[PATCH] D21774: [X86] Transform setcc + movzbl into xorl + setcc

[Michael Kuperstein via llvm-commits]

mkuper added inline comments.

================
Comment at: test/CodeGen/X86/sse42-intrinsics-x86.ll:1
@@ -1,1 +1,2 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc < %s -mtriple=i386-apple-darwin -mattr=-avx,+sse4.2 | FileCheck %s
----------------
RKSimon wrote:
> Regenerate this first so that we can see the diffs?
Will do.

But it's basically the same as in avx-intrinsics-x86.ll - some of the pcmpestr tests get pessimized, because we can't setcc directly into %al (since we can't xor it). So we use an additional register, and get the additional mov and pop (since it's the register's only use).


http://reviews.llvm.org/D21774