<html><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><div><br></div><div>This file that was attached did not go through I guess,</div><div>so I send my comments again :))</div><div><br>On Feb 12, 2010, at 6:14 PM, David van der Spoel wrote:<br><br><blockquote type="cite">On 2/12/10 5:32 PM, Sander Pronk wrote:<br></blockquote><blockquote type="cite"><blockquote type="cite">Now that the last issues have been resolved with the threading code, thread-based parallelization has been turned on by default. To disable all the threading code, use<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">--disable-threads.<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">in configure, or turn the option GMX_THREADS off with ccmake.<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Running mdrun with just one thread (the default) is almost exactly the same as running it without threading code: the only thing that's different is that the few remaining global variables are protected by mutexes.<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Performance-wise mdrun runs very slightly faster with threads than with OpenMPI when Nthreads<=Ncores (and there is no other processes on the computer). When Nthreads>Ncores (or other processes are running), the thread code is much faster than OpenMPI, but the total runtime is still smaller than when Nthreads==Ncores.<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">If there's any problems in getting things running, or with performance, I'd very much like to hear about it.<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite">Sander<br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><br></blockquote></blockquote><blockquote type="cite">Great work Sander!<br></blockquote><blockquote type="cite"><br></blockquote><blockquote type="cite">What about the rumours that enabling hyperthreading gives you superscaling beyond Ncores?<br></blockquote>my experience on poweer-PCs is that using hyperthreading indeed gives you<br>back some of the lost performance due to unperfect scaling. With few cpus you get</div><div>superscaling :))</div><div><br></div><div>In France they also have a power-PC supercomputer and use the hyperthreading </div><div>as default ... <br><br>I attach a file showing some tests I made some time ago.<br>(Huygens is the power-PC thing<br><br><br><blockquote type="cite"><br></blockquote><blockquote type="cite">-- David van der Spoel, Ph.D., Professor of Biology<br></blockquote><blockquote type="cite">Dept. of Cell & Molec. Biol., Uppsala University.<br></blockquote><blockquote type="cite">Box 596, 75124 Uppsala, Sweden. Phone:<span class="Apple-tab-span" style="white-space: pre; "> </span>+46184714205. Fax: +4618511755.<br></blockquote><blockquote type="cite"><a href="mailto:spoel@xray.bmc.uu.se">spoel@xray.bmc.uu.se</a><span class="Apple-tab-span" style="white-space: pre; "> </span><a href="mailto:spoel@gromacs.org">spoel@gromacs.org</a> <a href="http://folding.bmc.uu.se/">http://folding.bmc.uu.se</a><br></blockquote><blockquote type="cite">-- gmx-developers mailing list<br></blockquote><blockquote type="cite"><a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a><br></blockquote><blockquote type="cite"><a href="http://lists.gromacs.org/mailman/listinfo/gmx-developers">http://lists.gromacs.org/mailman/listinfo/gmx-developers</a><br></blockquote><blockquote type="cite">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">gmx-developers-request@gromacs.org</a>.<br></blockquote><div><br></div></div></body></html>