From oliver.sander at tu-dresden.de Thu Nov 11 08:12:37 2021 From: oliver.sander at tu-dresden.de (Oliver Sander) Date: Thu, 11 Nov 2021 08:12:37 +0100 Subject: [dune-functions] Parallelization of GlobalBasis In-Reply-To: <2134c106-f1c1-40a1-0524-cf8216b16e23@tu-dresden.de> References: <2134c106-f1c1-40a1-0524-cf8216b16e23@tu-dresden.de> Message-ID: <0feef671-09d9-98f7-016f-991965fc833b@tu-dresden.de> Hi Simon, > So far so good. [The GlobalGridView] works fine and is implemented in the branch > feature/parallel-grid-view. thank you! I recently realized that this is what I need to replace a lot of hand-written goo in a central part of my code. Would you be able to open a merge request to make code review easier? Best, Oliver -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 5198 bytes Desc: S/MIME Cryptographic Signature URL: From simon.praetorius at tu-dresden.de Thu Nov 11 11:01:49 2021 From: simon.praetorius at tu-dresden.de (Simon Praetorius) Date: Thu, 11 Nov 2021 11:01:49 +0100 Subject: [dune-functions] Parallelization of GlobalBasis In-Reply-To: <0feef671-09d9-98f7-016f-991965fc833b@tu-dresden.de> References: <2134c106-f1c1-40a1-0524-cf8216b16e23@tu-dresden.de> <0feef671-09d9-98f7-016f-991965fc833b@tu-dresden.de> Message-ID: <1e99e3c0-2cc8-f437-87ee-c1b306a4aec7@tu-dresden.de> While the GlobalGridView seems to work (what ever this means), some other parts of the MR are more experimental and, e.g., the parallel poisson equation does not work correctly. Also, it is quiet inefficient, since the same communication needs to be performed multiple time to build up the parallel indexset. Essentially, to build up the GlobalIndexSet we need at least two communications over the grid and then the same communications again on the ParallelIndexSet and RemoteIndices. This is not very efficient. If the GlobalIndexSet is just used to identy uniquely all DOFs, it is not worth building this up. A better approach is something that can be computed locally, to have a global indentification of all DOFs, e.g. a global ID-Set for basis functions. With just the globalIndexSet one cannot directly build any parallel linear algebra. One needs more information about the overlap of DOFs and who has to communicate with whom. This is not encoded in the GlobalIndexSet, but in the RemoteIndices. A ScalarProduct also needs the information whether the local DOF should be added up to the global value or not. There are a lot of open questions. I do not have answers for all of them. Am 11.11.21 um 08:12 schrieb Oliver Sander: > > Hi Simon, > >> So far so good. [The GlobalGridView] works fine and is implemented in >> the branch >> feature/parallel-grid-view. > > thank you!  I recently realized that this is what I need to replace > a lot of hand-written goo in a central part of my code.  Would you > be able to open a merge request to make code review easier? > > Best, > Oliver > > > _______________________________________________ > dune-functions mailing list > dune-functions at lists.dune-project.org > https://lists.dune-project.org/mailman/listinfo/dune-functions -- Dr. Simon Praetorius Institut für Wissenschaftliches Rechnen Fakultät Mathematik Technische Universität Dresden Tel.: TUD-34432 Mail: simon.praetorius at tu-dresden.de Web: https://tu-dresden.de/mn/math/wir/das-institut/beschaeftigte/simon-praetorius -------------- next part -------------- An HTML attachment was scrubbed... URL: -------------- next part -------------- A non-text attachment was scrubbed... Name: smime.p7s Type: application/pkcs7-signature Size: 5204 bytes Desc: S/MIME Cryptographic Signature URL: