I came across some issues with newer OpenMPI versions, e.g. 5.0.x. After valgrind-ing the code I tracked it down to some issue with the Zoltan_PartDist_MPIOp created in lb_balance.c:865. After reviewing this part of the code I implemented a version defining a separate MPI data type available in #9 . This seems to fix the problem, but I'm not sure if this is the best possible fix.
Has anyone else encoutered problems like this?
I came across some issues with newer OpenMPI versions, e.g. 5.0.x. After valgrind-ing the code I tracked it down to some issue with the
Zoltan_PartDist_MPIOpcreated in lb_balance.c:865. After reviewing this part of the code I implemented a version defining a separate MPI data type available in #9 . This seems to fix the problem, but I'm not sure if this is the best possible fix.Has anyone else encoutered problems like this?