method · first seen 13 Nov 2026
copy engine data movement
Moving GPU data asynchronously with dedicated DMA copy engines instead of compute kernels, freeing SMs for overlapped computation.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Zero-SM NCCL: A Systematic Guide to Designing and Deploying SM-Free GPU CommunicationPapers · 13 Nov 2026 · Zhenhao He, Siyuan Shen, Ke Wen, Prajwal Singhania, Shiqing Fan, Youngeun Kwon,…
Latest sources
- Zero-SM NCCL: A Systematic Guide to Designing and Deploying SM-Free GPU CommunicationPapers · 13 Nov 2026 · Zhenhao He, Siyuan Shen, Ke Wen, Prajwal Singhania, Shiqing Fan, Youngeun Kwon,…