Hawkynt

XorWow

XorWow combines George Marsaglia’s XorShift algorithm with a Weyl sequence for improved statistical properties. It has a period of 2^192-2^32 and is used as the default PRNG in NVIDIA’s CUDA cuRAND library for GPU-accelerated random number generation.

Properties

Property	Value
Category	Random Number Generators
Sub-category	Deterministic PRNG
Security status	🎓 Educational Only
Complexity	Intermediate
Inventor	George Marsaglia
Year	2003
Origin	🇺🇸 United States
Source	algorithms/random/xorwow.js

Parameters

Parameter	Supported values
Seed sizes	1 byte (8 bits) to 24 bytes (192 bits)

Capabilities

Flag	Value
IsDeterministic	Yes
IsCryptographicallySecure	No

Security

Status: 🎓 Educational Only

No vulnerabilities are recorded for this implementation.

Documentation

• NVIDIA CUDA cuRAND Documentation
• Original Paper: Xorshift RNGs (Marsaglia, 2003)
• Wikipedia: Xorshift
• TestU01 Results for XorWow

References

• CUDA Toolkit Source Code
• SimpleRandom Library (Reference Implementation)
• Random123 Library (Alternative High-Quality PRNGs)

Test vectors

4 vectors ship with this algorithm and run in the test suite. Byte values are hexadecimal.

Vector 1 — Seed (1): First 5 outputs - verified against C# implementation

Field	Value
seed	0000000000000001
outputSize	20
input	null
expected	acbed487b97556a081f6278d17aec601c87f3b27

Vector 2 — Seed (123456789): First 5 outputs - verified against C# implementation

Field	Value
seed	00000000075bcd15
outputSize	20
input	null
expected	f43c1487a09d56a0450e278d8d50c6016e23fb27

Vector 3 — Seed (0xDEADBEEF): First 5 outputs - common test value

Field	Value
seed	00000000deadbeef
outputSize	20
input	null
expected	3416e0475731bd202a80e40d8042ec81fa2c61e7

Vector 4 — Seed (1000000): First 5 outputs - large seed value

Field	Value
seed	00000000000f4240
outputSize	20
input	null
expected	780c3c87e80af6a073acc78df627660119333327

---

← All algorithms

This site is open source. Improve this page.