The article presents the software library PARMONC (an acronym of the PARallel MONte Carlo) developed for effective parallelization of a wide range of Monte Carlo method applications with the large computational complexity. The library uses native coarse-grain parallelism of the Monte Carlo method. The «core» of the library is a well-tested, fast and reliable long-period parallel pseudorandom number generator. The PARMONC is an easy-to-use tool to perform distributed stochastic simulation. It does not require any knowledge of the MPI language since the parallelization of complicated sequential programs is carried out in an easy way. The PARMONC enables one to scale Monte Carlo simulations to a practically infinite number of processors, the computational load being evenly distributed across all cores.
|
