Hardware accelerated Scalable Parallel Random Number Generators for Monte Carlo methods

Abstract

Monte Carlo methods often demand the generation of many random numbers to provide statistically meaningful results. Because generating random numbers is time consuming and error-prone, the Scalable Parallel Random Number Generators (SPRNG) library is widely used for Monte Carlo simulation. SPRNG supports fast, scalable random number generation with good statistical properties. In order to accelerate SPRNG, we develop a hardware accelerated version of SPRNG that produces identical results. To demonstrate HASPRNG for Reconfigurable Computing (RC) applications, we develop a Monte Carlo pi-estimator for the Cray XD1 and XUP platforms. The RC MC pi-estimator shows 8.1 times speedup over the 2.2 GHz AMD Opteron processor in the Cray XD1.