Generalized map reduce for parallel computers (not distributed computing). Can process in parallel an infinite stream of elements.
Can process a very large file in parallel on a multicore computer; provided there is a way to cut your file into independent blocks (the 'demux' function). The processing function is called 'work'. The function gathering the results is called 'mux'. The number of processors running your computation in parallel is called 'nprocs'. The chunk size (number of items) processed by one call to the 'work' function is called 'csize'.
There is a minimalist Parmap module, if you want to switch to/from Parmap easily.
Read the corresponding ocamldoc before using.
USING THIS LIBRARY IN A MISSION CRITICAL, LONG-RUNNING SOFTWARE THAT IS NEVER SUPPOSED TO CRASH IS NOT ADVIZED. WHILE THIS LIBRARY IS HIGH PERFORMANCE, IT IS ALSO DANGEROUS. USE AT YOUR OWN RISKS.
Published: 28 Apr 2022
Reverse Dependencies (15)