1 paper · 1 filter
Lannie Dalton Hough, Emir Gencer, Hoffmann Muki +1
Collective communication is essential to high performance computing and machine learning workloads, yet libraries such as NCCL do not exploit sparsity in message payloads. Sending…