Scalable Algorithms Enabled by Problem Structure

of results: This project extended the framework of timing analysis that we developed under a previous NSF award (MIP-9014058) by adding sufficient functional (logical) content in the models to permit more accurate (less pessimistic) performance estimation while maintaining, or not significantly reducing, the level of coverage and efficiency of the functionless analysis. The most significant result of this work has been the development of a symbolic functional waveform model that provides a consistent mathematical basis for timing analysis and simulation centered around a differential calculus that allows us to describe the transient behavior of logic waveforms in terms of appropriate time derivatives. This model highlights the intrinsically conditional nature of signal delay and can be shown to encompass most delay models proposed to date. It also, for the first time, allows precise functional and temporal abstractions to be made in order to generate compact “system-level” timing views with guaranteed accuracy bounds. An unanticipated but highly significant result has been the development of GRASP, a SAT solver that is central to the current proposal scalable algorithm development. GRASP was originally developed as part of an efficient search-based functional timing analysis methodology. We have since found it to be applicable to, and surprisingly effective in, many other problem areas. Two students who were supported on this project have graduated: João P. Marques Silva and V. Chandramouli.