<div>it is really helpful</div>
<div>Thank you so much..<br><br></div>
<div class="gmail_quote">On Tue, Apr 13, 2010 at 10:28 AM, Darius Buntinas <span dir="ltr"><<a href="mailto:buntinas@mcs.anl.gov">buntinas@mcs.anl.gov</a>></span> wrote:<br>
<blockquote style="BORDER-LEFT: #ccc 1px solid; MARGIN: 0px 0px 0px 0.8ex; PADDING-LEFT: 1ex" class="gmail_quote"><br>Socket connections are created for process management (between the mpiexec process and MPD processes and between the MPD processes and the MPI processes). However communication between the MPI processes will use shared memory. We don't use sysv shared memory because of the limits on the number and size of segments. Instead we create temporary files and mmap them.<br>
<br>You should, however, see a significant difference in performance on a multicore machine. Try a latency benchmark (e.g., osu_latency from the OMB: <a href="http://mvapich.cse.ohio-state.edu/benchmarks" target="_blank">http://mvapich.cse.ohio-state.edu/benchmarks</a> ). Using nemesis, you can force all communication to go through the network by setting the MPICH_NO_LOCAL environment variable for comparison:<br>
<br>For shared memory performance:<br> mpiexec -n 2 osu_latency<br>For sockets performance:<br> MPICH_NO_LOCAL=1 mpiexec -n 2 osu_latency<br><br>Here's a website on tips for measuring shared memory latency:<br><a href="http://wiki.mcs.anl.gov/mpich2/index.php/Measuring_Nemesis_Performance" target="_blank">http://wiki.mcs.anl.gov/mpich2/index.php/Measuring_Nemesis_Performance</a><br>
<font color="#888888"><br>-d</font>
<div class="im"><br><br><br>On 04/13/2010 06:25 AM, Dai-Hee Kim wrote:<br></div>
<blockquote style="BORDER-LEFT: #ccc 1px solid; MARGIN: 0px 0px 0px 0.8ex; PADDING-LEFT: 1ex" class="gmail_quote">
<div>
<div></div>
<div class="h5">Hello, everyone<br>I am testing MPICH2 on 24 cores SMP machine.<br>And, I installed MPICH2 with three devices (nemesis, ssm,<br>shm) separately and run parallel program with each device.<br>However, it seems that no device was using shared memory when I checked<br>
the performance<br>and network status using netstat command (a lot of sockets created for<br>self connection)<br>Of course, I could not see any shared memory segments through ipcs -m<br>and free commands<br>I compiled MPICH2 with three devices by below configuration options<br>
respectively<br>./configure --prefix=.../nemesis --enable-fast=03 --with-device=ch3:nemesis<br>./configure --prefix=.../ssm --enable-fast=03 --with-device=ch3:ssm<br>./configure --prefix=.../shm --enable-fast=03 --with-device=ch3:shm<br>
and complied and run the parallel program using scripts (mpif90 and<br>mpiexec) in different prefix directory depending on which device I used<br>for testing.<br>Do I need to put some another options for installing MPICH2 or running<br>
the parallel program with shared memory?<br>Is there anything I missed?<br>I really appreciate for your concern.<br>Thank you.<br><br><br><br></div></div>
<div class="im">_______________________________________________<br>mpich-discuss mailing list<br><a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.gov</a><br><a href="https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss</a><br>
</div></blockquote></blockquote></div><br>