[gmx-users] Gromacs with Intel Xeon Phi coprocessors ?

[Szilárd Páll]

Hi Chris,

You should be able to run on MIC/Xeon Phi as these accelerators, when used
in symmetric mode, behave just like a compute node. However, for two main
reasons the performance will be quite bad:
- no SIMD accelerated kernels for MIC;
- no accelerator-specific parallelization implemented (asymmetric/"offload"
mode).

Note that OpenMP alone won't help much.

Both are work in progress (see redmine #1181 and #1187), but there is no
target date for the availability of an efficient MIX/Phi-optimized GROMACS
version. I personally hope that we will have something in form of preview
release based on 4.6 (but probably not included) and if it works well
enough, perhaps included in 5.0.


Let me take the opportunity to say that if anybody is interested in
contributing to the MIC acceleration (or other HPC, computing, or
scientific aspects of GROMACS or MD in general), we are in the process of
defining more or less independent projects for new
contribution/collaboration:
http://www.gromacs.org/Projects_available_for_new_contributors#Intel_MIC_support.3a.c2.a0implementing_asymmetric_offload_mechanism

Cheers,

--
Szilárd


On Sat, Mar 9, 2013 at 2:43 AM, Christopher Neale <
chris.neale at mail.utoronto.ca> wrote:

> Dear users:
>
> does anybody have any experience with gromacs on a cluster in which each
> node is composed of 1 or 2 x86 processors plus an Intel Xeon Phi
> coprocessor? Can gromacs make use of the xeon phi coprocessor? If not, does
> anybody know if that is in the pipeline?
>
> Thank you,
> Chris.
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> * Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>