Mapping Specification-level primitives to IP-primitives: A Case Study

In system-level platform-based embedded systems design, it is crucial to model accurately and efficiently all steps needed when mapping an application specification onto a pre-specified architecture specification. The availability of an accurate high-level mapping method leads to: 1) a fast and accurate performance estimation and 2) an efficient designspace exploration. While our previous work was directed mainly towards the first goal (i.e., a fast and accurate performance estimation), this paper introduces a method to refining primitives from a task in an application-specification without rewriting the original code. Instead, we refine a task representation consisting of a symbolic program and a control trace