MIT Lincoln Laboratory's Embedded Digital Systems Group (102) is involved in the design and development of advanced signal processing software technology for rapid prototyping of real-time embedded systems. One of the key challenges is the software-to-hardware mapping and optimization of the application software. The problem is exacerbated when the application requires multiple processors in order to meet throughput requirements. To address these challenges, Lincoln Laboratory has been designing and developing over the past year and a half an intelligent, automatic parallelization library pMapper for the MATLAB programming environment. pMapper allows MATLAB users to quickly and easily parallelize their code in order to achieve significant performance increases. Currently pMapper is still in the research phase, and consequently several ideas and directions are being pursued including the use of different mapping strategies, the implementation of a robust simulation system, and the eventual migration of the library to the laboratory community. The focus of this Major Qualifying Project (MQP) is on the design, implementation and testing of a new intelligent mapping strategy ulitizing artificial neural networks. More specifically, the tradeoffs between this new strategy and the baseline heuristic programming mapping strategy are compared and analyzed in detail.

• This report represents the work of one or more WPI undergraduate students submitted to the faculty as evidence of completion of a degree requirement. WPI routinely publishes these reports on its website without editorial or peer review.

Creator

• Jennings, Daniel S.

Publisher

• Worcester Polytechnic Institute

Identifier

• 05B010M

Advisor

• Ciaraldi, Michael J.

Year

• 2005

Center

• MIT Lincoln Laboratory-Lexington, Massachusetts Project Center - MQP

Sponsor

• MIT Lincoln Laboratory

Date created

• 2005-01-01

Resource type

• Major Qualifying Project

Major

• Electrical & Computer Engineering

Rights statement

• In Copyright