Yiltan / MPI-Partitioned-Microbenchmarks

MPI Partitioned Microbenchmarks
4 stars 4 forks source link

Efficient Multi-Path NVLink/PCIe-Aware UCX based Collective Communication for Deep Learning #3

Open zfy3000163 opened 3 months ago

zfy3000163 commented 3 months ago

Is there any corresponding code implementation for this paper? Thanks https://ieeexplore.ieee.org/document/9547041/figures#figures

Yiltan commented 3 months ago

for which part?

zfy3000163 commented 3 months ago

Multi-Path Copy Algorithm with UCX. Or any other related implementation?

zfy3000163 commented 3 months ago

about : The Proposed Hierarchical MPI_Allreduce with Multi-Path Copy. Thanks