You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We are organizing a new working group focused on GPUs (for both AI/compute and graphics use-cases). I have been talking to folks in the community, getting feedback, and signing up participants. Here is the current thinking:
@jamesmunns mentioned that the Rust project is phasing out these "classic domain" WGs and we should decide the best route forward.
One option would be to scope down to just the Rust on GPUs compiler backends, which more closely matches "current day" WGs. The GPU working group folks had discussed this earlier (see the gist) and decided we would prefer to not limit it to just the compiler backends because we think design needs to span across host and GPU code (and sometimes CPU rust will be talking to GPU rust, sometimes not). We want to try a wholistic approach to GPUs vs the targeted approach everyone is currently doing separately.
Another option would be to add this as a "classic domain WG" and have it spin off "current day WGs" when we have concrete actionable chunks of work (like, say, memory spaces or vector / matrix traits). This too seems not ideal if the Rust project is trying to kill these "classic domain" WGs.
We don't care too much about being an "official" working group, except that 1) we would like to post on the official blog so non-GPU Rust people and non-Rust GPU people / companies know the initiative exists and can join 2) GPU and AI companies are looking for an obvious / "official" place to plug into Rust development and as they aren't Rust native, having a big ol' neon "talk to these folks" sign is extremely valuable.
Hello! One of the maintainers of rust-gpu here (and rust-cuda, see https://rust-gpu.github.io/blog/2025/01/27/rust-cuda-reboot/).
We are organizing a new working group focused on GPUs (for both AI/compute and graphics use-cases). I have been talking to folks in the community, getting feedback, and signing up participants. Here is the current thinking:
https://gist.github.com/LegNeato/a1fb3e3a9795af05f22920709d92e456
@jamesmunns mentioned that the Rust project is phasing out these "classic domain" WGs and we should decide the best route forward.
One option would be to scope down to just the Rust on GPUs compiler backends, which more closely matches "current day" WGs. The GPU working group folks had discussed this earlier (see the gist) and decided we would prefer to not limit it to just the compiler backends because we think design needs to span across host and GPU code (and sometimes CPU rust will be talking to GPU rust, sometimes not). We want to try a wholistic approach to GPUs vs the targeted approach everyone is currently doing separately.
Another option would be to add this as a "classic domain WG" and have it spin off "current day WGs" when we have concrete actionable chunks of work (like, say, memory spaces or vector / matrix traits). This too seems not ideal if the Rust project is trying to kill these "classic domain" WGs.
We don't care too much about being an "official" working group, except that 1) we would like to post on the official blog so non-GPU Rust people and non-Rust GPU people / companies know the initiative exists and can join 2) GPU and AI companies are looking for an obvious / "official" place to plug into Rust development and as they aren't Rust native, having a big ol' neon "talk to these folks" sign is extremely valuable.
More details in the original Zulip thread.
The text was updated successfully, but these errors were encountered: