TY  - RPRT
TI  - Optimized Broadcast for Deep Learning Workloads on Dense-GPU InfiniBand Clusters: MPI or NCCL?
AU  - Ammar Ahmad Awan
AU  - Ching-Hsiang Chu
AU  - Hari Subramoni
AU  - Dhabaleswar K. Panda
PY  - 2017
UR  - https://arxiv.org/abs/1707.09414
ID  - 1707.09414
ER  -