[PATCH] D34130: [AMDGPU] Eliminate SGPR to VGPR copy when possible

[Stanislav Mekhanoshin via Phabricator via llvm-commits]

rampitec created this revision.
Herald added subscribers: t-tye, tpr, dstuttard, yaxunl, nhaehnle, wdng, kzhuravl.

SGPRs are generally cheaper, so try to use them over VGPRs.


Repository:
  rL LLVM

https://reviews.llvm.org/D34130

Files:
  lib/Target/AMDGPU/SIFixSGPRCopies.cpp
  test/CodeGen/AMDGPU/llvm.amdgcn.mqsad.pk.u16.u8.ll
  test/CodeGen/AMDGPU/llvm.amdgcn.qsad.pk.u16.u8.ll
  test/CodeGen/AMDGPU/opt-sgpr-to-vgrp-copy.ll
  test/CodeGen/AMDGPU/sdwa-peephole.ll
  test/CodeGen/AMDGPU/sint_to_fp.i64.ll
  test/CodeGen/AMDGPU/uint_to_fp.i64.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D34130.102279.patch
Type: text/x-patch
Size: 6455 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170613/2f45a1a4/attachment.bin>