I’m trying to achieve parallel compressed input output for my Fortran simulation result.
As the last post about this topic is more than one year old, I would like to ask if anybody here achieved this goal.
I read here http://hdf-forum.184993.n3.nabble.com/Parallel-I-O-Compression-td4025790.html#a4025798 that utilizing third part compressione library, such as pigs, pzlib2 or pxz, parallel compression writing seems to be possible.
My pattern is not very complex:
each core has part of a variable, and this part are sequential (can be read with hyperslab) i.e
* core 1 from 1 to 100
* core 2 form 101 to 200
* and so on
they also need to be written in this order inside the dataset.
I also read about the possibility to create a plugin utilizing H5Z_register but it’s not clear if have to do this before install HDF5 in the system, or just when i compile my application.
Also, my C is a bit rusty, so i should work on hit just to do this thing.
Can anybody help on this?