Kan, could you install callgring /vallgrind/kcachegrind and run this command :
valgrind --tool=callgrind ./yade-svn1258 -n -- -f /home/bruno/YADE/data/scene.xml -v 1000 -p -m 2 with scene.xml = your 100k spheres simulation. It will make only two iterations then stop, and you will get a nice file in /bin that you can open with kcachegrind. It will show which function takes all this time. And yes, I suspect physicalActionContainer too. Bruno p.s. saving/loading in binary format would be faster probably. Václav Šmilauer a écrit : >> I also need to do large scale simulation which may need to do around >> 500,000 for field engineering case. but instead of using YADE I have >> to develop my own code to do this, :(. >> In my opinion, the CPU computing time is the most important issue in >> DEM, while the memory is determined by the total particle number, and >> you data-structure design, 2GB can do a lot for a 2D case, for 3D case >> you may need more memory. but the momory requirements will not change >> much when your total particle number is defined. >> > Since Bruno just began hunting bottlenecks, Yade could be useful even > for 5e5 - in the current state, one iteration would take about 5 seconds > (by extrapolation). > >> >> hi, Vaclav, >> is it possible that the 5 mins used for packing or sample preparation >> to fill the container? I just guess, for my case, it takes some time >> for this step and then enter the cycling. >> > No, I was loading that file from .xml and it took that time after > starting the simulation that had been already loaded. I think Janek > (ping Janek ;-) ) could comment on this, the PhysicalAction containers > can take a lot of time to initialize. > > Vaclav > > _______________________________________________ > Yade-users mailing list > [email protected] > https://lists.berlios.de/mailman/listinfo/yade-users > > -- _______________ Chareyre Bruno Maitre de conference Institut National Polytechnique de Grenoble Laboratoire 3S (Soils Solids Structures) - bureau E145 BP 53 - 38041, Grenoble cedex 9 - France Tél : 33 4 56 52 86 21 Fax : 33 4 76 82 70 43 ________________ _______________________________________________ Yade-users mailing list [email protected] https://lists.berlios.de/mailman/listinfo/yade-users
