Performance analysis of current parallel programming models for many-core systems
Yangjie Cao, Baodong Wu, Yongcai Tao, Lei Shi · 2013
Recent developments in microprocessor design show a clear rend towards multi-core and many-core architectures. Nowadays, processors consisting of dozens of general-purpose cores are already available in the market, and with the rapid development of semiconductor technology and increasing computing demand, it will be very common to have a processor consisting of hundreds or even thousands of cores in the near future. This kind of processors is commonly referred as many-core processors. In the many-core processor era, however, exploiting all the advantages offered by these processors will not be trivial. In this paper, we focus on studying and analyzing several typical parallel programming models for many-core systems, and summarize the common performance-optimized problems faced by current parallel programming models.