[Pw_forum] Re: Woodcrest vs Opteron performance in pwscf calc.
Alexander Shaposhnikov
shaposh at isp.nsc.ru
Thu Aug 3 12:55:41 CEST 2006
My experience with Pathscale compiler shows that it does not give any
substantial improvement over Intel compiler for AMD, in most cases.
Intel compiler is also more compatible with different FORTRAN codes,
and, last but not least, it is free for non-commercial use.
By the way, gcc-compiled ACML is compatible with Intel compiler.
I use it. Intel compiler can even be used to compile and run code on AMD
with -xP optimization, i.e. with SSE3 instructions. But performance
improvement (if any) is negligible.
I will run tests and post results as soon as i get access to the
machines.
Best Regards,
A. Shaposhnikov
On Thu, 2006-08-03 at 17:10 +0800, Huiqun Zhou wrote:
> Hi, Axel and list users,
>
> >
> > actually, it is. some people here spend a lot of money
> > on new machines and figuring out what is the best deal
> > for a specific application needs a lot of testing. so
> > every contribution is important.
> >
>
> So, I continue to post here ;-)
>
> >
> > one thing to check when using intel MKL is, whether it is running in
> > multi-threaded mode and thus getting better results on a 'half-loaded'
> > machines. for that, you may want to re-run the jobs with the
> > environment variable OMP_NUM_THREADS set to 1. secondly, memory
> > contention is a problem, so it would be interesting to see the
> > performance, if you run 4 serial jobs at the same time.
> >
>
> OMP_NUM_THREADS had been set to 1 in all my tests.
>
> >
> > there is a way to make (the gcc) ACML compatible with the intel
> > compiler (at least for packages that use only double precision
> > functions), see:
> > https://www.liniac.upenn.edu/wiki/tiki-index.php?page=acml+for+CMM
> >
> > it would be nice to see, how using ACML would affect
> > the performance in this case.
> >
>
> If anyone can contribute a makefile for using PathScale compiler, the test
> will
> be handy. I have pathscale version of ACML 2.8. Unfortunately, I haven't
> succeded in compiling QE with this compiler yet.
>
> Best Regards,
>
>
> Huiqun Zhou
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum
More information about the Pw_forum
mailing list