Global weighted scheduling and allocation algorithms

Abstract

Scheduling and allocation are very complex problems in a high-level synthesis system. It was proven, in related work, that the two are NP-complete optimization problems. The authors introduce a new global approach for scheduling and allocation. The approach uses graphs to formulate the two problems and applies a partitioning procedure on these graphs to find the minimal number of cliques. The obtained cliques correspond to the time steps in scheduling and to hardware elements required in allocation. The partitioning procedure is made more efficient by weighting the graph edges by the profit to group nodes together. The procedure was programmed in C++ and experimental results are given to show its efficiency to solve both scheduling and allocation.<>