Open nicolasaunai opened 1 year ago
@nicolasaunai did you find a solution for this? I am having the same issue.
@aslangil not really. I actually got it last week again, in my case it seems to occur only for simulations for which there are a few patches per processes. Do you see the error disappear if you use fewer processes OR larger domain?
When I use a single node (less processor) or use multi node but just 2 AMR levels instead of 3 (kind of larger domain), it dissappears. So my experience is also leading to the similar conclusion with you.
I didn't really investigate, but my (limited) understanding is that finding a patch decomposition and its distribution across (too) few MPI processes could be too constrained to find a solution and not all corner cases may not be well covered... I only saw these crashes in "test runs" that are abnormally small.
I regularly see this message as the last of my logs before simulations appear to be crashed or stalled :
I only have this message for simulations running with the TileClustering, never with BergerRigoutsos.
any hint regarding what may be the source of this?