under supervision by
Dr. Christoph Kowitz (Infineon Technologies AG),
Dr. Gokcen Mahmutoglu (Infineon Technologies AG),
Friedrich Menhorn, M.Sc. (BGCE)
Introduction
The design of increasingly complex integrated circuits (ICs) demands sophisticated software along with powerful hardware to enable their simulation. As the exponential growth in single core performance comes to an end, adapting circuit simulation software to parallel chip architectures becomes increasingly relevant.
This project continues the work of the previous BGCE generation on a C++ analog circuit simulator (ACS) . The ACS is split into two parts: the compiler which parses netlists and transforms them into an intermediate representation (IR), and the kernel which performs analyses on a given IR circuit using the method of modiļ¬ed nodal analysis (MNA).
The core objective of this project is to parallelize the kernel of the ACS to speed up the computations. Different parallel numerical methods and libraries which can solve the types of systems of nonlinear and linear equations that result from the MNA approach are reviewed. A parallel programming paradigm best suited for those algorithms is implemented.