Automatic program parallelization for multicore processors

Abstract

With the advent of multi-core processors the problem of designing application that efficiently can utilize it performance become more and more important. Moreover developing programs for these processors requires from the programmers some additional, specific knowledge about the processor architecture. In multi-core systems efficient program execution is the main issue. It can even happen that switching from sequential to parallel computation can lead to decreasing of performance. The paper deals with the short description of SliCer, the hardware independent tool that parallelizes serial programs in automatic way depending on the number of available processing units by creating the proper number of threads that can be later execute in parallel. © 2010 Springer-Verlag Berlin Heidelberg.