Decoupling Compute and Memory for Async GPUs
4 comments
67% less kernel code is the more interesting number here — Hopper's async capabilities have been underutilized largely because the programming model is painful. Curious how it handles cases where compute and memory phases aren't cleanly separable.
This seems like a better version of CUDA, for Hopper GPUs?
Paper: "VDCores: Resource Decoupled Programming and Execution for Asynchronous GPU" arXiv:2605.03190