Transaction Level Modeling of Computation

The design of embedded computing systems faces great challenges due to the huge complexity of these systems. The design complexity grows exponentially with the increasing number of components that have to cooperate properly. One solution to address the complexity problem is the modeling at higher levels of abstraction. However, it is generally not clear which features to abstract (and to what extend), nor how to use the remaining features to create an executable model that allows meaningful, efficient and accurate analysis of the intended system. Transaction Level Modeling (TLM) is widely accepted as an efficient technique for abstract modeling of communication. TLM offers gains in simulation speed of up to four orders of magnitude, usually however, at the price of low accuracy. So far, TLM as been used exclusively for communication. In this work, we propose to apply the concepts of TLM to computation. While the two aspects of embedded system models, computation and communication, are different in nature, both share the same concepts, namely functionality and timing. Thus, TLM, which is based on the separation of functionality and timing, is equally applicable to both, communication and computation. In turn, the tremendous advantages of TLM can be utilized as well for the abstract modeling of computation. While traditional work largely has focused on refinement and synthesis tasks, this work addresses the modeling of systems towards efficient accurate estimation and rapid design space exploration. The results of this work will be directly applicable to established design flows in the industry.