Just a report, and sort of a fix....<br><br>For parallel md runs of some large systems I get a segmentation fault when using 8 processors (but not 4).<br>Using fftw 3.2.2 and openmpi 1.3.3 and gromacs 4.0.5.<br><br>I tried different versions of openmpi and fftw, but the only thing that helped was using mpich2 instead :(<br>
<br>The main error seems to be "Signal code: Address not mapped (1)". <br>Searching for that, I found that it shows up on several websites involving parallel md and quantum packages, but no real solution is given. It might have to be inherent to openmpi or maybe some threading library?... <br>
<br><br>For reference, here is the output:<br><br>Back Off! I just backed up md.log to ./#md.log.27#<br>Reading file topol.tpr, VERSION 4.0.5 (single precision)<br>Making 1D domain decomposition 8 x 1 x 1<br><br>Back Off! I just backed up traj.trr to ./#traj.trr.27#<br>
<br>Back Off! I just backed up traj.xtc to ./#traj.xtc.27#<br><br>Back Off! I just backed up ener.edr to ./#ener.edr.27#<br><br>WARNING: This run will generate roughly 9919 Mb of data<br><br>starting mdrun 'Protein'<br>
5000000 steps, 10000.0 ps.<br>[node39:12247] *** Process received signal ***<br>[node39:12247] Signal: Segmentation fault (11)<br>[node39:12247] Signal code: Address not mapped (1)<br>[node39:12247] Failing at address: 0x2020bdeb0<br>
[node39:12247] [ 0] /lib64/libpthread.so.0 [0x2b9190d56a90]<br>[node39:12247] [ 1] mdrun(gmx_pme_do+0x5b5) [0x4ccab5]<br>[node39:12247] [ 2] mdrun(do_force_lowlevel+0x10f8) [0x4922c8]<br>[node39:12247] [ 3] mdrun(do_force+0x10fc) [0x4ebe2c]<br>
[node39:12247] [ 4] mdrun(do_md+0x4cfe) [0x4397fe]<br>[node39:12247] [ 5] mdrun(mdrunner+0x81f) [0x433fcf]<br>[node39:12247] [ 6] mdrun(main+0x3c5) [0x43b045]<br>[node39:12247] [ 7] /lib64/libc.so.6(__libc_start_main+0xe6) [0x2b9190f82586]<br>
[node39:12247] [ 8] mdrun [0x41c3e9]<br>[node39:12247] *** End of error message ***<br>--------------------------------------------------------------------------<br>mpirun noticed that process rank 0 with PID 12247 on node node39 exited on signal 11 (Segmentation fault).<br>
--------------------------------------------------------------------------<br>Finished simulation<br>Tue Dec 1 15:56:15 CET 2009<br><br>