Simultaneous memory and bus partitioning for SoC architectures

There has been a continued proliferation in the demand for application specific system on chip cores in the recent years. Meeting the power budget constraint continues to be a major challenge for the designers architecting such systems. In this work, we demonstrate that simultaneous partitioning of the bus and memory subsystem into smaller segments can be an effective mechanism for reducing the energy consumption of a SoC. We present a genetic algorithm based search mechanism to determine a system configuration that is energy-efficient and validate the effectiveness of the configuration using a cycle-accurate virtual platform for a multiprocessor SoC. Our results using various applications show that the proposed approach gives significant energy savings and accentuates the benefits of previously proposed bus and memory partitioning schemes applied individually or in combination.