This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
RE: Help for compiling PERFECTLY SPECjbb2000 with gcj
- From: "Boehm, Hans" <hans_boehm at hp dot com>
- To: "'Anthony Green'" <green at redhat dot com>,Tony Reix <Tony dot Reix at bull dot net>
- Cc: java at gcc dot gnu dot org, Tony dot Reix at frec dot bull dot fr
- Date: Thu, 28 Aug 2003 11:58:52 -0700
- Subject: RE: Help for compiling PERFECTLY SPECjbb2000 with gcj
Do -ffast-math -mfpmath=sse preserve Java fp semantics? The benchmark is very sensitive to floating point rounding. In my experience the initial correctness testing run fails if you get that appreciably wrong, though it probably doesn't change the amount of work done by later stages of the benchmark very much.
Hans
> -----Original Message-----
> From: Anthony Green [mailto:green@redhat.com]
> Sent: Thursday, August 28, 2003 11:29 AM
> To: Tony Reix
> Cc: java@gcc.gnu.org; Tony.Reix@frec.bull.fr
> Subject: Re: Help for compiling PERFECTLY SPECjbb2000 with gcj
>
>
> On Wed, 2003-08-27 at 09:05, Tony Reix wrote:
> > Done on a 2xia32 (RedHat 9 , 2.4.20-8smp) 1.266 GHz and
> 512K cache, gcc 3.3.1 .
>
> Try compiling with -O3 -fomit-frame-pointer -march=pentium4
> -finline-functions -finline-limit=800 -ffast-math -mfpmath=sse
>
> AG
>
> --
> Anthony Green <green@redhat.com>
> Red Hat, Inc.
>