DOCUMENTATION

Learn Miri

Start with a five-minute quickstart, take the full tour of the language, then move to first-class GPU programming — all in one language, no shader files, no FFI.

▲ 5 MIN

Quickstart

Install the compiler, write hello.mi, and run your first program. The fastest path from zero to a running Miri binary.

Install & run →
◆ GUIDE

Language Guide

The complete tour — variables, functions, pattern matching, structs, traits, generics, the memory model, and modules. Read start to finish or jump to a topic.

Tour the language →
⛁ GPU

GPU Programming

Device-resident bindings, forall kernels, on-device reduction, shared memory, atomics, and interactive gpu frame loops. Starts from your first kernel — no GPU background assumed.

Write a kernel →
▶ LIVE

GPU Playground

See Miri GPU programs running live in your browser — Mandelbrot, fluid, particles, a black hole, and more, each compiled to WebGPU from Miri source.

Open the playground →

Browse by topic

Jump straight to any part of the reference.

Language basics
Variables — let & var Functions Control flow Pattern matching Strings
Types & data
Structs Enums Collections Queue & Stack Option types Result<T, E>
Memory & ownership
Memory model Cloneable & .clone() Resource types & drop Use-after-move The out keyword
Abstraction
Classes Traits Closures Generics Imports & modules
Standard library
system.text — regex system.json system.fs — files system.os — env & args system.time
Tooling & testing
Assertions Attributes The miri test runner Memory diagnostics Multi-file projects
GPU essentials
The residency surface Launching kernels Getting data back The four cost classes Cookbook
GPU advanced
On-device reduction Shared memory & barriers Atomics Interactive frames Building for the browser

Looking for a short answer rather than a guide — what works today, which GPUs are supported, how memory is managed? Read the FAQ →