Ithaca, NY — The Cornell Theory Center (CTC) announces the release of the Cornell Multitask Toolbox for MATLAB (CMTM). The MATLAB software environment is used in colleges and corporations around the world as a tool for data analysis, simulation, and visualization. Typically, users select from an extensive suite of tools and customize the software environment for their needs, whether financial risk analysis or a laboratory engineering course in wave science. CMTM provides a new, user-friendly set of development programming tools that extends the power of MATLAB to parallel computing.
According to lead developer John Zollweg, CMTM allows the user to harness multiple MATLAB tasks for working on large problems using parallel programming methods. For example, researchers seeking to improve the performance of electric power grids look for optimal routing solutions. This type of optimization problem requires the computer to generate tens or hundreds of thousands of solutions and then analyze them. This can take a long time on any computing system, but multitasking can help. Typically the wall-clock time to solve a problem using CMTM is only a fraction of what it would be using a single task. Zollweg developed the Toolbox with CTC research associate Dr. Arun Verma.
CMTM combines the programming ease of MATLAB with the message-passing ability of MPI. MPI is the informal standard for message passing used in parallel computing. The self-describing matrix feature of MATLAB allows one to make MPI calls, with fewer arguments than needed when using traditional programming languages (C and FORTRAN), through the CMTM, thus making parallel applications generated using CMTM more efficient.
Additional features of CMTM support a variety of programming styles. For example, code can be written that uses either static or dynamic load-balancing, users can monitor output from multiple tasks in one console, and the toolbox supports remote master task control for cluster computing.
Robert J. Thomas, professor of electrical engineering at Cornell, studies problems within large-scale electrical grid systems. Members of his research team, among the early users for CMTM, have already benefited from the toolbox. “As part of his research on the problem of unit-commitment subject to AC power-flow constraints in electric power systems, Carlos Murillio-Sanchez discovered a natural parallelization of a new optimization algorithm,” says Thomas.
CMTM is available for Windows NT or 2000 systems with MATLAB 5.3 and MPI/Pro 1.5. A time-limited beta version of the toolbox can be downloaded free of charge at http://www.tc.cornell.edu/UserDoc/Cluster/Software/CMTM/ . CTC is a high-performance computing and interdisciplinary research center located at Cornell University. CTC receives funding from Cornell University, New York State, a number of federal agencies, and Corporate Program members.