Running Jaguar Jobs over Multiple Processors

There are two ways to use multiple processors to make Jaguar jobs run faster: distribution of multiple subjobs over processors and threading with OpenMP. The first method runs independent subjobs simultaneously, each with one processor, in an "embarrassingly parallel" fashion. This is common for many products and is the most efficient method. The second method uses multiple OpenMP threads to accelerate individual subjobs. Both strategies can be combined, and the -PARALLEL option provides automatic selection of the mix. Here, the term "processor" is used to cover any hardware unit that can run a process, such as a single core on a CPU unit.