[PATCH] D58928: [AMDGPU] Prepare for introduction of v3 and v5 MVTs

[Tim Renouf via Phabricator via llvm-commits]

tpr created this revision.
Herald added subscribers: llvm-commits, t-tye, dstuttard, yaxunl, nhaehnle, wdng, jvesely, kzhuravl, arsenm.
Herald added a project: LLVM.

AMDGPU would like to have MVTs for v3i32, v3f32, v5i32, v5f32. This
commit does not add them, but makes preparatory changes:

- Fixed assumptions of power-of-2 vector type in kernel arg handling, and added v5 kernel arg tests.

- Added v5 tests for cost analysis.

Some of this patch is from Matt Arsenault, also of AMD.

Change-Id: I7279d6b4841464d2080eb255ef3c589e268eabcd


Repository:
  rL LLVM

https://reviews.llvm.org/D58928

Files:
  lib/Target/AMDGPU/AMDGPUISelLowering.cpp
  test/Analysis/CostModel/AMDGPU/add-sub.ll
  test/Analysis/CostModel/AMDGPU/extractelement.ll
  test/Analysis/CostModel/AMDGPU/fabs.ll
  test/Analysis/CostModel/AMDGPU/fadd.ll
  test/Analysis/CostModel/AMDGPU/fdiv.ll
  test/Analysis/CostModel/AMDGPU/fmul.ll
  test/Analysis/CostModel/AMDGPU/fsub.ll
  test/Analysis/CostModel/AMDGPU/mul.ll
  test/CodeGen/AMDGPU/kernel-args.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D58928.189215.patch
Type: text/x-patch
Size: 10478 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190304/d88025bc/attachment.bin>