Title: Overview of the Model-based Integrated Simulation (milan) Framework

The Model-based Integrated simuLAtioN framework (MILAN) is a model-based, extensible environment that facilitates rapid evaluation of different system performance metrics, such as power, latency, and throughput, at multiple levels of granularity, of a large class of embedded systems by seamlessly integrating widely-used simulators into a unified environment. The MILAN framework is aimed at the design of embedded high-performance computing platforms, of System-on-Chip (SoC) architectures for embedded systems, and for the hardware/software co-design of heterogeneous systems. MILAN will be implemented with Model-Integrated Computing technology. The technology includes two major components—graphical models and automatic program/data synthesizers. The system will be built using the MultiGraph Architecture framework, a technology base that has been developed and applied over the past 15 years by a consortium of industry, government, and academia. These tools support multi-aspect model building and automatic application synthesis. The expected results of this effort are 1) characterization and representation of embedded systems (application, computing hardware and communication resources) using domain-specific graphical modeling environments where the entire design space is represented; 2) integration and/or development of power models for system components; 3) unified, multi-aspect representation of system models through a common database that acts as a repository for user-specified system parameters and results/statistics generated by the simulators; 4) a vertical, multi-resolution, extensible simulator integration framework that allows user-specified levels of simulation for different system components; 5) a high-level, system-wide, parameterized power-performance simulator to allow rapid evaluation of the target architecture with reasonable accuracy; 6) a constraint-based design space exploration and pruning tool that will narrow down the set of possible design alternatives based on a user-specified set of latency, throughput and power consumption constraints; and 7) demonstrations of both the vertical and horizontal simulator integration capabilities of the framework for use-case DoD applications such as Model-based Automatic Target Recognition. This report discusses the architectural design of MILAN, as well as the modeling concepts governing the use of MILAN, including application-and resource modeling. Constraint specification and representation are examined, as is design space pruning, as a means to reduce the size of the design space while presenting candidate solutions for detailed simulation and analysis. This report covers the use of MILAN to model both design-and processing alternatives, and discusses how those alternatives are to be used during both high-and low-level simulations of a system's performance and functional behavior. Integrating existing simulation and analysis packages such as MATLAB, SystemC, and SimpleScalar with MILAN is explored, …