We are surrounded by embedded computing systems, ranging from video-enabled mobile phones over real-time automotive applications to reliable medical devices. Just as the quality of an architectural blue-print determines the quality of the resulting building, the model of an embedded system is the key to its successful implementation. This project moves research and education on embedded system design forward in the area of system-level specification and modeling. While traditional work largely has focused on simulation and synthesis from a given system model, this project addresses the creation and optimization of system models for effective use in existing design processes. The results of this project are directly applicable to established system design flows in industry and fit well into existing and new courses in computer engineering education.
This project optimizes the modeling of embedded systems by use of four novel techniques. First, it advances a new model of computation, named ConcurrenC, which refines the generic capabilities of common C-based system-level description languages. Second, the creation of the system model is automated by computer-aided re-coding that derives an executable model directly from reference code. Third, the efficiency of the model is optimized using Result-Oriented Modeling (ROM), which, in contrast to traditional Transaction-Level Modeling (TLM), offers gains in simulation speed of multiple orders of magnitude and highest accuracy at the same time. Fourth, this project investigates TLM of computation, an area where it has not been applied before.
Everyday we are surrounded by embedded computing systems, ranging from video-enabled mobile phones over real-time automotive applications to reliable medical devices. Just as the quality of an architectural blue-print determines the quality of the resulting building, the model of an embedded system is the key to its successful implementation. This project has concentrated on two focus areas, namely (A) ConcurrenC and (B) Recoding. For focus area (A) ConcurrenC, we have studied a novel model of computation and have developed an out-of-order parallel simulator for embedded systems that is an order of magnitude faster than previous techniques. Our technique not only yields faster validation of embedded system models, it also fully preserves the accuracy in functionality and timing, and supports the traditional system-level description language semantics. For focus area (B) Recoding, we we have designed and developed novel model analysis methods and source code transformations. Our prototype integrated development framework allows the system-designer to quickly analyze and restructure his application models. Complex static dependency and conflict analysis can be performed in the editor and necessary source-code transformations can be applied with the click of a button. Our results on a set of multi-media applications, including a H.264 video codec, demonstrate the feasibility and show significant reduction in design specification and modeling time. Both developed techniques are applicable to the production of everyday computing devices and, in the end, result in better devices developed at less cost and in shorter amount of time. This research project has resulted two PhD dissertations and a total of 16 technical reports, 3 presentations and tutorials, 1 book chapter, 15 conference papers, 4 journal articles, and 1 book. Most notably, we are proud to report that our work has been recognized by the EDA community with an EDAA Outstanding Dissertation Award 2013 and a DATE Best Paper Award 2014. Intellectual Merit This project has advanced the research and education on embedded system design in the area of system-level modeling and simulation. While traditional work largely has focused on design and synthesis from a given system model, this project addressed the efficient creation, optimization, and validation of system models for effective use in existing design processes. The results of this project are directly applicable to established system design flows in industry and fit well into existing and new courses in computer engineering education. Broader impact In our everyday life, we are constantly surrounded by embedded computers. The model of such an embedded system is the key to its successful implementation. Thus, the better model generation and faster simulation techniques developed in this project directly improve the technical systems around us and the qualitity of life for everyone in our society.
