM/EXIT_PSEUDO_WM markers instead.
These are treat similarly by WWM register pre-allocation pass,
but do not manipulate EXEC or use registers to save EXEC state.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D136813

Files:
  llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
  llvm/lib/Target/AMDGPU/SIInstructions.td
  llvm/lib/Target/AMDGPU/SIPreAllocateWWMRegs.cpp
  llvm/lib/Target/AMDGPU/SIWholeQuadMode.cpp
  llvm/test/CodeGen/AMDGPU/wqm.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D136813.471010.patch
Type: text/x-patch
Size: 10191 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20221027/217d845a/attachment.bin>