-
Notifications
You must be signed in to change notification settings - Fork 66
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Support boost 1.69.0-1.70.0 #1296
Comments
I have found a bug in boost mpi, I have send a fix with this pull request boostorg/mpi#84 . I hope it will be accepted before the 1.70 release. |
unfortunately, boost mpi has still some issues. quite hard to reproduce because not appears all the times.
|
The problem comes from overtaking msg of some irecv when applying with std::vector (the same tag is used to send the size and data). But the MPI API says: Order Messages are non-overtaking. So it should work and it's the case almost every time. The bug detected is due to the fact that in partitionio reader, there are two steps of isend/irecv and the first one seems sometimes disturb the second step. Put a barrier between can fix also the issue. I finally fix the bug by replacing vector by an array because the size was known. I close the issue but we need to review our use of boost mpi (maybe it can appear in other part) |
thanks for the clear explanation |
I ran into the same issue, here is the backtrace
|
- can pass a mesh support instead of a mesh : allow to force some elements extracted (that are isolated in mesh support) to be ghost (can fix some special case of partitioning) - up mpi comm for #1296
Require some fixes :
The text was updated successfully, but these errors were encountered: