Shang Wang 写道: > Hi Steven and other Meep users, > > We are new to Meep. Right now the most concern for us is the simulation > time and memory usage. > It seams one single server cannot handle very large simulation. > For example: the simulation of a 600 x 20 x 900 structure takes ~8-10 > hours to finish on a 4GB RAM server. (I don't know if that's a typical > number for other users, but we think it's large.) But actually I need to > simulate a structure 5 times larger than that. In that case, the > simulation just cut down by itself. I guess that's because we don't have > enough memory. > > So my question is: Is there anyway to reduce the simulation time or > memory usage? Or we can only upgrade our machine or use computer cluster? > > Other than that, can I break the structure into several pieces, and > simulate it separately? In that case, how to specify the sources? > Actually, that's another problem that bothers me a lot: Can I specify > the source using E-field or H-field? Or can I use one simulation's > output as another simulation's input? Thank you very much! > > Best, > Shang > > _______________________________________________ > meep-discuss mailing list > [email protected] > http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss > > > Hi, One task consuming 1 GB Mems may take several hours. Your simulation time is on the same order of magnitude. :-) . Yet, there exist different ways to speed up your simulations. if you are simulating symmetry structures, then use symmetry in meep. If your computers are multi-core or larger clusters are available, then use meep-mpi to do what you say: " break the structure into several pieces, and simulate it separately ".
And the source question has been discussed many times in this thread, i.e., see http://www.mail-archive.com/[email protected]/msg02087.html or other replies. Zheng 2009-1-8 _______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
