[gmx-users] Testing gromacs on IBM POWER8

Szilárd Páll pall.szilard at gmail.com
Thu Aug 27 15:03:32 CEST 2015


A few more things to add:
- we don't have CPU SIMD kernels for Power8, so comparing the plain C
kernels against AVX kernels on AMD is not exactly fair;
- as Mark pointed out, to use resources efficiently, you'll have to manage
thread affinity on Power8 (by default you'll need a stride of 8 to use one
thread per core, a stride of 4 to use two, etc.);
- for a realistic performance measurement use ~100-1000 atoms/core; the
"0000.96" benchmark in a 960 atom input, running that cross 192 threads
makes little sense.

What motivates your choice of compiler flags?


--
Szilárd

On Thu, Aug 27, 2015 at 9:17 AM, Mark Abraham <mark.j.abraham at gmail.com>
wrote:

> Hi,
>
> I have no idea what you're trying to show with these graphs. Your vertical
> axis of time makes it looks like a 2.5 year old AMD chip is walking all
> over POWER8?
>
> Other points
> 1) GROMACS doesn't use configure or fortran, so don't mention that
> 2) these simulations do not use lapack or blas, so don't mention that
> 3) you need to clarify what a "CPU" is... core, hardware thread?
> 4) when you're using fewer cores than the whole node, then you need to
> report how you are managing thread affinity
> 5) the splines on the graphs are misleading when reporting discrete
> quantities
> 6) you need to report times that are measured after auto-tuning completes
> 7) you need to report whether you are using MPI, thread-MPI or OpenMP to
> distribute work to threads.
>
> Mark
>
>
> On Tue, Aug 25, 2015 at 11:54 PM Fabricio Cannini <fcannini at gmail.com>
> wrote:
>
> > Hello there
> >
> > We've been testing gromacs-5.1-beta1 in an IBM POWER8. The input used is
> > 'pme.mdp' within the "0000.96" directory of the following benchmark:
> >
> > ftp://ftp.gromacs.org/pub/benchmarks/water_GMX50_bare.tar.gz
> >
> >
> >
> > The following link shows our results:
> >
> > http://suporte.versatushpc.com.br/power/gromacs.php
> >
> >
> >
> > These are our first results. We're now running more tests using IBM XL
> > compilers, the just launched 5.1.0 version and OpenBLAS port to POWER8.
> >
> > Please feel free to ask questions, suggests improvements and spot errors.
> > --
> > Gromacs Users mailing list
> >
> > * Please search the archive at
> > http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> > posting!
> >
> > * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
> >
> > * For (un)subscribe requests visit
> > https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> > send a mail to gmx-users-request at gromacs.org.
> >
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>


More information about the gromacs.org_gmx-users mailing list