Crate oxicuda_rand 

OxiCUDA Rand – GPU-Accelerated Random Number Generation

This crate provides GPU-accelerated random number generation, serving as a pure Rust equivalent to NVIDIA’s cuRAND library. It supports multiple RNG engines (Philox, XORWOW, MRG32k3a), various distributions (uniform, normal, log-normal, Poisson), and quasi-random sequences (Sobol).

Engines

• Philox-4x32-10 – Counter-based PRNG, the cuRAND default.
• XORWOW – XORshift with Weyl sequence, fast and suitable for most Monte Carlo workloads.
• MRG32k3a – Combined multiple recursive generator with the highest statistical quality.

Distributions

• Uniform [0, 1) in f32/f64
• Normal (Gaussian) via Box-Muller transform
• Log-normal via exponentiation of normals
• Poisson (Knuth for small lambda, normal approximation for large lambda)

Quasi-random

• Sobol sequences for low-discrepancy sampling in Monte Carlo integration

Re-exports

pub use error::RandError;

pub use error::RandResult;

pub use generator::RngEngine;

pub use generator::RngGenerator;

pub use graph_gen::AdjacencyList;

pub use graph_gen::BarabasiAlbertGenerator;

pub use graph_gen::ErdosRenyiGenerator;

pub use graph_gen::GraphStats;

pub use graph_gen::GraphType;

pub use graph_gen::RandomRegularGenerator;

pub use graph_gen::StochasticBlockModelGenerator;

pub use graph_gen::WattsStrogatzGenerator;

pub use matrix_gen::CorrelationMatrixGenerator;

pub use matrix_gen::GaussianMatrixGenerator;

pub use matrix_gen::MatrixLayout;

pub use matrix_gen::OrthogonalMatrixGenerator;

pub use matrix_gen::RandomMatrix;

pub use matrix_gen::SymmetricPositiveDefiniteGenerator;

pub use matrix_gen::WishartGenerator;

pub use monte_carlo::BlackScholesParams;

pub use monte_carlo::HamiltonianMC;

pub use monte_carlo::McmcResult;

pub use monte_carlo::MetropolisHastings;

pub use monte_carlo::MonteCarloConfig;

pub use monte_carlo::MonteCarloResult;

pub use monte_carlo::SamplerState;

pub use quasi::HaltonGenerator;

pub use quasi::LatinHypercubeSampler;

pub use quasi::ScrambledSobolGenerator;

pub use quasi::SobolGenerator;

Modules

distributions

Distribution transform utilities for GPU random number generation.

engines

RNG engine implementations.

error

Error types for OxiCUDA Rand operations.

generator

High-level RNG generator wrapping engine PTX generators.

graph_gen

Random graph generation for GNN workloads and network analysis.

matrix_gen

Random matrix generation for statistical applications.

monte_carlo

Monte Carlo simulation primitives.

prelude

Prelude for convenient imports.

quasi

Quasi-random number generation for Monte Carlo methods.

statistical_tests

Statistical quality tests for OxiCUDA random number generators.