[mpich-discuss] Socket closed

Tim Kroeger tim.kroeger at cevis.uni-bremen.de
Mon Nov 16 04:08:23 CST 2009

Dear all,

The "Socket closed" problem that I discussed here two weeks ago is 
still not solved.  I thought about a memory leak, but this turned out 
to be not true; at least, when I print out the contents of 
/proc/meminfo regularly within my application, nothing appears to 
converge to zero therein.  Also, the crashes apprear rather 
predictibly after a fixed number of time steps.  This number does 
depend on the number of threads and number of nodes that I use, but it 
does not depend on the actual time step size, so that a crash due to 
numerical problems seems unlikely to me.

Are there any possibilities to let mpich2 do some more logging that 
allows me to find out what is going on?

Best Regards,


Dr. Tim Kroeger
tim.kroeger at mevis.fraunhofer.de            Phone +49-421-218-7710
tim.kroeger at cevis.uni-bremen.de            Fax   +49-421-218-4236

Fraunhofer MEVIS, Institute for Medical Image Computing
Universitaetsallee 29, 28359 Bremen, Germany

More information about the mpich-discuss mailing list