[PATCH] D26330: [X86] Lower AVX512 and SSE intrinsics for CVTTPD2DQ to X86ISD::CVTTPD2DQ.

[Craig Topper via llvm-commits]

craig.topper updated this revision to Diff 77159.
craig.topper added a comment.

Turns out those movqs were caused by the vzmovl patterns being rooted with a bitcast that didn't exist for the return values being 2xi64. To fix that remove the bitcasts, but then we hit priority problems with vzmovl patterns due to lots of AddedComplexity nodes on other patterns starting with vzmovl including the most vanilla vzmovl pattern. So threw in another AddedComplexity on these patterns to fix that. I'll try to scrub the vzmovl pattern mess sometime and see if we can't get these AddedComplexity's out.

I also added in a vzmovl cvtpd2ps pattern I noticed was missing


https://reviews.llvm.org/D26330

Files:
  lib/Target/X86/X86InstrAVX512.td
  lib/Target/X86/X86InstrSSE.td
  lib/Target/X86/X86IntrinsicsInfo.h
  test/CodeGen/X86/avx-intrinsics-x86.ll
  test/CodeGen/X86/sse2-intrinsics-x86.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D26330.77159.patch
Type: text/x-patch
Size: 12483 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20161108/95886c9f/attachment.bin>