Signal Processing Algorithms for Heterogeneous Architectures
Abstract
As stated in our proposal, our research goal was to derive architectures which preserve the VLSI implementation advantages of regular meshes while still obtaining sufficient increases in performance and/or decreases in size and cost. Our problem domain was signal and image processing applications which require either significant computational resources or which must be solved in real time. Our research has been successful in that we have derived a number of such architectures for important problem areas in image and signal processing. These architectures are outlined in more detail in the final report. We even successfully prototyped one of our architectures, the Arithmetic Cube, which involved both VLSI chip and PCB board design and fabrication. The Arithmetic Cube, directly solves both the discrete Fourier transform and cyclic convolution. In parallel with our architecture work, we also investigated algorithms which map efficiently onto the various derived architectures and which take advantage of their particular architectural aspects.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 1993
- Accession Number
- ADA271248
Entities
People
- Mary J. Irwin
- Robert M. Owens
Organizations
- Pennsylvania State University