search

cwpearson / stencil

A prototype MPI/CUDA stencil communication library
Boost Software License 1.0
10 stars 3 forks source link

Periodic hang during test/test_cuda_mpi in NodeAware Placement #13

Closed cwpearson closed 4 years ago

cwpearson commented 4 years ago

Hang is somewhere in here, based on last message from rank 0

https://github.com/cwpearson/stencil/blob/880272c196e598f7edaa9f656e069645c2a41395/include/stencil/partition.hpp#L653-L701

cwpearson commented 4 years ago

This may have been corrected in be6cc1fcbfa9389be6127e5144753d4f9c865596. The problem could have been that not all ranks were using the same placement strategy.