This is the mail archive of the gcc-patches@gcc.gnu.org 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]

Re: [patch] enabling vectorization by default at -O3

From: Vladimir Makarov <vmakarov at redhat dot com>
To: Diego Novillo <dnovillo at google dot com>
Cc: gcc-patches at gcc dot gnu dot org
Date: Thu, 13 Sep 2007 17:14:29 -0400
Subject: Re: [patch] enabling vectorization by default at -O3
References: <OFF0ECF488.A8736665-ONC2257354.001F3A08-C2257354.002A3B8F@il.ibm.com> <46E9864F.8050308@codesourcery.com> <b798aad50709131155p6c923739g39bafd6ad92e76a7@mail.gmail.com>

Diego Novillo wrote:

On 9/13/07, Mark Mitchell <mark@codesourcery.com> wrote:

Do any of the other optimization maintainers have a feeling about this? Should we turn on vectorization at -O3, or are we not quite there yet?
Anything inside of 5% in SPEC scores is noise. Enabling vectorization

I guess you meant 0.5% not 5%. For SPEC2000 my observation that on some platforms it is even less (e.g. 0.3% for Intel Core2 server available to me).

at -O3 sounds like a fine idea to me.  After all, we always say that
-O3 is a "hopefully better" setting.
IMO, the biggest gain is the stability it will bring to the vectorizer and the incentive to keep tuning the cost model.

Follow-Ups:
- Re: [patch] enabling vectorization by default at -O3
  - From: Diego Novillo

References:
- Re: [patch] enabling vectorization by default at -O3
  - From: Dorit Nuzman
- Re: [patch] enabling vectorization by default at -O3
  - From: Mark Mitchell
- Re: [patch] enabling vectorization by default at -O3
  - From: Diego Novillo

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