ABSTRACT
System level modeling and faster architectural exploration is becoming very essential to achieve a better cost-performance trade off, short time to market, to determine the success and failure of the system and to narrow down the design space early in the design stage [1]. Simulation based system level modeling and analysis involves dynamic and statistical computation of design matrices at different workload events and transitions. This method is faster than RTL (Register Transfer Level) and more accurate than Analytical method [2]. Difficulty in simulation based system level modeling is that the integrating the models of system components at different levels of details with different modeling languages. That leads to increase in learning curve, modeling effort and modeling time. To address this problem, we need to have common platform and method to integrate the analysis specific system component models at different levels of modeling details. Work carried out is to explore the modeling flow for simulation based system level performance modeling, analyzing and optimization using analysis specific generalized parametric models of system components. In this work, system architecture and task mapping is determined based on the analysis and optimization of end to end latency, resource cost and utilization. System components such as multi-core processor, memory, and bus and IO unit can be characterized by generalized standard abstract modeling libraries.
