A shared-memory multiprocessor logic simulator

A simple multiprocessor logic simulation kernel is described and preliminary benchmark results presented. Communication between processors is through shared global memory, with locking/unlocking provided by the host operating system. Partitioning the gate evaluation list and current event queue among processors gives a dynamic self-balancing effect which tends to evenly distribute the load associated with large simulations. The kernel is written entirely in C and currently runs on a Balance 8000 system under the DYNIX 2.1 operating system. For a 3800-gate combinational circuit, 90% efficiency is achieved for a two-processor simulation.<
>