CPU utilization analysis of selected genetic algorithms in multi-core systems for a certain class of problems

Abstract

This work was carried out in order to examine and compare selected models of genetic algorithms (through the implementation), using the latest tools and libraries that allow for multithreaded programming in a.NET environment. Implemented algorithms were then tested for the use of available resources, such as CPU cycles/cores consumption and the time at which they are able to provide the quality results at acceptable pace. With a choice of multi-core processors—allowing for parallel calculations on their cores, as well as genetic algorithms, one should think about how to implement the chosen algorithm so as to avoid the deadlocks and bottlenecks to make optimal use of the computing power of cores. There are many approaches to deal with such issues—a lot of tools and software libraries facilitate the implementation of such algorithms. This paper tries to address two essential questions what algorithms fit the best into multicore architecture, and which one benefits the best from available logical/physical cores producing the best possible results.