[llvm-commits] [PATCH] LoopVectorizer

[Hal Finkel]

----- Original Message -----
> From: "Shivarama Rao" <Shivarama.Rao at amd.com>
> To: "Ralf Karrenberg" <karrenberg at cdl.uni-saarland.de>, "Nadav Rotem" <nrotem at apple.com>
> Cc: "llvm-commits at cs.uiuc.edu LLVM" <llvm-commits at cs.uiuc.edu>
> Sent: Wednesday, October 17, 2012 8:20:13 AM
> Subject: Re: [llvm-commits] [PATCH] LoopVectorizer
> 
> Hi Nadav,
> 
> Some minor comments:
> 
> a) Some functions such as isConsecutiveGep, runOnLoop are inlined
> within the class. It will be good to see them defined outside.
> 
> b) The function CopyEmptyLoop can renamed to createEmptyLoop for
> better meaning of the function.
> 
> c) There are some inconsistencies in function name declarations. The
> functions CopyEmptyLoop, VectorizeLoop, DeleteOldLoop and
> ScalarizeInstructions are not consistent with
> getConsecutiveVector,isConsecutiveGep (first leter).  They can be
> made consistent.
> 
> d) Do we need to take any extra care of intrinsic calls?. In the
> current code, I guess they will be handled in ScalarizeInstruction
> function and I am not sure how they work. You can have a look on
> this.

In the current patch, these are not handled and they prevent vectorization.

 -Hal

> 
> Regards,
> Shivaram
> 
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> 

-- 
Hal Finkel
Postdoctoral Appointee
Leadership Computing Facility
Argonne National Laboratory