Sampling-Based Cache-Efficient Parallel Sorting on Multi-Core Systems

Provided by: Binary Information Press
Topic: Hardware
Format: PDF
According to the capacity of L2 cache and L1 cache, and the number of processing cores on multi-core machine, this paper presents a novel data partitioning approach which distributes properly the data sequence to be sorted in main memory to the shared L2 cache and private L1 cache of each processing core, and designs a sampling-based cache-efficient thread-level parallel sorting algorithm. The execution performance of the algorithm is further improved by thread-binding and thread-level parallel searching method.

Find By Topic