[PATCH] D30213: [X86][SSE] Enable post-legalize vXi64 shuffle combining on 32-bit targets

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: test/CodeGen/X86/dagcombine-cse.ll:1-2
 ; REQUIRES: asserts
-; RUN: llc < %s -march=x86 -mattr=+sse2 -mtriple=i386-apple-darwin -stats 2>&1 | grep asm-printer | grep 13
+; RUN: llc < %s -march=x86 -mattr=+sse2 -mtriple=i386-apple-darwin -stats 2>&1 | grep asm-printer | grep 12
 
----------------
andreadb wrote:
> Do you think that there is still value in keeping this test?
> 
> If we want to keep this test, then I suggest to rewrite the RUN line (i.e. use FileCheck instead of grep, and regenerate the test using the update script).
It seems to be interested in the stats dump as opposed to the codegen so we can't as easily pipe into FileCheck (and update_llc_test_checks.py). But maybe I'm over thinking this and they really were just wanting to count the number of instructions.....

I'll investigate alternatives because I too have a loathing for the old grep style tests.....


Repository:
  rL LLVM

https://reviews.llvm.org/D30213