SGER: Adaptive Multi-Objective Algorithms for High-Performance Designs

University Of Minnesota-Twin Cities, Minneapolis MN

Investigators

Abstract

The overall goal of this research is to introduce a new class of algorithms for solving complex optimization problems on graphs. Intellectual Merit This project contributes to applications by solving an important problem in high performance computing. The adaptive algorithm: -computes a trajectory for achieving a target (clock cycle period) -creates control points to correct a trajectory (guidance algorithm function) -changes trajectory relative to inherent deviations -timing delay accumulated along data paths when logic gates interconnect on the chip -accumulated timing compared with projections, and bounds on distances between not yet placed components are reevaluated -new bounds consistent with original clock cycle period used to find locations for a new group of components Broader Impact Preliminary experiments have shown that this methodology is indeed capable of producing chip design with better timing than any of existing systems for the test cases tried.