The open-source Rust CUDA project has been “rebooted” to get back onto the effort of allowing NVIDIA CUDA compute kernels to be coded within the Rust programming language.
The Rust CUDA project is on enabling CUDA kernels on NVIDIA GPUs via Rust code and in turn that compiler targeting NVIDIA’s NVVM IR as a derivative of LLVM IR that is focused on GPU compute kernels and in turn translated to PTX code for NVIDIA GPU execution.
There hasn’t been a new release of this open-source Rust CUDA project since its v0.3 milestone in February 2022, but three years later the developers are getting back to work. They are also hoping to find more open-source developers to work on this project. The developers hope to eventually move at least some shared assets of Rust CUDA with the Rust GPU project that is focused on targeting SPIR-V / Vulkan. There’s also the possibility of greater collaboration with rustc’s PTX back-end and cudarc as a CUDA programming language abstraction within Rust.
The Rust CUDA developers are hoping to merge outstanding pull requests soon, updating dependencies, launching a new project web-site, and take care of other items on their TODO list. Medium-term work includes Rust and C++ interoperability, PTX back-end collaboration, and the Rust GPU project collaboration. Longer term they plan to explore Rust compiler integration, a unified GPU API for Rust, and evolving the Rust language to better support GPU programming.
Those wanting to learn more about this reboot to the Rust CUDA project can do so via the Rust-GPU GitHub blog by Christian Legnitto.
