adam-sim-dev
adam-sim-dev
Hi @psocratis , I also see the same issue on a cluster with mfem-4.5, mumps-5.5.1. I just test with ex25p. The number of finite element unknowns is only 14544, so...
FYI, MUMPS-5.2.1 works. @psocratis
FYI, mumps-5.2.1 works. The possible bugs may be in the MFEM_MUMPS_VERSION >= 530 code part. @psocratis
Yes, but just give me some time
I cannot reproduce the problem on a small workstation using Intel and GCC compilers. I have some difficulty obtaining the stack trace on that cluster because I am not so...
mumps.cpp Line 213 213 id->job = 3; (gdb) 214 dmumps_c(id); (gdb) Thread 1 "ex25p" received signal SIGSEGV, Segmentation fault. __memcpy_generic () at ../sysdeps/aarch64/multiarch/../memcpy.S:198 198 ../sysdeps/aarch64/multiarch/../memcpy.S: No such file or directory.
Hi @jandrej ， just ask a related usage question. If I can get a SubMesh or ParSubMesh, how can I add a bounding box outside the SubMesh and then construct...
The usage situation is that a SubMesh is only a small (refined) Mesh, which does not fulfill the Dirichlet boundary conditions on the boundary, so I need to add some...
I forget the specific command lines. I also can not reproduce the problem on my shared-memory workstation but on a distributed-memory cluster.
Hi @sebastiangrimberg , MUMPS depends on parmetis only for both parallel and sequential analysis. So parmetis may be not required.