SIAM PP08 Conference

MS5: Parallel Computing with MATLAB

MATLAB has emerged as one of the languages most commonly used by scientists for technical computing, with ~1,000,000 users worldwide. The primary benefits of MATLAB are reduced code development time via high levels of abstractions, interactive programming, and powerful mathematical graphics. Computationally intensive MATLAB tasks can significantly benefit from the increased performance offered by parallel computing. There exists a number of parallel programming solutions for MATLAB. This workshop will provide a unique opportunity to interact between researchers from the growing field of parallel MATLAB computing and the technical leaders in this area.

Organizer: Piotr Luszczek
The MathWorks, Inc
Jeremy Kepner
Massachusetts Institute of Technology

10:00-10:25 Parallel Computing Toolbox (PCT) and Parallel Programming in MATLAB
Piotr Luszczek, The MathWorks, Inc
Abstract:
Parallel Computing Toolbox (PCT) addresses computationally and data-intensive problems using MATLAB and Simulink in a multiprocessor computing environment. The toolbox allows both several independent tasks or a single parallel computation by harnessing computing clusters and a variety of batch queuing software implementation. The toolbox provides high-level constructs, such as parallel loops and algorithms, and MPI-based functions. Also, low-level constructs for resource management are included. The Parallel Command Window provides interactive environment for developing parallel applications.
[PDF]
10:30-10:55 Parallel Programming in MATLAB: Best Practices
Jeremy Kepner, Massachusetts Institute of Technology
Abstract:
Matlab is one of the most commonly used languages for scientific computing with approximately one million users worldwide. The Lincoln pMatlab library (http://www.ll.mit.edu/pMatlab), The Mathworks DCT, and StarP from ISC have brought parallel computing to the this community using the distributed array programming paradigm. This talk provides an introduction to distributed array programming and will describe the best programming practices for using distributed arrays to produce well performing parallel Matlab programs.
[PPT]
11:00-11:25 Parallel MATLAB in Production Supercomputing with Applications in Signal and Image Processing
Ashok Krishnamurthy, David Hudak, John Nehrbass, Siddharth Samsi, and Vijay Gadepally, Ohio Supercomputer Center
Abstract:
Parallel MATLAB enables the large community of MATLAB users to harness the increased computing capacity and memory of distributed memory clusters. At the Ohio Supercomputer Center we provide our users with three varieties of Parallel MATLAB. In this talk, we will describe how we run these Parallel MATLAB environments within a traditional batch oriented queuing system. We will also describe our experiences in developing three signal and image processing applications within this environment.
[PPT]
11:30-11:55 Interactive Data Exploration with Star-P
Viral B. Shah, University of California, Santa Barbara
Abstract:
High performance applications increasingly combine numerical and combinatorial algorithms. Past research on high performance computation has focused mainly on numerical algorithms, and there is a rich variety of tools for high performance numerical computing. On the other hand, few tools exist for large scale combinatorial computing. We describe our efforts to build a common infrastucture for numerical and combinatorial computing by using parallel sparse matrices to implement parallel graph algorithms.
[PDF | PPT]