Dear HDF5 users,
I’m trying to write on a single dataset only a certain part of a whole domain, which involves just a few processors. It is there any possibility write chunked dataset of only a partial number of processors?
If the whole domain is 4x4x8 and y want to save a 4x4x2 from the end of the domain, it is possible to do?
To clarify: you are to run a job scheduled on 4x4x8 = 64 processors and you want only quarter of them: 16 to participate in a collective IO call, while the other processes doing something else: [something useful | busy wait | ???], then you want to synchronise the processes and MPI exit?
That’s exactly what I want. Only a part of the domain it’s needed to be written so the other processes which don’t involve these information should be idle in the collective IO call. I tried to make an independent call with the processes involved but the chunk definition it is not possible due to the final dataset dimension.
Thank you in advance
