mcrand

What problem does it solve?

mcrand provides efficient GPU-based random number generation, enabling fast statistical computation and simulation workflows that rely on high-quality randomness.

Core Features & Use Cases

• Fast GPU RNG: Generate uniform, normal, log-normal, and Poisson-distributed random numbers directly on GPU for AI training, Monte Carlo simulations, and scientific computing.
• Flexible Algorithms: Supports multiple high-quality generators like XORWOW, MRG32k3a, Philox, Sobol, and more, suitable for various precision and reproducibility needs.
• Use Case: Imagine running massive Monte Carlo simulations for financial modeling or Bayesian inference, with rapid and reproducible random data streams.

Quick Start

Use mcRAND to generate normal distribution samples on GPU by creating a handle, setting a seed, and calling the appropriate function with your data array.