IS&T RCS Tutorial - Matlab Performance Optimization (Hands-on)

For many programs speed is the main research productivity factor, meaning the difference between hours versus months in getting results. For many applications it is possible to write MATLAB programs that are as fast or only a factor of 2-3 times slower than a compiled C++ or Fortran version, while requiring dramatically fewer lines of code and/or being faster to develop. For existing MATLAB programs, optimizing their performance can provide dramatic speedups. This tutorial will take a "case study" hands-on approach, examining several example programs and optimizing their performance. In the process this will demonstrate useful speedup techniques including vectorization, bsxfun, memory management, and "big O" algorithmic improvements. Basic knowledge of MATLAB is assumed.