Parallel i/o

From openpipeflow.org
Revision as of 04:26, 20 July 2015 by Apwillis (talk | contribs) (Created page with "Currently statefiles are written from rank 0. Each process, however, reads (only) the data section it needs. NetCDF-4 has HDF5 support - it calls the appropriate HDF5 rout...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Currently statefiles are written from rank 0. Each process, however, reads (only) the data section it needs.

NetCDF-4 has HDF5 support - it calls the appropriate HDF5 routines that the user would call, and so overhead is nominal. Adding parallel i/o though netCDF would actually be simplification to the save routine, but since the serial write unlikely to be a bottleneck when using fewer than 1000 cores, and HDF5 installation can be tricy, I've not yet included the option.

Please contact me if you are interested in seeing this feature added.

Retrieved from ‘http:///index.php?title=Parallel_i/o&oldid=231’