[PATCH] Vectorize intrinsic math function calls in SLPVectorizer

[Nadav Rotem]

Hi Karthik, 

I read your patch and I noticed that you moved lots of code into VectorUtils.h.  Please split your patch to two parts: 1. the part where you move code around and 2. the part where you add new functionality to the vectorizer. This will make the task of reviewing the patch much easier. 

Thanks,
Nadav


On Apr 29, 2014, at 4:25 AM, Karthik Bhat <kv.bhat at samsung.com> wrote:

> Hi nadav, aschwaighofer, hfinkel, rsilvera,
> 
> Hi All,
> Added a patch to vectorize intrinsic math functions in SLPVectorizer.
> If any intrinsic math function can be replaced with llvm intrinsic function we do so after doing the relevent check and vectorize the same if vectorization turns out to be benificial.
> This is quite similar to how we handle math functions in LoopVectorizer.
> 
> Could you please let me know your inputs on this patch?
> 
> Thanks
> Karthik Bhat
> 
> http://reviews.llvm.org/D3535
> 
> Files:
>  include/llvm/Transforms/Utils/VectorUtils.h
>  lib/Transforms/Vectorize/LoopVectorize.cpp
>  lib/Transforms/Vectorize/SLPVectorizer.cpp
>  test/Transforms/SLPVectorizer/X86/call.ll
> <D3535.8910.patch>