On Mon, Oct 29, 2012 at 2:51 PM, Berk Hess <span dir="ltr"><<a href="mailto:hess@kth.se" target="_blank">hess@kth.se</a>></span> wrote:<br><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hi,<br>
<br>
Compiling in GPU functionality has no effect on performance.<br>
We have never timed the effect of compiling in MPI and OpenMP,<br>
but I would expect it to be very small, probably less than 1%.<br>
But it would be good to try that once on a small system.<br></blockquote><div><br></div><div>I've just tried (again) the performance difference in thread-MPI-only runs using mdrun compiled with vs without is negligible (<0.5%).</div>
<div><br></div><div>--<br>Szilárd<br></div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Note that pure OpenMP can be much faster than MPI.<br>
<br>
The difference in performance between SSE2 and SSE4.1 is minor,<br>
less than 5% (AVX is a lot faster though).<br>
So I think you can use a single installation.<br>
<br>
Cheers,<br>
<br>
Berk<div class="HOEnZb"><div class="h5"><br>
<br>
On 10/29/12 14:47 , Carsten Kutzner wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Dear developers,<br>
<br>
we run Gromacs on an inhomogeneous cluster with different node groups -<br>
old and new ones, where each group requires an individual acceleration<br>
setting, e.g. SSE2 vs. SSE4.1. If I set GMX_ACCELERATION to SSE2 then I<br>
won't get the optimal performance on some of the nodes, whereas with SSE4.1<br>
the code won't run on all nodes. One way around seems to build different<br>
mdruns, and to select the optimal one after the node has been assigned by the<br>
queue. Is there a more elegant way like including alternate code paths,<br>
where the optimal path is selected at run time?<br>
<br>
2nd question: From a performance point of view, does it hurt to have support<br>
for GPU, MPI, and threads compiled in - even if it is not used in many cases?<br>
Or would it be better to have also separate executables for that?<br>
<br>
Thanks for your thoughts,<br>
Carsten<br>
<br>
<br>
--<br>
Dr. Carsten Kutzner<br>
Max Planck Institute for Biophysical Chemistry<br>
Theoretical and Computational Biophysics<br>
Am Fassberg 11, 37077 Goettingen, Germany<br>
Tel. <a href="tel:%2B49-551-2012313" value="+495512012313" target="_blank">+49-551-2012313</a>, Fax: <a href="tel:%2B49-551-2012302" value="+495512012302" target="_blank">+49-551-2012302</a><br>
<a href="http://www.mpibpc.mpg.de/grubmueller/kutzner" target="_blank">http://www.mpibpc.mpg.de/<u></u>grubmueller/kutzner</a><br>
<br>
</blockquote>
<br>
-- <br>
gmx-developers mailing list<br>
<a href="mailto:gmx-developers@gromacs.org" target="_blank">gmx-developers@gromacs.org</a><br>
<a href="http://lists.gromacs.org/mailman/listinfo/gmx-developers" target="_blank">http://lists.gromacs.org/<u></u>mailman/listinfo/gmx-<u></u>developers</a><br>
Please don't post (un)subscribe requests to the list. Use the www interface or send it to <a href="mailto:gmx-developers-request@gromacs.org" target="_blank">gmx-developers-request@<u></u>gromacs.org</a>.<br>
</div></div></blockquote></div><br>