This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

PING: Re: [patch] implement simd loops in trunk (OMP_SIMD)

On 07/09/13 18:02, Aldy Hernandez wrote:

PING PING folks. Mr. Biener and/or Mr. Henderson. Could either of you look at this, especially the vectorizer bits. This is needed for the CILK_SIMD patchset that Jason is currently looking at.


Hi folks.

I have distilled the relevant machinery to implement and handle simd
loops from the gomp-4_0-branch.  This will be used by both gomp4
(#pragma omp simd) and Cilk Plus (#pragma simd).

All of it is Jakub's code.  There is nothing new here, as I believe
Jakub and the Richards have talked about all this at length.  Included
in the patch are the vectorizer changes that have been previously
discussed, all from the gomp-4_0-branch.

It would be ideal to get an ok from either Mr. Biener or Mr. Henderson
on the vectorizer changes.

The patch I am formally submitting is the attached patch named OMP_SIMD.
  I am including the second (CILK_SIMD), which I will submit
independently, as reference to how it will be used (if for some reason
it isn't clear from the gomp4 branch usage of OMP_SIMD).

Tested on x86-64 Linux with and without the Cilk Plus bits, not to
mention that this has been extensively tested on the gomp4 branch.

OK for trunk?

Attachment: OMP_SIMD
Description: Text document

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]