MATCH: A MATLAB Compilation Environment for Distributed Heterogeneous Adaptive Computing Systems
Project Funding:
DARPA Adaptive Computing Systems Program
DARPA Program Manager Dr. Jose Munoz, DARPA ITO
AFRL Program Manager: Dr. Ralph Kohler, Air Force Research Lab
Project Leaders:
Prof. Prithviraj Banerjee,
Prof. Alok Choudhary,
Prof. Scott Hauck.
Prof. Nagaraj Shenoy.
Technology Transfer to a Startup Company
An exclusive license of the Match Compiler has been transferred to a startup
company called AccelChip, Inc. (formerly called Mach Design Systems), located in Schaumburg, Illinois.
The company has been founded by some professors and students
working on the MATCH compiler at Northwestern University.
For more information about the company check here
AccelChip
News Release on MATCH Project:
Compiler shortens DSP design time, startup claims
SCHAUMBURG, Ill.: Mach Design Systems, a startup, claims to
have "cracked the code" for synthesizing DSPs using C code from the
Matlab simulator of The Mathworks Inc. Based on a compiler developed
at Northwestern University, Mach Design said it can convert Matlab
code into register-transfer level (RTL) VHDL that can be used to
synthesize ASICs or FPGAs. The compiler can also generate C code that
can execute on general-purpose embedded processors, said Prith
Banerjee, president and chief executive officer of the new company.
See more details at
News Article in EE Times Aug.
2000 ,
Project Overview
Adaptive computing systems are composed of reconfigurable hardware
and are capable of adaptation to various computation
requirements of DOD aplications.
It is widely known that purely FPGA-based systems are unsuitable for complete
algorithm implementation for complex applications.
A practical approach to build adaptive computing system
is to build it out of commercial-off-the-shelf (COTS) FPGAs, DSP processors
and embedded processors all integrated within one system.
A key problem that needs to be addressed is how to map a given
computation onto such a heterogeneous architecture
without expecting the application programmer
to get into the low level details of the architecture
and program the system at the level of assembly code
for the embedded and DSP processors, or at the level of
VHDL programming for the FPGA logic.
The main objective of the MATCH (MATlab Compiler for Heterogeneous
adaptive computing systems) project is to make it easier for the
users to develop efficient codes for adaptive
computing systems.
We will develop a compiler that will take
in the user's applications written in a high-level language (MATLAB) and generate efficient
low level code that will run on COTS FPGAs, COTS embedded processors, COTS
DSP processors.
The specific goals of the MATCH project are to:
-
Enable the users to reduce code development times for adaptive applications from weeks
(using manual approaches) to hours (using compiler tools).
This will allow the users to focus on
the high-level algorithmic developments in MATLAB and leave the details
of the implementations to the compiler.
-
Produce efficient codes that are within a factor of 2-4 of the
best manual approach with respect to
optimizing resources (number of processors, FPGAs, code sizes) under performance constraints (latency, throughput),
or optimizing performance under resource constraints.
The project was funded by DARPA from March 1998 to August 2001. The project is now continuing with support from NASA.
Mathtools.net is the technical
computing portal for all scientific and engineering needs. The portal
is provided as a free service to the scientific community and contains
useful links for technical computing programmers.
Updated by Prith Banerjee Feb. 25, 2002.
Send any questions to
Professor Banerjee