Hybrid parallel task placement in X10

Jeeva Paudel, Olivier Tardieu, José Nelson Amaral
2013 Proceedings of the third ACM SIGPLAN X10 Workshop on - X10 '13  
This paper presents a hybrid parallel task-placement strategy that combines work stealing and work dealing to improve workload distribution across nodes in distributed shared-memory machines. Existing work-dealing-based load balancers suffer from large performance penalties resulting from excessive task migration and from excessive communication among the nodes to determine the target node for a migrated task. This work employs a simple heuristic to determine the load status of a node and also
more » ... o detect a good target for migration of tasks. Experimental evaluations on applications chosen from the Cowichan and Lonestar suites demonstrate a speedup, with the proposed approach, in the range of 2% to 16% on a cluster of 128 cores over the state-of-the-art work-stealing scheduler.
doi:10.1145/2481268.2481277 dblp:conf/pldi/PaudelTA13 fatcat:h5lruyqmfvdkddzgtfgv2we33u