Simulation time reduction techniques for computationally intensive algorithms

Complex algorithms particularly for real-time computer vision applications require significant amount of simulation time when implemented using Hardware Description Languages (HDLs). This increase in simulation time is mainly due to the use of OEM supplied IP Cores. In this paper we propose a methodology involving: 1) design partitioning, and 2) use of our proposed functional models; to perform simulation. We take Locally Normalized Cross Correlation (LNCC) Algorithm which is a good candidate for computer vision applications in Unmanned Aerial Systems (UAS). The proposed methodology resulted in 90x faster simulations when compared to the conventional non-partitioned, synthesizable IP approach.