GPU acceleration for the web browser based evolutionary computing system

This paper presents a novel approach for the acceleration of distributed computing system entirely based on web browsers. We propose two strategies of embedding GPU kernels into the JavaScript code that is run by clients' machines (computing agents) participated in the computing grid, and analyze the speed increase resulting from the application of these methods. The computational experiments are performed on the basis of the two standard optimization problems: a travelling salesman problem and a flowshop scheduling problem. According to the obtained results the calculation of a fitness function accelerated by GPU may bring up to 50% reduction in execution time, while a local search process accelerated by GPU may be reduced tenfold.