Affinity-Aware Work-Stealing for Integrated CPU-GPU Processors

作者：Farooqui Naila^*; Barik Rajkishore; Lewis Brian T; Shpeisman Tatiana; Schwan Karsten

来源：ACM Sigplan Notices, 2016, 51(8): 363-364.

DOI：10.1145/2851141.2851194

Recent integrated CPU-GPU processors like Intel's Broadwell and AMD's Kaveri support hardware CPU-GPU shared virtual memory, atomic operations, and memory coherency. This enables finegrained CPU-GPU work-stealing, but architectural differences between the CPU and GPU hurt the performance of traditionallyimplemented work-stealing on such processors. These architectural differences include different clock frequencies, atomic operation costs, and cache and shared memory latencies. This paper describes a preliminary implementation of our work-stealing scheduler, Libra, which includes techniques to deal with these architectural differences in integrated CPU-GPU processors. Libra's affinity-aware techniques achieve significant performance gains over classicallyimplemented work-stealing. We show preliminary results using a diverse set of nine regular and irregular workloads running on an Intel Broadwell Core-M processor. Libra currently achieves up to a 2x performance improvement over classical work-stealing, with a 20% average improvement.

• 出版日期2016-8