Hi number crunchers, here's a little hack that might even be useful, but before I turn it into a function package I would be happy to hear comments. From a user's point of view it's supposed to be a tool to run a numerically heavy, long-running function in parallel on the CPU's different cores with MPI, without knowing anything about MPI. (Well, it has to be installed and ready to use...) The limitation is that the function _must_ have a matrices array as the return type. (But note that a single matrix can be wrapped in a one-element array, and a single number is a 1x1 matrix; so basically all functions with numerical results should be wrappable. In the future such wrappers may even be auto-created, but don't hold your breath...) Another limitation is that since the function can have any number of arguments, specifying and handing them over to my tool involves specifying a suitable bundle and a strings array, which may look a little cumbersome. I hope the attached test script clarifies the usage. As I said, comments welcome. cheers sven