On Thu, May 24, 2007 at 06:05:03PM +0300, Tahir Malas wrote: > Each process in our parallel FORTRAN > 90 program holds an identical array that is used in all parts of the > program. However, when the size of the problem gets larger and larger, this > memory cost has started to become a memory bottleneck for us.
This is actually a fairly frequent question. Some people use hybrid MPI+OpenMP in this situation. However, another way to attack it is to create a shared memory segment and put this array into it. Or, alternately, you can mmap() a file into all the processes with this data. -- greg _______________________________________________ Beowulf mailing list, [email protected] To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
