The Community for Technology Leaders
Cluster Computing and the Grid, IEEE International Symposium on (2012)
Ottawa, Canada
May 13, 2012 to May 16, 2012
ISBN: 978-0-7695-4691-9
pp: 384-392
Pipelining is necessary for efficient do-across parallelism but the use is difficult to automate because it requires send-receive analysis and loop blocking in both sender and receiver code. The blocking factor is statically chosen. This paper presents a new interface called delta send-recv. Through compiler and run-time support, it enables dynamic pipelining. In program code, the interface is used to mark the related computation and communication. There is no need to restructure the computation code or compose multiple messages. At run time, the message size is dynamically determined, and multiple pipelines are chained among all tasks that participate in the delta communication. The new system is tested on kernel and reduced NAS benchmarks to show that it simplifies message-passing programming and improves program performance.
MPI, communication-computation overlapping, dynamic pipelining
Yaoqing Gao, Chen Ding, Roch Archambault, Bin Bao, "Delta Send-Recv for Dynamic Pipelining in MPI Programs", Cluster Computing and the Grid, IEEE International Symposium on, vol. 00, no. , pp. 384-392, 2012, doi:10.1109/CCGrid.2012.113
92 ms
(Ver 3.3 (11022016))