Closed samuelkgutierrez closed 3 years ago
@Lcjacobpd, this may be of interest to you.
I just realized that there is potentially a cleaner way of dealing with this issue. Let me see if I can write that version up. Please hold off on merging this request until further notice.
I just realized that there is potentially a cleaner way of dealing with this issue. Let me see if I can write that version up. Please hold off on merging this request until further notice.
Scratch that. It wasn't a clearer solution after all. Sorry about the noise. Please consider the original pull request.
Looks good to me. This is another case of "How did this go so long without being noticed?"
Looks good to me. This is another case of "How did this go so long without being noticed?"
Thank you, @cferenba. I agree!
Fix crashes when the number of MPI processes is greater than 1. The problem appears to come from taking the address of the first element of zero-length std::vectors during collective operations. The problematic arrays are located on MPI processes with MPI_COMM_WORLD_RANK != 0.