IMPORTANT: To view this page as Markdown, append `.md` to the URL (e.g. /docs/manual/basics.md). For the complete Mojo documentation index, see llms.txt.
Skip to main content
Version: Nightly
For the complete Mojo documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /docs/manual/basics.md).

Random

struct Random[rounds: Int = 10]

A high-performance random number generator using the Philox algorithm.

The Philox algorithm is a counter-based random number generator designed for parallel computing. It provides high-quality random numbers with excellent statistical properties and works efficiently on both CPU and GPU.

Parameters​

  • ​rounds (Int): Number of mixing rounds to perform. Higher values provide better statistical quality at the cost of performance. Default is 10.

Implemented traits​

AnyType, Copyable, ImplicitlyDestructible, Movable

Methods​

__init__​

def __init__(out self, *, seed: UInt64 = UInt64(67280421310721), subsequence: UInt64 = UInt64(0), offset: UInt64 = UInt64(0))

Initialize the random number generator.

Args:

  • ​seed (UInt64): Initial seed value for reproducible sequences. Default is 0.
  • ​subsequence (UInt64): Subsequence number for generating independent streams. Default is 0.
  • ​offset (UInt64): Starting offset in the sequence. Default is 0.

step​

def step(mut self) -> SIMD[DType.uint32, 4]

Generate 4 random 32-bit unsigned integers.

Returns:

SIMD[DType.uint32, 4]: SIMD vector containing 4 random 32-bit unsigned integers.

step_uniform​

def step_uniform(mut self) -> SIMD[DType.float32, 4]

Generate 4 random floating point numbers uniformly distributed in [0,1).

Returns:

SIMD[DType.float32, 4]: SIMD vector containing 4 random float32 values in range [0,1).

step_uniform_unbiased​

def step_uniform_unbiased(mut self) -> SIMD[DType.float32, 4]

Generate 4 uniform float32 values in (0, 1), unbiased.

Uses all 32 raw bits of each Philox output via the conversion u = (raw + 0.5) / 2^32, which is fused into a single FMA. The resulting range is (2^-33, 1 - 2^-33) — never exactly 0 or 1. Compared to step_uniform, this uses one extra bit of randomness and is bounded away from 0, which removes the need for a 1.0 - u guard before log(u) in Box-Muller transforms.

Returns:

SIMD[DType.float32, 4]: SIMD vector containing 4 random float32 values in (0, 1).

View source