How to Correctly Deal With Pseudorandom Numbers in Manycore Environments - Application to GPU programming with Shoverand

TL;DR

This paper introduces ShoveRand, a framework for reliable and efficient pseudorandom number generation on GPUs, addressing the lack of quality RNG implementations in GPU-based stochastic simulations.

Contribution

It presents a general framework for high-quality RNGs on GP-GPU platforms, facilitating reliable stochastic simulations with an easy-to-use interface.

Findings

ShoveRand provides efficient RNG implementations on GPUs.

The framework simplifies RNG development for GPU applications.

Experimental results show improved performance and reliability.

Abstract

Stochastic simulations are often sensitive to the source of randomness that character-izes the statistical quality of their results. Consequently, we need highly reliable Random Number Generators (RNGs) to feed such applications. Recent developments try to shrink the computa-tion time by relying more and more General Purpose Graphics Processing Units (GP-GPUs) to speed-up stochastic simulations. Such devices bring new parallelization possibilities, but they also introduce new programming difficulties. Since RNGs are at the base of any stochastic simulation, they also need to be ported to GP-GPU. There is still a lack of well-designed implementations of quality-proven RNGs on GP-GPU platforms. In this paper, we introduce ShoveRand, a frame-work defining common rules to generate random numbers uniformly on GP-GPU. Our framework is designed to cope with any GPU-enabled development platform and to expose a straightfor-ward interface to users. We also provide an existing RNG implementation with this framework to demonstrate its efficiency in both development and ease of use.