Synthetic Circuit Generation and Circuit Characterization

One of the big issues in both FPGA architecture research and CAD research is obtaining enough realistic circuit examples to test architectures and algorithms. This project is concerned with both the determination of what those realistic characteristics are, and the construction of a software tool that will randomly generate circuits with those kind of characteristics specified. (The obvious ones, like size and fanout distribution, and the not-so-obvious ones like "shape" and edge-length distribution and sequential level. This is a brand-new area of research with many challenging problems. Our first work on the subject was published in [Hutton96] . More recently, we have extended this work to characterize and generate sequential circuits. [Hutton97] .

Circuit characterization and generation software is available.

We have extended this work to characterize and generate circuits based on clusters. Here the idea is to partition circuits into clusters (some number typically between 4 and 8, likely growing with circuit size) and to characterize the connectivity between those clusters. When generating synthetic circuits from that kind of characterization, we generate clusters independently and then connect them together. Paul Kundarewich's Master's thesis, "Synthetic Circuit Generation Using Clustering and Iteration" , describes this work. The software arising from this work can be found here.

Research on FPGAs at the University of Toronto.

Return to Jonathan Rose's Page .Computer Group.