During recent CONUS baseline testing computation, it was discovered that when running ngen in MPI, the std output/screen dump write out the realization file multiple times. The number of output realization file equal to the number of MPI processing CPU cores used. This is redundant and also making examining the screen output more difficult.
It appears that this is not a CONUS issue, but occurs at any scale.
stcui007
commented
2 months ago
During recent CONUS baseline testing computation, it was discovered that when running ngen in MPI, the std output/screen dump write out the realization file multiple times. The number of output realization file equal to the number of MPI processing CPU cores used. This is redundant and also making examining the screen output more difficult.
It appears that this is not a CONUS issue, but occurs at any scale.
Current behavior
As describe above.
Expected behavior
A single copy of the realization file.
Steps to replicate behavior (include URLs)
Build a MPI executable and run a test.
Screenshots