Timing Optimization Through Clock Skew Scheduling

The focus of this book is on timing analysis and optimization techniques for circuits with level-sensitive memory elements (registers). Level-sensitive registers are becoming significantly more popular in practice as integrated circuit densities are increasing and the `performance-per-power' metric for integrated circuits becomes a key issue.

Therefore, techniques for understanding level-sensitive based circuits and for optimizing the performance of such circuits are increasingly important. The book includes the following major topics in the timing analysis and optimization of level-sensitive circuits: A linear programming (LP) formulation applicable to the timing analysis of large scale circuits.

The formulation uses a variation of the big M method - called the modified big M method - to transform the non-linear constraints in the problem formulation into solvable linear constraints. This LP formulation is computationally efficient and demonstrates significant circuit performance improvement.

By making maximum use of cycle stealing, operation at a higher clock frequency (reduced clock period) is possible. A delay insertion methodology that improves the efficiency of clock skew scheduling in level-sensitive circuits. It is shown that re-convergent paths limit the improvement of circuit performance that can be achieved through clock skew scheduling.

The described delay insertion method mitigates the limitations cause by re-convergent data paths and improves the results of timing optimization (for increased clock frequency). A summary of circuit partitioning, placement and synchronization methodologies that enables the implementation of high speed, low power circuits synchronized with ultra modern resonant clocking technology (such as traveling oscillators/waves).

The described framework includes the particular circuit partitioning and placement methodologies that permit the hierarchical application of non-zero clock skew system timing in resonant clocking based circuits. A framework for and results from implementing the described timing optimization algorithms in a parallel computing environment.

As multi-core microprocessors become commonplace, computationally intense algorithms can benefit greatly by exploiting this available parallelism. The framework uses a heuristic approach to generate circuit partition and solve them independently on processors/computers working on parallel. This is one of the first such applications of explicit parallelism in Electronic Design Automation (EDA), and, will be of great interest to practicing EDA engineers.