Hi there, FYI [1] is showing a concrete MPI execution (with just two threads...). The MPI run is approximately 10x slower than the reference non-MPI run on my laptop, which is encouraging given the lazy implementation (e.g. large excess of useless data going through messages with serialization/deserialization on both sides). I'm currently extending to N threads, which should only need little changes.
This is you last chance to join yade-mpi team since next messages will go there directly. :) Cheers Bruno [1] https://github.com/bchareyre/yade-mpi/blob/master/examples/mpi/testMPI.py On 8 August 2018 at 09:25, Eulitz, Alexander < alexander.eul...@iwf.tu-berlin.de> wrote: > Hi, > I am really looking forward to such feature of Yade. This is an important > step for broadening the field of possible applications of the software. For > us in Berlin it would be a great thing if it enables us so simulate some > hundred tousands of particles in our manufacturing processes. > If you need some assistance besides programming let me know. I hope, I can > contribute somehow. At the moment we applied for some funding in that > direction, hopefully we will be successfull and can support the project a > bit. > Best regards > Alex > > -----Ursprüngliche Nachricht----- > Von: Yade-dev [mailto:yade-dev-bounces+alexander.eulitz=iwf.tu- > berlin...@lists.launchpad.net] Im Auftrag von Bruno Chareyre > Gesendet: Donnerstag, 19. Juli 2018 18:50 > An: yade-dev; Kneib François; Václav Šmilauer; Pierre-Antoine Bouttier > Betreff: [Yade-dev] yade-MPI > > Hi all, > Hopefully there will be some progress in the near future regarding a > distributed memory parallelization of Yade. > For the moment the preferred approach is to handle this all at the > python level using mpi4py (porting some critical functions to c++ might > be considered later). > Some changes (hopefully limited) are necessary on the c++ side > nevertheless, and I just forked a branch to this purpose [1]. > You can find main ideas in the wiki of this branch [2]. > > This project is still very much in a brainstorming stage, every comment > or question are welcome (especially if it helps anticipating > shortcomings of some decisions!). > You can subscribe to the yade-mpi mailing list if you'd like to discuss > it or just keep track of the progress. > > Cheers > > Bruno > > https://github.com/bchareyre/yade-mpi/ > https://github.com/bchareyre/yade-mpi/wiki > https://launchpad.net/~yade-mpi > > -- > _______________ > Bruno Chareyre > Associate Professor > ENSE³ - Grenoble INP > Lab. 3SR > BP 53 > 38041 Grenoble cedex 9 > Tél : +33 4 56 52 86 21 > Fax : +33 4 76 82 70 43 > ________________ > > Email too brief? > Here's why! http://emailcharter.org > > > > _______________________________________________ > Mailing list: https://launchpad.net/~yade-dev > Post to : yade-dev@lists.launchpad.net > Unsubscribe : https://launchpad.net/~yade-dev > More help : https://help.launchpad.net/ListHelp > _______________________________________________ > Mailing list: https://launchpad.net/~yade-dev > Post to : yade-dev@lists.launchpad.net > Unsubscribe : https://launchpad.net/~yade-dev > More help : https://help.launchpad.net/ListHelp >
_______________________________________________ Mailing list: https://launchpad.net/~yade-dev Post to : yade-dev@lists.launchpad.net Unsubscribe : https://launchpad.net/~yade-dev More help : https://help.launchpad.net/ListHelp