>BTW, Sega, I wanted to ask you independently: could you document your >findings on advantages of parallelizing? Did you parallelize >constitutive law as well, etc?
Yes, I will do it, but not as quickly as hoped (most likely, after the new year), because in December I have a PhD and now I am busy with this basically. Now I have some observations but need more tests. Raw data is: 1. Previously I report about 0.44 iter/sec for 90k particles with ParallelEngine and about 0.35 iter/sec without ParallelEngine. But these data were obtained for a static heap problem, i.e. where there is no intense change the position of particle relative to one another . 2. For dynamic problem (a tumbling mill) I get follow: a) with ParallelEngine - 0.34 iter/sec b) without ParallelEngine - 0.33 iter/sec c) Parallel simulation with OpenMP - 0.4 iter/sec So... It seems that the program needs not so much in MetaEngine cycles parallelization, but needs optimizing algorithms, such as an algorithm for determining collisions... Another observation: when parallelization with OpenMP a processor load is only about 120%, at the same time EDEM load a processor for all 190%. I think this shows the need for the engines parallelization... -- Best regards, Sergei D. _______________________________________________ Mailing list: https://launchpad.net/~yade-users Post to : [email protected] Unsubscribe : https://launchpad.net/~yade-users More help : https://help.launchpad.net/ListHelp
