[PATCH] D33862: [x86] avoid flipping sign bits for vector icmp by using known bits

[Sanjay Patel via Phabricator via llvm-commits]

spatel updated this revision to Diff 101358.
spatel added a comment.

Patch updated:

1. Instead of changing the predicate from unsigned to signed, just remove the FlipSigns requirement when both operands are known non-negative. This lets the min/max/subus logic work as-is, so any case where we would have used those instructions should be unchanged.
2. Add a TODO comment to see if we can do more with the knownbits knowledge.


https://reviews.llvm.org/D33862

Files:
  lib/Target/X86/X86ISelLowering.cpp
  test/CodeGen/X86/avx512-cvt.ll
  test/CodeGen/X86/vector-unsigned-cmp.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D33862.101358.patch
Type: text/x-patch
Size: 13275 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170604/e4c52ec8/attachment.bin>