This research is on finding pipeline schedulers for the behavioral description of an iterative or recursive algorithm or loops. This extends the scheduling algorithm to more realistic resource models and to graphs with conditionals. The goal is to expose the parallelism in an iterative or recursive algorithm to provide information for optimization. Focus is on the innermost loops or iterations, which are the most time-critical for applications. The approach is to study the structure of cyclic data-flow graphs with edge delays in order to optimize behavioral transformation, scheduling and partitioning.
