Automatic Sequential to Parallel Code Conversion

Abstract

The way software programs are being written has been redefined since the introduction of multicore processors. Software developers have started writing parallel programs that are robust and scalable. This would ensure use of processor power being made available in the form of multiple cores. Though this trend is increasing, there are legacy applications that have been developed over the past few decades. Most of these applications are inherently sequential making no use of multithreading or parallel programming. If such applications are ported to execute on the multicore hardware as they are then optimal usage of all cores is not guaranteed. Such applications would ideally utilize only one core and the other cores would remain idle, unless the operating system supports some parallelism while scheduling. Hence there is a need to convert such legacy sequential codes to their parallel versions so that multicore hardware is exploited to the fullest. In this paper we present a tool that we have developed to automatically convert a sequential C code to parallel code. This Sequential to Parallel (S2P) tool is still in the development phase. We also discuss other parallelization tools available today, compare such tools with S2P tool and present our performance analysis results on different kind of multicore hardware.