[llvm-commits] [Patch] Optimize integer ABS on X86

[Manman Ren]

Hi Chandler,

Thanks for reviewing.

Manman

On Jun 6, 2012, at 6:30 PM, Chandler Carruth wrote:

> Minor comment on the patch itself:
> 
> --- test/CodeGen/X86/iabs.ll	(revision 157837)
> +++ test/CodeGen/X86/iabs.ll	(working copy)
> @@ -1,13 +1,18 @@
>  ; RUN: llc < %s -march=x86-64 -stats  |& \
> -; RUN:   grep {5 .*Number of machine instrs printed}
> +; RUN:   grep {4 .*Number of machine instrs printed}
> +; RUN: llc < %s -march=x86-64 | FileCheck %s
> 
> Don't mix and match 'grep' and FileCheck based tests.
> 
> Actually, I would ask that any time you're about to touch a test case that currently uses grep, first convert it to use FileCheck, submit that, and then build your patch on top of it. We would like the test suite to slowly move away from all 'grep' based assertions.
I will use CHECK and CHECK-NEXT to make sure number of machine instrs printed is 4.
If that is not what you have in mind, please let me know.
> 
> +; rdar://10695237
> 
> I'd also attach rdar and bug links to the actual test code & filecheck assertions relevant for that particular issue…
Will do.
>