Model compiler driven device modeling and circuit simulation

Model compilers had been proposed and developed in recent years to automatically compile a mathematical description into implementation that can be used directly for circuit simulation. But those model compilers have a severe limitation: the compiled models are not as efficient as manually crafted models. In this dissertation, various techniques---function strength reduction, function call minimization, division reduction, and efficient Automatic differentiation---are proposed and implemented in model compiler MCAST. With those techniques, MCAST can generate compiled models that are as efficient as or better than manually implemented models.; In this dissertation, I also presented a framework that combines both model compiler MCAST with high level hardware description languages VHDL-AMS to model dynamic nonlinear systems and compile them into simulatable modules that can be either directly or indirectly linked to available simulation engines for further analysis. As a case study, an industry design (6 bit ADC) was modeled and simulated in mixed level (transistor and behavioral level) to analyze its nonlinearity.; In addition, this dissertation presents a new simulation approach capable of simulating hundreds and thousands Process, Voltage and Temperature (PVT) corners with the computational cost comparable to or even less than that of a few corner simulations, yet with the same simulation accuracy and robustness. The proposed approach is based on the combination of the LU-factorization based direct method and Krylov subspace based iterative methods to explore the common characteristics shared by a circuit under all PVT corners. The key novelty is a systematic method that uses as few LU based direct solving as possible for underlying linearized systems, and then solves the rest of linearized systems across the entire PVT linear system space using Krylov subspace based iterative methods with preconditioners computed from those LU factors.; MCAST is enhanced to support automatic generation of piece wise non-linear (PWNL) models for the iterative solver based simulators. Iterative solvers require preconditioners and PWNL models are used in constructing efficient preconditioners. With the support from MCAST, iterative solvers can be more widely applied to simulate circuits with new devices.