[PATCH] D50069: AMDGPU: Improve hack for packing conversion ops

[Matt Arsenault via Phabricator via llvm-commits]

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

Mutate the node type during selection when it
doesn't matter. This avoids an intermediate bitcast
node on targets with legal i16/f16.

      

Also fixes missing output modifiers on v_cvt_pkrtz_f32_f16,
which I assume are OK.


https://reviews.llvm.org/D50069

Files:
  lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
  lib/Target/AMDGPU/SIISelLowering.cpp
  lib/Target/AMDGPU/SIInstrInfo.td
  lib/Target/AMDGPU/VOP2Instructions.td
  test/CodeGen/AMDGPU/clamp-modifier.ll
  test/CodeGen/AMDGPU/coalescer-subranges-another-prune-error.mir
  test/CodeGen/AMDGPU/couldnt-join-subrange-3.mir
  test/CodeGen/AMDGPU/subreg-split-live-in-error.mir

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D50069.158266.patch
Type: text/x-patch
Size: 17418 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20180731/f064f4d4/attachment-0001.bin>