[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: repeat you test on P4
There is a line a little bit down saying
nb=pars.get_fastest_gemm('p4',pre)
change p4 to p3, and then change the directory a little bit further down
from Linux_P4 to what ever you are using. That should do it.
If you just want Atlas performance numbers just look in
ATLAS/bin/<arch>/INSTALL_LOG/SUMMARY.LOG
and you will get the exact same numbers.
Cheers,
Peter.
On Wed, 25 Jul 2001 Jenwei_Hsieh@Dell.com wrote:
> Peter,
>
> That's very nice of you. One question for you. What should I change to use
> your test script for other platforms, like PIII? I'm trying to figure out
> the performance difference, in term of Gflops, between your test script and
> HPL using BLAS from ATLAS. Thanks again.
>
> Regards,
>
> --
> Jenwei
>
> -----Original Message-----
> From: Peter Soendergaard [mailto:soender@cs.utk.edu]
> Sent: Tuesday, July 24, 2001 4:38 PM
> To: Jenwei_Hsieh@exchange.dell.com
> Subject: RE: repeat you test on P4
>
>
> Glad I could help. It is nice to have ones software tested and used by
> others.
>
> Cheers,
>
> Peter.
>
> On Tue, 24 Jul 2001 Jenwei_Hsieh@Dell.com wrote:
>
> > Peter,
> >
> > It works now with a new download. Thanks a lot for your help. I used your
> > test to make sure that our P4 boxes do not have the overheat problem.
> >
> > Regards,
> >
> > --
> > Jenwei
> >
> > -----Original Message-----
> > From: Peter Soendergaard [mailto:soender@cs.utk.edu]
> > Sent: Monday, July 23, 2001 2:54 PM
> > To: Jenwei_Hsieh@exchange.dell.com
> > Subject: RE: repeat you test on P4
> >
> >
> > I checked the homepage and it is updated alright. Maybe you will have to
> > force a reload of the page or something, html is not my speciality.
> >
> > Anyway, the updated file starts with a long description of what to do, so
> > you should be able to recognize it.
> >
> > Cheers,
> >
> > Peter.
> >
> > On Mon, 23 Jul 2001 Jenwei_Hsieh@Dell.com wrote:
> >
> > > Peter,
> > >
> > > I got the same problem with the Syntax Error. Here's the error message:
> > >
> > > [root@p330 gemm]# pwd
> > > /root/ATLAS/tune/blas/gemm
> > > [root@p330 gemm]# ./runhot.py
> > > File "./runhot.py", line 11
> > > nb=pars.('p4',pre)
> > > ^
> > > SyntaxError: invalid syntax
> > >
> > > Is there something I didn't do correctly? Thanks,
> > >
> > > Regards,
> > >
> > > --
> > > Jenwei
> > >
> > > -----Original Message-----
> > > From: Peter Soendergaard [mailto:soender@cs.utk.edu]
> > > Sent: Sunday, July 22, 2001 6:28 AM
> > > To: Jenwei_Hsieh@exchange.dell.com
> > > Cc: atlas-comm@cs.utk.edu
> > > Subject: Re: repeat you test on P4
> > >
> > >
> > > Hello Jenwei,
> > >
> > > I am back from my holliday, so if you are still insterested I have fixed
> > > the P4 throtelling test. If you download the tar-file from my homepage
> > > http://www.cs.utk.edu/~soender/atlas/src/ps_atlas.tgz, untar it on top
> of
> > > your atlas directory and run the script it should work now, otherwise
> > > please respond and I will fix what ever I forgot. Instructions on output
> > > and configuring the script is included in the script, runhot.py
> > >
> > > Cheers,
> > >
> > > Peter.
> > >
> > > On Tue, 3 Jul 2001 Jenwei_Hsieh@Dell.com wrote:
> > >
> > > > Peter,
> > > >
> > > > I read you e-mail ("titled P4 slowdown on May 7) regarding you test on
> > > > P4. I tried to repeat your test but got Syntax Error right after
> > executed
> > > > runhot.py. I am wondering if you could provide more instruction on how
> > > > to run your test for a Python dummy like me. Thanks a lot!
> > > >
> > > > Regards,
> > > >
> > > > --
> > > > Jenwei
> > > >
> > > >
> > > > -------cut from original mail-------------
> > > > If anyone wants to repeat my test the can grab my codegenerator at
> > > > www.cs.utk.edu/~soender/atlas/src/ps_atlas.tgz
> > > >
> > > > Untar it on top of your ATLAS tree, go to ATLAS/tune/blas/gemm and
> > execute
> > > > runhot.py (requires python) and look at the file runhotdata. First
> > column
> > > > is seconds, second column is mflops. If you have gnuplot installed
> > execute
> > > > 'gnuplot gnuplotrunhot' and look at runhot.eps
> > > >
> > > >
> > >
> > >
> > >
> >
> >
>
>