CUDA-oxide: Nvidia's official Rust to CUDA compiler

#1

Post by adamnemecek » Mon, May 11, 2026, 3:55 PM UTC

CUDA-oxide: Nvidia's official Rust to CUDA compiler
nvlabs.github.io

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#2

Post by arpadav » Mon, May 11, 2026, 5:09 PM UTC

This is amazing.. ive been working with custom CUDA kernels and https://crates.io/crates/cudarc for a long time, and this honestly looks like it could be a near drop-in replacement.

im especially curious how build times would compare? Most Rust CUDA crates obv rely on calling CMake or nvcc, which can make compilation painfully slow. coincidentally, just last week i was profiling build times and found that tools like sccache can dramatically reduce rebuild times by caching artifacts - but you still end up paying for expensive custom nvcc invocations (e.g. candle by hugging face calls custom nvcc command in their kernel compilation): https://arpadvoros.com/posts/2026/05/05/speeding-up-rust-whi...

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#3

Post by rvz » Mon, May 11, 2026, 5:18 PM UTC

This is a bit good for Rust if you want to use the language with CUDA. The problem is, it still doesn't really move the needle if you really don't like running closed source drivers and runtime binaries and care about open source.

Continuing from this discussion [0], this only makes it a Rust or a CUDA problem rather than a Python, CUDA and a PyTorch one if there bug in one of them.

Yet at the end of the day, it still uses Nvidia's closed source CUDA compiler 'nvcc' which they will never open source. A least Mojo promises to open source their own compiler which compiles to different accelerators with multiple backend support.

Unlike this...but uses Rust.

[0] https://news.ycombinator.com/item?id=48067228

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#4

Post by bigyabai » Mon, May 11, 2026, 5:25 PM UTC

rvz wrote: post #3

This is a bit good for Rust if you want to use the language with CUDA. The problem is, it still doesn't really move the needle if you really don't like running closed source drivers and runtime binaries and care about open source. Continuing from this discussion [0], this only makes it a Rust or a CUDA problem rather than a Python, CUDA and a PyTorch one if there bug in one of them. Yet at the end of the day, it stil…

> it still doesn't really move the needle if you really don't like running closed source drivers and runtime binaries

Those people probably did not buy an Nvidia GPU for themselves. It should be common knowledge that the "Open" Nvidia drivers still run gigantic firmware blobs to dispatch complex workloads. And Nouveau is close to useless for GPGPU compute.

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#5

Post by rowanG077 » Mon, May 11, 2026, 5:29 PM UTC

Personally I really don't want new GPU languages that do not have AD as a first class citizen. I mean rust is an improvement over C++ CUDA but still.

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#6

Post by pjmlp » Mon, May 11, 2026, 5:30 PM UTC

rvz wrote: post #3

This is a bit good for Rust if you want to use the language with CUDA. The problem is, it still doesn't really move the needle if you really don't like running closed source drivers and runtime binaries and care about open source. Continuing from this discussion [0], this only makes it a Rust or a CUDA problem rather than a Python, CUDA and a PyTorch one if there bug in one of them. Yet at the end of the day, it stil…

Mojo remains to be seen if it isn't another Swift for Tensorflow, apparently 1.0 won't even support Windows properly.

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#7

Post by TallGuyShort » Mon, May 11, 2026, 5:31 PM UTC

rowanG077 wrote: post #5

Personally I really don't want new GPU languages that do not have AD as a first class citizen. I mean rust is an improvement over C++ CUDA but still.

Sorry, what is AD in this context?

edit: oh, automatic differentiation?

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#8

Post by huflungdung » Mon, May 11, 2026, 5:33 PM UTC

TallGuyShort wrote: post #7

rowanG077 wrote: post #5

Personally I really don't want new GPU languages that do not have AD as a first class citizen. I mean rust is an improvement over C++ CUDA but still.

Sorry, what is AD in this context? edit: oh, automatic differentiation?

Active Directory

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#9

Post by semiinfinitely » Mon, May 11, 2026, 5:33 PM UTC

pjmlp wrote: post #6

rvz wrote: post #3

This is a bit good for Rust if you want to use the language with CUDA. The problem is, it still doesn't really move the needle if you really don't like running closed source drivers and runtime binaries and care about open source. Continuing from this discussion [0], this only makes it a Rust or a CUDA problem rather than a Python, CUDA and a PyTorch one if there bug in one of them. Yet at the end of the day, it stil…

Mojo remains to be seen if it isn't another Swift for Tensorflow, apparently 1.0 won't even support Windows properly.

[flagged]

Re: CUDA-oxide: Nvidia's official Rust to CUDA compiler

#10

Post by cyber_kinetist » Mon, May 11, 2026, 5:33 PM UTC

I'm quite interested in how they dealt with Rust's memory model, which might not neatly map to CUDA's semantics. Curious what the differences are compared to CUDA C++, and if the Rust's type system can actually bring more safety to CUDA (I do think writing GPU kernels is inherently unsafe, it's just too hard to create a safe language because of how the hardware works, and because of the fact that you're hyper-optimizing all the time)