Skip to main content

Ctrl+K

cuda-oxide

GitHub

GitHub

Table of Contents

Getting Started

Installation
Writing Your First Kernel

Writing GPU Programs

The CUDA Execution Model
Kernels and Device Functions
Memory and Data Movement
Launching Kernels
Closures and Generics
Error Handling and Debugging

Safety on the GPU

The Safety Model

Async GPU Programming

The DeviceOperation Model
Combinators and Composition
Scheduling and Streams
Concurrent Execution

Building a Real Application

Project: Async MLP Pipeline

Advanced GPU Features

Shared Memory and Synchronization
Warp-Level Programming
Tensor Memory Accelerator (TMA)
Matrix Multiply Accelerators
Cluster Programming

Inside the Compiler

Architecture Overview
Pliron – Pliron IR (MLIR-like)
rustc_public – Stable MIR
The Code Generator: rustc-codegen-cuda
The MIR Importer
Pliron Dialects
The Lowering Pipeline
Adding New Intrinsics
Fuzzing and Differential Testing

Appendix

Building from Source
API Quick Reference
Supported Features
cuda-oxide vs CUDA C++
The Rust + GPU Ecosystem
Glossary

cuda-oxide vs CUDA C++

cuda-oxide vs CUDA C++#

Note

This chapter is under construction.

This appendix will offer a side-by-side comparison of cuda-oxide and CUDA C++.

previous

Supported Features

next

The Rust + GPU Ecosystem

Privacy Policy | Your Privacy Choices | Terms of Service | Accessibility | Corporate Policies | Product Security | Contact

Copyright © 2025-2026, NVIDIA.