[PATCH] D65325: [AMDGPU] Enable v4f16 and above for v_pk_fma instructions

[David Stuttard via Phabricator via llvm-commits]

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

If isel is presented with <2 x half> vectors then it will correctly select
v_pk_fma style instructions.
If isel is presented with e.g. <4 x half> vectors it will scalarize, unlike for
other instruction types (such as fadd, fmul etc.)

Added extra support to enable this. Updated one of the tests to include a test
for this (as well as extending the test to GFX9)


Repository:
  rL LLVM

https://reviews.llvm.org/D65325

Files:
  lib/Target/AMDGPU/SIISelLowering.cpp
  lib/Target/AMDGPU/SIISelLowering.h
  test/CodeGen/AMDGPU/fmac.sdwa.ll
  test/CodeGen/AMDGPU/llvm.fma.f16.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D65325.211906.patch
Type: text/x-patch
Size: 14933 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190726/052b7fd6/attachment.bin>