The interesting challenge with async/await on GPU is that it inverts the usual concurrency mental model. CPU async is about waiting efficiently while I/O completes. GPU async is about managing work distribution across warps that are physically executing in parallel. The futures abstraction maps onto that, but the semantics are different enough that you have to be careful not to carry over intuitions from tokio/async-std.
The comparison to NVIDIA's stdexec is worth looking at. stdexec uses a sender/receiver model which is more explicit about the execution context. Rust's Future trait abstracts over that, which is ergonomic but means you're relying on the executor to do the right thing with GPU-specific scheduling constraints.
Practically, the biggest win here is probably for the cases shayonj mentioned: mixed compute/memory pipelines where you want one warp loading while another computes. That's exactly where the warp specialization boilerplate becomes painful. If async/await can express that cleanly without runtime overhead, that is a real improvement.
show comments
starkiller
Love this.
You mention futures are cooperative and GPUs lack interrupts, but GPU warps already have a hardware scheduler that preempts at the instruction level. ARe you intentionally working above that layer, or do you see a path to a fture executor that hooks into warp scheduling more directly to get preemptive-like behavior?
xiphias2
Really cool experiment (the whole company).
Training pipelines are full of data preparation that are first written on CPU then moving to GPU and always thinking of what to keep on CPU and what to put on GPU, when is it worth to create a tensor, or should it be tiling instead. I guess your company is betting on solving problems like this (and async-await is needed for serving inference requests directly on the GPU for example).
My question is a little bit different: how do you want to handle the SIMD question: should a rust function be running on the warp as a machine with 32 long arrays as data types, or always ,,hope'' for autovectorization to work (especially with Rust's iter library helpers).
show comments
zozbot234
I'm not quite seeing the real benefit of this. Is the idea that warps will now be able to do work-stealing and continuation-stealing when running heterogenous parallel workloads? But that requires keeping the async function's state in GPU-wide shared memory, which is generally a scarce resource.
show comments
GZGavinZhao
One concern I have is that this async/await approach is not "AOT"-enough like the Triton approach, in the sense that you know how to most efficiently schedule the computations on which warps since you know exactly what operations you'll be performing at compile time.
Here with the async/await approach, it seems like there needs to be manual book-keeping at runtime to know what has finished, what has not, and _then_ consider which warp should we put this new computation in. Do you anticipate that there will be measurable performance difference?
show comments
textlapse
What's the performance like?
What would the benefits be of converting a streaming multiprocessor programming model to this?
show comments
the__alchemist
Et tu, GPU?
I am, bluntly, sick of Async taking over rust ecosystems. Embedded and web/HTTP have already fallen. I'm optimistic this won't take hold in GPU; well see. Async splits the ecosystem. I see it as the biggest threat to Rust staying a useful tool.
I use rust on the GPU for the following: 3d graphics via WGPU, cuFFT via FFI, custom kernels via Cudarc, and ML via Burn and Candle. Thankfully these are all Async-free.
show comments
ismailmaj
Warp specialization is an abomination that should be killed and I'm glad this could be an alternative.
I hope they can minimize the bookkeeping costs because I don't see it gain traction in AI if it hurts big kernels performance.
shayonj
Very cool to see this and something I have been curious about myself and exploring the space as well. I'd be curious what are some parallels and differentiations between this and NVIDIA's stdexec (outside of it being in Rust and using Future, which is also cooL)
Arch485
Very cool!
Is the goal with this project (generally, not specifically async) to have an equivalent to e.g. CUDA, but in Rust? Or is there another intended use-case that I'm missing?
show comments
firefly2000
Is this Nvidia-only or does it work on other architectures?
show comments
bionhoward
genius, great idea and follow through, please keep it up, this could improve the ML industry tremendously, maybe some einops inspired interface for this would be good?
The interesting challenge with async/await on GPU is that it inverts the usual concurrency mental model. CPU async is about waiting efficiently while I/O completes. GPU async is about managing work distribution across warps that are physically executing in parallel. The futures abstraction maps onto that, but the semantics are different enough that you have to be careful not to carry over intuitions from tokio/async-std.
The comparison to NVIDIA's stdexec is worth looking at. stdexec uses a sender/receiver model which is more explicit about the execution context. Rust's Future trait abstracts over that, which is ergonomic but means you're relying on the executor to do the right thing with GPU-specific scheduling constraints.
Practically, the biggest win here is probably for the cases shayonj mentioned: mixed compute/memory pipelines where you want one warp loading while another computes. That's exactly where the warp specialization boilerplate becomes painful. If async/await can express that cleanly without runtime overhead, that is a real improvement.
Love this.
You mention futures are cooperative and GPUs lack interrupts, but GPU warps already have a hardware scheduler that preempts at the instruction level. ARe you intentionally working above that layer, or do you see a path to a fture executor that hooks into warp scheduling more directly to get preemptive-like behavior?
Really cool experiment (the whole company).
Training pipelines are full of data preparation that are first written on CPU then moving to GPU and always thinking of what to keep on CPU and what to put on GPU, when is it worth to create a tensor, or should it be tiling instead. I guess your company is betting on solving problems like this (and async-await is needed for serving inference requests directly on the GPU for example).
My question is a little bit different: how do you want to handle the SIMD question: should a rust function be running on the warp as a machine with 32 long arrays as data types, or always ,,hope'' for autovectorization to work (especially with Rust's iter library helpers).
I'm not quite seeing the real benefit of this. Is the idea that warps will now be able to do work-stealing and continuation-stealing when running heterogenous parallel workloads? But that requires keeping the async function's state in GPU-wide shared memory, which is generally a scarce resource.
One concern I have is that this async/await approach is not "AOT"-enough like the Triton approach, in the sense that you know how to most efficiently schedule the computations on which warps since you know exactly what operations you'll be performing at compile time.
Here with the async/await approach, it seems like there needs to be manual book-keeping at runtime to know what has finished, what has not, and _then_ consider which warp should we put this new computation in. Do you anticipate that there will be measurable performance difference?
What's the performance like? What would the benefits be of converting a streaming multiprocessor programming model to this?
Et tu, GPU?
I am, bluntly, sick of Async taking over rust ecosystems. Embedded and web/HTTP have already fallen. I'm optimistic this won't take hold in GPU; well see. Async splits the ecosystem. I see it as the biggest threat to Rust staying a useful tool.
I use rust on the GPU for the following: 3d graphics via WGPU, cuFFT via FFI, custom kernels via Cudarc, and ML via Burn and Candle. Thankfully these are all Async-free.
Warp specialization is an abomination that should be killed and I'm glad this could be an alternative.
I hope they can minimize the bookkeeping costs because I don't see it gain traction in AI if it hurts big kernels performance.
Very cool to see this and something I have been curious about myself and exploring the space as well. I'd be curious what are some parallels and differentiations between this and NVIDIA's stdexec (outside of it being in Rust and using Future, which is also cooL)
Very cool!
Is the goal with this project (generally, not specifically async) to have an equivalent to e.g. CUDA, but in Rust? Or is there another intended use-case that I'm missing?
Is this Nvidia-only or does it work on other architectures?
genius, great idea and follow through, please keep it up, this could improve the ML industry tremendously, maybe some einops inspired interface for this would be good?