Steven G. Johnson wrote: > On Thu, 29 Nov 2007, João Luís Silva wrote: >> I've found that the parallelization of my simulations is being limited by >> add-flux. I'm using a dual Xeon 5335, and for the appended example it >> takes > > The problem is that your computation time is apparently dominated by the > flux computation (at every time step, it has to accumulate the field > Fourier transforms at every point in the flux region). The flux > computation is parallelized, but only in the sense that each processor > computes the Fourier transforms for flux points in its "own" chunk of the > grid. If the division of the grid between processors (into roughly equal > chunks) allocates most of the points where the flux is computed to one > processor, it is not going to parallelize.
Ok, now I can see why it seems to work serially. My simulations are for optical phenomena, and they need quite a few points to get a good spectra. I will just stop using add-flux and get the same information by post-processing the hdf5 files. I assumed add-flux to be an inexpensive operation and was surprised to find that it was limiting the performance of my simulations in such a large scale. Implementing a load balancing method that works properly in all cases would be nontrivial. Maybe some sort of warning in the documentation about add-flux computational requirements would be appropriate? Best regards, João L. Silva _______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
