It’s great that there is a page about current limitations [1], but I am afraid that what most people describe as “CUDA” is a small subset of the real CUDA functionality. Would be great to have a comparison table for advanced features like warp shuffles, atomics, DPX, TMA, MMA, etc. Ideally a table, mapping every PTX instruction to a direct RDNA counterpart or a list of instructions used to emulate it.
[1]: https://docs.scale-lang.com/manual/differences/