This research is on optimization algorithms for synthesis of multi- level loops, which occur in time and memory critical parts of scientific computing applications. The nested loops are modeled as multi-dimensional data-flow Graphs (MDFG); and algorithms taking advantage of the multi-dimensionality are being designed. By considering the multi-dimensional iteration space and the iteration body simultaneously, the transformation and optimization techniques are able to optimize throughput and memory requirement at the behavior level. Research topics include: graph transformation and optimization; data scheduling; and co-design. Polynomial-time algorithms for various graph models are being developed. This avoids exponential integer linear programming approaches.
