BOTS: Balanced-Objective Task Selector for Application Graph Scheduling in Reconfigurable Computing Systems

Hardware-based execution capability with resource reusing flexibility is one of the new features of Reconfigurable Computing (RC) systems. However, due to the reconfiguration constraints, it is difficult to fully exploit their advantages. For a better design with higher performance, several pivotal objectives should be investigated simultaneously such as task execution time, communication data between tasks, and required resources for implementation. In this paper, a new algorithm for task selection in an application graph has been introduced to balance objectives in RC system. This algorithm decides based on a selection merit to reduce the number of execution clusters and to save communication cost as much as possible. To remove forbidden solution, a set of free tasks has been constructed to satisfy the precedence constraints. The evaluation has been performed with different topological features such as parallelism degree and critical path. Results showed that the proposed method provides acceptable improvement in comparison with previous method for application execution time metric