Efficient and Scalable Thread-level Parallel Algorithms for Sorting Multisets on Multi-core Systems

Abstract

By distributing adaptively the data blocks to the processing cores to balance their computation loads and applying the strategy of “the extremum of the extremums” to select the data with the same keys, a cache-efficient and thread-level parallel algorithm for sorting Multisets on the multi-core computer is proposed. For the sorting Multisets problem, an aperiodic multi-round data distribution model is presented, which the first round scheduling assigns data blocks into the slave multi-core nodes according to the given distribution order and the other rounds scheduling will distribute data blocks into the slave multi-core nodes by first request first distribution strategy. The scheduling technique can ensure that each slave node can receive the next required data block before it finishes sorting the current data block in its own main memory. A hybrid thread-level and process-level parallel algorithm for sorting Multisets is presented on the heterogeneous cluster with multi-core nodes which have different amount of processing cores, different computation and communication abilities, and distinct size of main memory. The experimental results on the single multi-core computer and the heterogeneous cluster with multi-core computers show that the presented parallel sorting Multisets algorithms are efficient and they obtain good speedup and scalability.