CSU HAYWARD
DEPARTMENT OF MATHEMATICS AND
COMPUTER SCIENCE
COLLOQUIUM
Friday, November 9, 2001 noon-1pm ScN321
Speaker:
Phillip Nico, Cal Poly, San Luis Obisbo
Automatic Scheduling for Real-Time Signal Processing
In this talk I will present SERT, a language and compiler system for automatic parallelization, partitioning, and scheduling of real-time signal processing applications running on general-purpose hardware.
Traditionally, most real-time signal processing applications have required special-purpose hardware and software to deliver the performance demanded of them. In spite of the great cost of developing such systems, they soon become obsolete and have to be replaced. Real-time system development time and expense can be greatly reduced by using general-purpose processors in conjunction with high-level languages for system implementation. Advantages of this approach over current explicitly parallel or manual systems include increased ease of development and potentially greater accuracy with respect to timing analysis. In addition, our approach allows for rapid prototyping of new systems and simplifies porting to new computing platforms as they are developed. The ability to test different system configurations quickly will allow system implementors to determine minimal, (i.e., least expensive) hardware configurations easily, yielding potentially substantial economic benefits.
Please join us beforehand for pizza!!