Evaluation of Pseudo-Random Number Generation on GPU Cards

Monte Carlo methods rely on sequences of random numbers to obtain solutions many problems in science and engineering. In this work, we evaluate the performance different pseudo-random number generators (PRNGs) Curand library a modern Nvidia GPU cards. As numerical test, generate (PRN) non-uniform distributions using acceptance-rejection method. We consider GPU, CPU, hybrid CPU/GPU implementations. For additionally two implementations host device application programming interfaces (API). study how depends implementation parameters, including threads per block blocks streaming multiprocessor. To achieve fastest performance, one has minimize time consumed by PRNG seed setup state update. The duration increases with threads, while update decreases. Hence, is achieved optimal balance these opposing effects.