High efficiency analysis and optimization algorithms in electronic design automation

This thesis focuses on the development of efficient algorithms for solving three CAD problems encountered in the design of high performance VLSI circuits. In the first part of the thesis, several Green function-based thermal simulation algorithms are presented, and they can be used, respectively, to perform localized thermal analysis, full chip temperature profiling, and thermal simulations where the accuracy requirement differs from place to place over the same chip. The accuracy and efficiency of the algorithms are demonstrated through comparisons with the results from a commercial computational fluid dynamic software for thermal analysis. The second part of the thesis is concerned with efficient module assignment in pin-limited designs under the stacked-Vdd paradigm. The stacked-Vdd paradigm, in which modules are distributed to different Vdd domains so that currents consumed by modules in one Vdd domain can be recycled by modules in another Vdd domain, is a promising candidate in resolving the I/O pin limitation problem that plagues many VLSI circuits with high performance and high integration density. However, a significant amount of power may be wasted in a stacked-Vdd circuit if the module assignment is not performed carefully. In this part of the thesis, a graph partition-based algorithm is developed to assign modules to different Vdd domains, and experimental results show that it is much more effective in reducing the power waste in stacked-Vdd circuits than a bin-packing technique. Finally, in the last part of the thesis, the optimization of on-chip spiral inductors is studied. A sequential quadratic programming approach is used to optimize the quality factor Q of inductors. The optimization routine is built around an extraction engine that is based on the physical model of on-chip inductors, and experimental results show that high quality inductors can be obtained using this approach within a very reasonable amount of time.