Improving the Performance of Collective Operations in MPICH
By R. Thakur, William D. Gropp. PVM/MPI, 2003
300 citations recorded by the source corpus.
Subfields: Computer Networks and Communications, Hardware and Architecture
View Paper: https://www.semanticscholar.org/paper/1a7a3c3945947358e99257b3461d3eff8a06946d
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Computer Networks and Communications
- Jelena Pjesivac-Grbovic
- Edgar Gabriel
- Jack J. Dongarra
- George Bosilca
- Rolf Rabenseifner
- Graham Edward Fagg
- Thara Angskun
- Torsten Hoefler
In Hardware and Architecture
- Jelena Pjesivac-Grbovic
- Edgar Gabriel
- Jack J. Dongarra
- George Bosilca
- Rolf Rabenseifner
- Graham Edward Fagg
- Thara Angskun
- Torsten Hoefler
Citations in the corpus
- SparCML: High-Performance Sparse Communication for Machine Learning (2018, unknown): “In practice, allreduce implementations switch between different implementations depending on the message size and the number of processes [48].”
- Optimization of Collective Reduction Operations (2004, methodology): “It uses the pairwise exchange algorithm for reduce scatter and ring algorithm for allgather (for allreduce), as described in [12], and for reduce, all processes send their result segment directly to root.”
- Performance analysis of MPI collective operations (2005, background): “A profiling study [1] showed that some applications spend more than eighty percent of a transfer time in collective operations.”