I’m a rust-GPU maintainer and can say that shared types on host and GPU are definitely intended. We’ve mostly been focused on graphics, but are shifting efforts to more general compute. There’s a lot of work though, and we all have day jobs - we’re looking for help. If you’re interested in helping you should say so at our GitHub.
What is the intended distinguisher between this and WGPU for graphics? I didn't realize that was a goal; have seen it mostly discussed in context of CUDA. There doesn't have to be, but I'm curious, as the CUDA/GPGPU side of the ecosystem is less developed, while catching up to WGPU may be a tall order. From a skim of its main page, it seems like it may also focus on writing shaders in rust.
Tangent; What is the intended distinguishes between Rust-CUDA, and Cudarc? Rust shaders with shared data structures I'm guessing is the big one. That would be great! There of course doesn't have to be. More tools to choose from, and that encourages progress from each other.
