Application Analysis for Parallel Processing

Effective mapping of multimedia applications on massively parallel embedded systems is a challenging demand in the domain of compiler design. The software implementations of emerging multimedia applications are often huge and it is virtually impossible to analyze these applications for parallel processing without generic automated tools. This paper proposes a two step methodology for specifying multimedia applications (video applications in this paper) in form of parallel process networks. In the first step of the proposed methodology, an input program written in a high-level language is translated into a trace tree representation by dynamic analysis. Operations are performed on the trace tree representation of the application to get analysis results. In the second step, results obtained from application analysis are exploited to re-formulate the application in form of parallel process networks expressed in Avel framework. A well known multimedia application MPEG-2 decoder serves as a case study. Experimental results show the applicability of the proposed methodology.