Nvidia introduces CUDA-oxide, a Rust-to-CUDA compiler for GPU programming

May 11, 2026

AI Summary

CUDA-oxide is an experimental compiler that allows developers to write GPU kernels using Rust, compiling directly to PTX without the need for additional languages or bindings. The initial alpha release is intended for testing and feedback as it is still in development.

CUDA-oxide enables the writing of Single Instruction, Multiple Threads (SIMT) GPU kernels in Rust, focusing on safety and idiomatic usage.

The compiler compiles standard Rust code directly to PTX, eliminating the need for domain-specific languages or foreign language bindings.

The v0.1.0 release is an early-stage alpha version, which may contain bugs and incomplete features, and the developers encourage user feedback to guide future improvements.

Developers can use Rust's type system and ownership model to write GPU kernels, with safety being a primary goal despite the complexities of GPU programming.

The tool allows for the creation of lazy DeviceOperation graphs for GPU tasks, enabling scheduling across stream pools and awaiting results using Rust's async features.

nvidiarustcudacompilerprogramming

Nvidia introduces CUDA-oxide, a Rust-to-CUDA compiler for GPU programming

Related Stories

Exploring the Relevance of Python in an AI-Driven Coding Landscape

Digg Relaunches as AI-Focused News Aggregator After Previous Shutdown

OpenAI Launches Campus Network for Student Clubs

Frame Security raises $50 million to address human vulnerabilities in cybersecurity