Skip to main content

/

Docs

Nightly

v25.1

Log in

Mojo module

random

Random number generation for GPU kernels.

This module implements a high-performance random number generator using the Philox algorithm, which is designed for parallel and GPU computing. The Philox algorithm is a counter-based random number generator that provides high-quality random numbers with excellent statistical properties.

The main class is Random which generates both uniform random numbers and raw 32-bit integers. It supports:

  • Seeding for reproducible sequences
  • Multiple independent subsequences
  • Configurable number of rounds for quality vs performance tradeoff
  • Vectorized operations for efficiency

Example:

from gpu.random import Random
    rng = Random(seed=42)
    uniform_values = rng.step_uniform()  # Returns 4 random floats in [0,1)
    raw_values = rng.step()  # Returns 4 raw 32-bit integers
from gpu.random import Random
    rng = Random(seed=42)
    uniform_values = rng.step_uniform()  # Returns 4 random floats in [0,1)
    raw_values = rng.step()  # Returns 4 raw 32-bit integers

Structs​

  • ​Random: A high-performance random number generator using the Philox algorithm.