What question did this study set out to answer?

This study aims to evaluate and compare the statistical performance of various modern pseudo-random number generators (PRNGs) used in computing.

May 22, 2026Open Access

Assessing the Stochastic Properties of Modern Pseudo-Random Generators for Parallel Computing

View Full Paper

Ask AI

Bookmark

This study aims to evaluate and compare the statistical performance of various modern pseudo-random number generators (PRNGs) used in computing.
Tested over 10^3 streams of multiple PRNGs including Xoshiro, Philox, PCG, and MRG32k3a.
Utilized the BigCrush battery from the TestU01 library over 4.5 years of cumulative computing time.
Conducted all tests under consistent conditions to simulate real-world usage.
The highest success rate among the tests was 72%.
Most generators failed multiple tests, with detailed documentation of failures provided.
All test results are made available and reproducible through a git repository.

No takes yet. Share an insight, caveat, or question.

Randomized trial evaluates statistical performance of various PRNGs in high-performance computing, suggesting implications for AI applications.

Wartel et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark

Randomized trial evaluates statistical performance of various PRNGs in high-performance computing, suggesting implications for AI applications.