» MIT OpenCourseWare » Mathematics » Applied Parallel Computing (SMA 5505), Spring 2003

Projects

The project counts toward 50% of the course grade. Students are required to hand in 2 progress reports, and a final project report. Below are samples of final student reports.

PROJECT TITLES / AUTHORS	FINAL REPORTS
Comparison of Programming and Synchronization Techniques - Sean Lie	(PDF) Courtesy of Sean Lie.
FFTW and MATLAB®*P - Richard Hu	(PDF) Courtesy of Richard Hu. Used with permission.
Parallel Implementation of a Multi-Length Scale Finite Element Method - Trevor Tippetts	(PDF) Courtesy of Trevor Tippetts.
Java MPI in MATLAB®*P - Max Goldman and Da Guo	(PDF) Courtesy of Max Goldman and Darwin Guo.
A Parallel Hierarchical Solver for the Poisson Equation - R. Sudarshan and Seung Lee	(PDF) Courtesy of Raghunathan Sudarshan and Seung Lee.
Sparse Matrix Implementation on MATLAB®*P - Stu Blair	(PDF) Courtesy of Stu Blair.
Parallelizing Incremental Bayesian Segmentation (IBS) - Joseph Hastings, Siddhartha Sen

Some Ideas

A new (and better) visualization package for MATLAB®*P.
Grid enabled MPI applications - use MATLAB®*P as a test application.
Optimization package for MATLAB®*P - either implement a few general purpose functions or do one function complete with real applications.
Allow Octave to be used for mm mode - Octave has no engine mode, need some clever hack to get it work.
Allow a 'delayed evaluation' mode in MATLAB®*P. Say you do A = randn(100*p); B = chol(A); 2nd statement will return immediately to MATLAB®, even before completion of chol. The next command to server will block and wait for chol to finish. This allows MATLAB® commands to be completed while server is working. This requires modification to the server, multithreaded stuff, and could be hard.
Add sparse matrix support to MATLAB®*P - this is a huge project, but pieces could be carved out. Collaboration with John Gilbert (Sparse matrix guru of MATLAB® sparse fame) at UCSB.

Search

Projects