On Thu, 2 Aug 2007, Fernando Souto wrote: > Im trying to run some simulations using a linux-cluster system. Ive been > trying to find out what is the optimal number of processors to run my > simulations, but so far, the fastest simulation was performed by a single > processor followed by 8. Does anyone know if there is an optimum number of > processors? I have also tried to run the benchmark command included in meep,
The optimum number of processors depends not only on your hardware, but on the problem size (and various other details of the problem). As a rule of thumb, the bigger your problem is, the more processors it will parallelize over. If the problem is too small, on the other hand, then the computational parallelism of adding more processors is outweighed by the increased communication costs. Regards, Steven G. Johnson _______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
