PRAND: GPU accelerated parallel random number generation library: Using most reliable algorithms and applying parallelism of modern GPUs and CPUs

TL;DR

PRAND is a GPU-accelerated library offering reliable pseudorandom number generators with parallel streams, optimized for modern CPUs and GPUs, enabling high-performance simulations.

Contribution

It introduces a comprehensive library with both single and multi-threaded generators, supporting massive parallel streams and SIMD optimization for modern hardware.

Findings

Supports up to 10^19 independent streams

Significantly improves performance on GPUs and CPUs

Includes reliable generators from recent research

Abstract

The library PRAND for pseudorandom number generation for modern CPUs and GPUs is presented. It contains both single-threaded and multi-threaded realizations of a number of modern and most reliable generators recently proposed and studied in [1,2,3,4,5] and the efficient SIMD realizations proposed in [6]. One of the useful features for using PRAND in parallel simulations is the ability to initialize up to $10^{19}$ independent streams. Using massive parallelism of modern GPUs and SIMD parallelism of modern CPUs substantially improves performance of the generators.