All-reduce
All-Reduce is a collective communication operation in parallel computing that belongs to the class of reduction operations, distinguished by its requirement that the result of reducing data across all processing nodes is distributed back to all nodes, thereby ensuring uniform availability of the global reduced value.
20
Publications
1.7K
Citations
104
Authors
30
Institutions
Publications per year
2017–2026
14
Parents
20
A generic communication scheduler for distributed DNN training acceleration
Yanghua Peng, Yibo Zhu, Yangrui Chen et al. · 2019 · 319 citations
Alexander Rasmussen, Vinh The Lam, Michael Conley et al. · 2012 · 71 citations
1–5 of 20
104
2
h-index
2
h-index
2
h-index
1
h-index
1
h-index
1–5 of 104
30

University of Hong Kong
Pok Fu Lam, Hong Kong
2 h-index
7 publications
1.4K citations

City University of Hong Kong
Hong Kong, Hong Kong
1 h-index
2 publications
74 citations

Sun Yat-sen University
Guangzhou, China
1 h-index
1 publications
37 citations

Peking University
Beijing, China
1 h-index
1 publications
37 citations
IBM Research - Austin
Austin, United States
1 h-index
2 publications
42 citations
University of California, Santa Barbara
Santa Barbara, United States
1 h-index
1 publications
40 citations
The University of Texas at Austin
Austin, United States
1 h-index
1 publications
24 citations

Knoxville College
Knoxville, United States
1 h-index
1 publications
28 citations
1–8 of 30
Repository · 3 publications · 395 citations
3
h-index
1 publications · 34 citations
1
h-index
1 publications · 12 citations
1
h-index
1 publications · 24 citations
1
h-index
TON · 1 publications · 29 citations
1
h-index
1–5