Closed yousefmoazzam closed 2 months ago
To address the problem of if the value given to the --frames-per-chunk
flag exceeds the length of the slicing dim (ie, the chunk shape exceeds the boundaries of the data to be written), when this happens, the frames per-chunk value used is 1 instead of the flag value.
If other behaviour is preferred, feel free to mention, I'll open this for review now.
Actually by enabling export OMPI_MCA_fcoll=dynamic
fixed the PMIX related issue and I can confirm that I can run with more than one frame per chunk.
Built on top of work done in #271
This adds the
--frames-per-chunk
flag which enables the ability to:A value of 0 will turn chunking off (ie, write the hdf5 data as contiguous), and values > 0 will be the number of frames per-chunk.
Acceptance criteria checklist: