Software combining to mitigate multithreaded MPI contention
Abdelhalim Amer, Charles J Archer, Michael A Blocksome, Chongxiao Cao, Michael Chuvelev, Hajime Fujita, María Jesús Garzarán, Yanfei Guo, Jeff R. Hammond, Shintaro Iwasaki, Ken Raffenetti, Mikhail Shiryaev, Min Si, Kenjiro Taura, Sagar Thapaliya, Pavan Balaji · 2019
Efforts to mitigate lock contention from concurrent threaded accesses to MPI have reduced contention through fine-grained locking, avoided locking altogether by offloading communication to dedicated threads, or alleviated negative side effects from contention by using better lock management protocols. The blocking nature of lock-based methods, however, wastes the asynchrony benefits of nonblocking MPI operations, and the offloading model sacrifices CPU resources and incurs unnecessary software offloading overheads under low contention.