Tailoring MPI I/O views to match sparse trace headers
S-01564
The trace metadata that needs to be read from or written to a file is only the metadata requested by the programmer. Some time and space might be saved by creating a file view that shows only the bytes needed to be written, or perhaps a smaller contiguous region than the whole header (e.g. from the lowest requested byte to the highest)
Reduce the library memory requirements by creating file views (e.g MPI-IO file views) that map to the exact size required to be read from the trace headers rather than reading either the entire header or some extent from the min to the max.
Do this by adding an object layer call that takes the actual size of the block being written which will be used to create a datatype for hindexed_block.
CREATED ON - 19/01/2017