Open bboutkov opened 6 years ago
Can you try a git revert a555464
and see if the same problem occurs? It looks like we do indeed probably have a (hard-to-replicate) bug there.
If that's the case though then you might be able to get the same error on introduction_ex4, as in #1697. If so then please chime in there; at the moment I have no idea why this bug is affecting some users but not others!
The problem in this issue is independent of a555464 and it still occurs if I rebase the branch mentioned in the original post on master.
With regards to the problems in #1697, I have not managed to reproduce the issues brought up there with intro/ex4 when using as was mentioned -np 2 ./example-opt -n 10 -d 3.
Hello, while doing some further testing for the upcoming GMG work in #1568 I came across a situation where I can reliably hit the PETSc error below while constructing some projection matrixes in a 3D scenario. I've managed to reduce it to an initial 2x2x2 mesh with 2 refinements and can trigger the error with -np 4.
Since it depends on still ongoing work I think the fastest place for me to demonstrate this in the modified fem_system/ex4 which I have been using which you can find at: proj_mat_3d_badmalloc
Please allow me to make the reminder that in order to trigger the section of the code we need to provide some command line options so the minimal ones should be something like SOLVER_OPTIONS="--use_petsc_dm -pc_mg_levels 3"
Any thoughts on the matter would be much appreciated. Thanks in advanced!
Error Below: