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
Virtual Memory Management and Peer Access
Launching Kernels
Kernel Families
Closures and Generics
Error Handling and Debugging

Safety on the GPU

The Safety Model
Bounds Checks: Pay Once or Opt Out

Async GPU Programming

The DeviceOperation Model
Combinators and Composition
Scheduling and Streams
Concurrent Execution
Overlapping Transfers and Compute

Async Host Pipeline Walkthrough

Project: Async MLP Pipeline

CuTe in Rust (preview)

Overview: cute-rs and ctm-rs (preview)
Support Matrix (preview)
Build with the CUTLASS Compiler
From Python CuTeDSL to cute-rs
Layouts, Tiling, and Partitions by Example
TMA Pipelines, Tensor Cores, and ctm-rs Interop
Checks and Obligations

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
Compiler Optimizations
Pliron Dialects
The Lowering Pipeline
Adding New Intrinsics
Catalog-Generated Intrinsics
Fuzzing and Differential Testing

Appendix

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

Search

Ctrl+K

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

Copyright © 2025-2026, NVIDIA.