Dear HDF5 forum members,
I would like to share with you the following information and codes on parallel HDF5.
- collective vs. independent access pattern
- single dataset vs. one dataset per process
- array data per column vs. one data per column (in a dataset)
The codes will not compile because it lacks the other classes, but the concept can be applied to other applications using parallel HDF5.
Feel free to use them.
Thank you.
Regards,
Hoi Yen
parallelHDF5_collective_independent_rank_single.tar.gz (4.79 KB)
parallel_HDF5_collective_independent_rank_single.pdf (79.3 KB)