Open Snowthe opened 2 years ago
That's likely a serialization issue of those parameters through the MPI interface. The GML method on the dev branch has been completely rewritten with a modernized MPI structure. I'd suggest trying that. I haven't documented the all the new configuration variables since that's relatively new. I can send along my test configuration to give you the new variables, if it's helpful.
When trying to run a parallel optimization, i.e. with OstrichMPI.exe (I tried both the GML and ParaPADDS algorithm) where constant parameters are included in the tied parameters section, Ostrich crashes immediately after creating the worker folders, and before creating the model files from the template files, with the following error message:
"OstrichMPI.exe ended prematurely and may have crashed. exit code 0xc0000374"
As soon as I remove all of the constant parameters, OstrichMPI.exe runs as expected.
Also, running a serial optimization with Ostrich.exe works just fine (again, I tried the GML and PADDS algorithm) - with and without constant parameters. Hence, I suspect the above to be a bug?