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] k8 and amdfam10 cost model tuning

From: "Sjodin, Jan" <Jan dot Sjodin at amd dot com>
To: "Uros Bizjak" <ubizjak at gmail dot com>
Cc: "GCC Patches" <gcc-patches at gcc dot gnu dot org>
Date: Fri, 21 Dec 2007 08:49:27 -0600
Subject: RE: [patch] k8 and amdfam10 cost model tuning
References: <476A80A1.7030509@gmail.com> <9BCA02B0979C2A429C5965B8AE29269818C999@SAUSEXMB2.amd.com> <5787cf470712202326r7e2eaa80sa8ab1a80510aaf47@mail.gmail.com>

> I think you need -mfpmath=387 here (the default for 32bit compiles) to
> show the performance problem.
> 

Sorry I missed that part. Here is the new run:

21 Dec 2007  8:32:29 standardgfortransingle - Compile rnflow
command=gfortran -static -m32 -march=opteron -ffast-math -funroll-loops
-ftree-loop-linear -ftree-vectorize -mfpmath=387 -msse3 -O3 -w
rnflow.f90 -o rnflow
 
21 Dec 2007  8:32:45 standardgfortransingle - Execute rnflow
rnflow Run #  10    56.29950    56.27076 - Error=  0.0262%

Geometric Mean Execution Time =      56.27 seconds

Clearly something bad is happening.

- Jan

References:
- Re: [patch] k8 and amdfam10 cost model tuning
  - From: Uros Bizjak
- RE: [patch] k8 and amdfam10 cost model tuning
  - From: Sjodin, Jan
- Re: [patch] k8 and amdfam10 cost model tuning
  - From: Uros Bizjak

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