This is the mail archive of the
mailing list for the GCC project.
Re: [Patch] Splitting memory references during unrolling (resubmission)
- From: Revital Eres <ERES at il dot ibm dot com>
- To: David Edelsohn <dje at watson dot ibm dot com>
- Cc: Ayal Zaks <ZAKS at il dot ibm dot com>, Mircea Namolaru <NAMOLARU at il dot ibm dot com>, gcc-patches at gcc dot gnu dot org
- Date: Thu, 18 Nov 2004 15:05:31 +0200
- Subject: Re: [Patch] Splitting memory references during unrolling (resubmission)
David Edelsohn <firstname.lastname@example.org> wrote on 15/10/2004 05:47:41:
> What is the current SPEC performance benefit from this patch?
> Thanks, David
This patch gains ~1% improvements on overall CINTPEC2000 on
powerpc-linux when the number of times this optimization
is applied in each loop is restricted to one
(I can add this restriction if this patch will be accepted);
where gap gains ~5% improvements and bzip2 ~1%.
In CFPPEC2000 there was no significant improvements
except swim which gained 3% improvement but mesa,
facerec and wupwise show a regression of about 1%