MATCH Virtual Machine: An Adaptive RunTime System to Execute MATLAB

MATLAB is one of the most popular languages for desk-top numerical computations as well as for signal and im-age processing applic ations. Applying parallel processing techniques to improve performance of MATLAB codes has been the goal of many recent works. Most current frame-works require the user to specify parallelism and/or in-formation regarding type/shap eof the variables, thereby sacri cing the user friendliness which is one of the most popular MATLAB featur es. Other systems work on a re-stricted subset of MATLAB, thereby limiting the class of applic ationsMATLAB can supp ort. We present a run-time system capable of executing MATLAB code in par-allel without any user intervention. The runtime system performs automatic parallelization and typ e/shap einfer-ence of the code at runtime. A unique feature of the runtime system is its capability to automatically adapt to changes in the underlying architecture, making it par-ticularly useful for systems where predicting p erformance statically is dicult. We present exp erimental results ob-taine d for the runtime system running on SGI Origin2000 shared memory multiprocessor