Closed cjoycd closed 4 months ago
There's difference in the commands that one needs to send to GLVis in serial and in parallel -- make sure you are following one of the parallel examples in MFEM. For example, in serial, ex1
uses
sol_sock << "solution\n" << mesh << x << flush;
whereas in parallel ex1p
uses
sol_sock << "parallel " << num_procs << " " << myid << "\n";
sol_sock << "solution\n" << pmesh << x << flush;
Dear Dobrev,
Thank you for your reply. This issue happens for all standard "example/miniapp" problems available with the mfem package and also with my own codes (which were working fine previously). All these codes were working fine till I changed to a new OS (Ubuntu 24.04 LTS , gcc version 13.2.0 - 13.2.0-23ubuntu4) and with a fresh installation of [SDL2], [GLEW], [GLM], [Freetype], [Fontconfig] etc. Also, a new version of MFEM is installed (mfem-4.7). Previously it was version 4.6.
The issue is mainly due to improper MPI initialization. After the MPI initiation
Mpi::Init(argc, argv); const int myid = Mpi::WorldRank();
The values assigned/obtained for myid and Mpi::Root() are 0 and 1, respectively for all the processors.
This makes all sections being executed by all processors, irrespective of their rank ..
Regards --Sijoy
The issue has been resolved after a fresh MPI (mpich-4.2.2) installation manually. The issue was not related to GLVis or MFEM.
Thanks --Sijoy
Dear all,
I have recently installed GLVis on Ubuntu 24.04 (LTS) with the gcc version (Ubuntu 13.2.0-23ubuntu4) 13.2.0. The installation is done. However, while plotting outputs (live while running the code), it opens plot windows for each processor separately (duplicates the plot window equal to the number of processors used). Can someone help to fix the problem ?
Thank you in advance --Sijoy