[PATCH] X86: sink splat-shuffle into block doing a shift.

[Nadav Rotem]

+ If
+  /// this is the case, LLVM will try to sink shufflevectors implementing a
+  /// splat before CodeGen to permit the better output.

LGTM, except that I would omit the line above.  The shufflevector sinking is one example of a usage of this hook. 


On Feb 18, 2014, at 4:54 AM, Tim Northover <t.p.northover at gmail.com> wrote:

>  I've tried to improve the clarity here, with slightly different TargetLowering function name and different doxygen text.
> 
>  Better? Worse?
> 
>  Cheers.
> 
>  Tim.
> 
> http://llvm-reviews.chandlerc.com/D2816
> 
> CHANGE SINCE LAST DIFF
>  http://llvm-reviews.chandlerc.com/D2816?vs=7164&id=7174#toc
> 
> Files:
>  include/llvm/Target/TargetLowering.h
>  lib/Target/X86/X86ISelLowering.cpp
>  lib/Target/X86/X86ISelLowering.h
>  lib/Transforms/Scalar/CodeGenPrepare.cpp
>  test/Transforms/CodeGenPrepare/x86-shuffle-sink.ll
> <D2816.3.patch>