Next: Functions for parallel execution over a cluster of machines, Previous: Installation hints, Up: General documentation for the parallel package for Octave [Index]
Explicit local parallel execution, with the intent to exploit more than one local processor(-core), is performed by calling a user-defined function in parallel with several different arguments. This is done in parallel processes. Note that there is no way to pass global variables to or from these processes.
The interface of the functions for local parallel execution is similar
to Octaves cellfun
and parcellfun
functions.
Note that some operations in Octave, particularly some matrix operations, may already be performed in parallel threads. This may limit the advantage yielded by explicit local parallel execution. Also, RAM access can be a bottleneck which limits computation speed of multicore computing.