Closed rhyspoulton closed 4 years ago
This update seems to require too many updates. The H5OutputFile class could handle all of this internally where if the mpi_write_comm is of size 1, then do not open up the file in parallel.
The IO and SO should be separate pull requests. Also the description of the update in the SO is incorrect. The loop always started at the centre of the halo but only began searching for overdensity points out past a small number of particles in order to save checking all particles. This could result in the issue outlined. Now code simple starts searching for overdensity points at 2nd particle from the centre.
This update is required since the code would fail if MPI_number_of_tasks_per_write==1 and NMPI>1, since the code would only create 1 file but would attempt to write to NMPI files causing it to crash.