Efficient threads mapping on multicore architecture

Abstract

Considering today's hardware performance, in order to obtain best results, a proper programming strategy for optimum mapping of all processes to existing resources is necessary. The presence of multiple cores in a single chip requires applications with a higher level of parallelism. The use of suited mapping algorithms can lead to a great performance improvement considering computing time at a smaller energy consumption. We've realized a comparison between the parallel computing with an efficient mapping algorithm of threads to specific cores and parallel computing with threads mapping maintained by Linux kernel process scheduler. We observed that a good strategy regarding thread mapping on different processing units, balancing all available cores and allocating a specific amount of work, can lead to improved computational time. In our simulation we used Kubuntu Linux operating system, a system with Intel Core 2 Duo processor and another system with an Intel Quad Core.