[PATCH] D110376: [AMDGPU] Always reserve flat scratch SGPR for architected flat scratch

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec created this revision.
rampitec added reviewers: kzhuravl, msearles.
Herald added subscribers: foad, kerbowa, hiraditya, t-tye, tpr, dstuttard, yaxunl, nhaehnle, jvesely, arsenm.
rampitec requested review of this revision.
Herald added a subscriber: wdng.
Herald added a project: LLVM.

With architected flat scratch it becomes readonly. We must always
reserve SGPR pair for it even if we do not use scratch at all since
an attempt to write to SGPRs mapped to FLAT_SCRATCH results in
memory violation.

This is not needed since GFX10 with architected flat scratch though
since special SGPRs are not carving space from normal SGPRs.


https://reviews.llvm.org/D110376

Files:
  llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
  llvm/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
  llvm/test/CodeGen/AMDGPU/flat-scratch-reg.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D110376.374703.patch
Type: text/x-patch
Size: 4734 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210924/b52de5ea/attachment-0001.bin>