A fast and accurate instruction-oriented processor simulation approach

This paper proposes an efficient instruction-oriented approach to improve processor simulation speed while maintaining perfect timing accuracy. Unlike traditional approaches, we simulate only necessary states instruction-by-instruction, instead of updating all states cycle-by-cycle. This approach can be applied to various types of processors, including superscalar processors. The experimental result shows that the simulation performance is about twenty times faster than the traditional Cycle-Accurate approach while with 100% timing accuracy.