IS&T RCS Tutorial - MATLAB for High-Performance Computing

MATLAB programs can be exceptionally fast if they are well-designed, and painfully slow if not. Fortunately, it does not take a professional programmer to write an efficient MATLAB program or to take advantage of multi-core processors, GPUs, and computer clusters. This tutorial will introduce (1) using MATLAB on the BU SCC, (2) optimizing MATLAB serial codes, and (3) using the MATLAB Parallel Toolbox. The first part will cover how to use the MATLAB platform remotely and how submitting batch jobs for MATLAB works. In the second part, some useful skills for removing unnecessary computation and optimizing memory usage will be introduced. The third part will introduce how to parallelize MATLAB codes using parallel processing tools, such as parfor, spmd, and distributed array types. The skills you learn should enable you to solve bigger problems faster using MATLAB. Basic knowledge of MATLAB is assumed.