A method of simulating a multiprocessor system using cycle accurate or cycle approximate simulation

摘要

A method of simulating a multiprocessor system using cycle accurate or cycle approximate simulation, wherein processor instruction sets are simulated independently of each other for cycles that do not include inter-processor communication and (b) each processor instruction set is simulated step-by-step for the entire system for cycles that do include inter-processor communication. Trace data may be stored for each processor during a simulation, and may be merged together to generate cycle accurate information for the entire system. Preferably, the simulation involves the step of rewinding back a simulation. Conventionally, simulation would be necessary and inter-locked across all processors for all cycles, which is computationally very intensive, whereas simulating step by step for an entire multi-processor system for all cycles is more computationally more demanding than simulating as defined above.