Replies: 1 comment 7 replies
|
@zivlie Would you be willing to share the script (and the associated data) that's giving you the error? |
|
Sorry for the waffling - while this doesn't throw the segmentation fault, it now seems like the mpirun never actually starts. Please see the attached files (note extensions changed to .txt for upload but the script refers to extant_ranges_sp.nex and c-ir-comb-scaled.tre). |
|
Sorry for the late reply @zivlie. The script you sent fails with the following error, at least for me: Do you still experience issues once this has been fixed? |
|
Hi @davidcerny, oops, looks like I uploaded the wrong version of the script. But yes, the problem persists and is also the same for other scripts that have worked in the past, e.g., the FBD that you also recently helped me out with. |
|
I'm afraid that might be a problem with your MPI setup, then. The script from #776 runs just fine for me on 4 cores using an MPI executable compiled from the current tip of the development branch (4483f3f).
I don't think so. The issues mostly had to do with checkpointing and MC^3, and none of them would have prevented you from running the analysis in the first place. Maybe try using your MPI executable to run an integration test, e.g., |
|
OK, I suspect you are right! I'll use the test you provided and then look into fixing up my MPI setup. |
Uh oh!
There was an error while loading. Please reload this page.
Hi again,
Sorry for the many subsequent questions! Not sure if this ought to go to Issues or if I'm just missing something.
I just downloaded the development branch and had no issues with cmake for rb and rb-mpi. However, when I try to use mpirun with more than 1 process, I get a segmentation fault (see below). The script runs just fine with -np 1, or with the non-mpi RB.
I know that there were some other issues with mpi that seem to have been fixed recently. Could I have gotten the wrong branch? I just used the instructions to git clone --branch development.
I have Open MPI v4.1.2.
[Apologies for the long path to rb-mpi making this a bit hard to read!]
All reactions