This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] GCSE after reload - resubmit (SPECint +1.5% fp -0.5%).
- From: Mostafa Hagog <MUSTAFA at il dot ibm dot com>
- To: Vladimir Makarov <vmakarov at redhat dot com>
- Cc: Ayal Zaks <zaks at il dot ibm dot com>, David Edelsohn <dje at watson dot ibm dot com>, gcc-patches at gcc dot gnu dot org, Roger Sayle <roger at eyesopen dot com>
- Date: Wed, 25 Feb 2004 13:35:00 +0200
- Subject: Re: [PATCH] GCSE after reload - resubmit (SPECint +1.5% fp -0.5%).
Vladimir Makarov <vmakarov@redhat.com> wrote on 24/02/2004 18:43:57:
> This optimization is really useful. But I think it would be better
> remove redundency based on availability only (not partial redundency).
> Putting insns on critical edges results in generation of additional
> branches which sometimes is worse that removing some loads.
This decision should be based on a cost model. We already have a hard-coded
cost model for the partial redundancy elimination, which based on
feedback.
We should add a special consideration to critical edges.
We also intend to do some tuning in GCSE after reload pass, whenever the
mainline is stable in means of performance. Then we can provide several
parameters (using --param option) to make the tuning easier.
Mostafa