This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [patch] Loop-aware SLP
- From: Ira Rosen <IRAR at il dot ibm dot com>
- To: bonzini at gnu dot org
- Cc: Paolo Bonzini <bonzini at gnu dot org>, Daniel Berlin <dberlin at dberlin dot org>, Dorit Nuzman <DORIT at il dot ibm dot com>, gcc-patches at gnu dot org
- Date: Tue, 14 Aug 2007 20:01:55 +0300
- Subject: Re: [patch] Loop-aware SLP
Paolo Bonzini <paolo.bonzini@lu.unisi.ch> wrote on 14/08/2007 19:09:06:
>
> > However, COMPONENT_REFs are usually non-consecuitive
> > accesses, and currently non-consecuitive accesses can only be
vectorized if
> > they are strided. Hence, I don't think this will increase vectorization
> > potential too much.
>
> Well, not so much "usually". I would think that they are consecutive
> exactly in the kind of code that can benefit from COMPONENT_REF SLP.
> Though again, strided accesses are probably the majority.
Our SLP analysis is based on groups strided accesses. So, by
"non-consecuitive accesses can only be vectorized if they are strided", I
meant both SLP vectorization and strided accesses vectorization.
Thanks,
Ira
>
> Thanks for all your work!
>
> Paolo