Summary, from someone who uses CUDA on rust in several projects (Computational chemistry and cosmology simulations):

  - This lib has been in an unusable and unmaintained state for years. I.e., to get it working, you need to use specific, several-years-old variants of both rustc, and CUDA.
  - It was recently rebooted. I haven't tried the Github branch, but there isn't a release yet. Has anyone verified if this is working on current Rustc and CUDA yet?
  - The Cudarc library (https://github.com/coreylowman/cudarc) is actively maintained, and works well. It does not, however, let you share host and device data structures; you will [de]serialize as a byte stream, using functions the lib provides. Works on any (within past few years at least) CUDA version and GPU.

I’ve been using the cudarc crate professionally for a while to write and call cuda from rust. Can highly recommend. You don’t have to use super old rustc versions. Although I haven’t looked exactly what you do need to use recently.

Very cool to see this project get rebooted. I'm hoping it will have the critical mass needed to actually take off. Writing CUDA kernels in C++ is a pain.

In theory, since the NVVM IR is based on LLVM IR, rust in CUDA should be quite doable. In practice, though, of course it is an extreme amount of work.

Shouldn't it be called RUDA?

For this to take off it has to be supported by Nvidia, at feature parity with C++. Otherwise it'll remain a niche tool, unfortunately. I do hope Nvidia either supports this or rolls out their own alternative. I've been looking to get off the C++ train for years, but stuff like this keeps me there.

Looks like a dead end. Why CUDA? There should be some way to use Rust for GPU programming in general fashion, without being tied to Nvidia.