Also prof. at K.U.Leuven

This paper presents an exploration algorithm which ex-amines execution time and energy consumption of a given application, while considering a parameterized mem-ory architecture. The input to our algorithm is an appli-cation given as an annotated task graph and a specifi-cation of a multi-layer memory architecture. The algo-rithm produces Pareto trade-off points representing dif-ferent multi-objective execution options for the whole application. Different metrics are used to estimate pa-rameters for application-level Pareto points obtained by merging all Pareto diagrams of the tasks compos-ing the application. We estimate application execu-tion time although the final scheduling is not yet known. The algorithm makes it possible to trade off the qual-ity of the results and its runtime depending on the used metrics and the number of levels in the hierarchical com-position of the tasks ’ Pareto points. We have evaluated our algorithm on a medical image processing applica-tion and randomly generated task graphs. We have shown that our algorithm can explore huge design space and ob-tain (near) optimal results in terms of Pareto diagram quality. 1