Live data from Hacker News

Decoupling Compute and Memory for Async GPUs

news.ycombinator.com

1–5 of 5 posts

Decoupling Compute and Memory for Async GPUs

#1
Cool open-source project that introduces a new programming model for decoupling compute and memory for NVIDIA GPUs that supports asynchronous memory operations (e.g., Hopper). 12% perf improvement over SOTA and 67% less kernel code.

Paper: "VDCores: Resource Decoupled Programming and Execution for Asynchronous GPU" arXiv:2605.03190