An Efficient Instruction-Level Energy Estimation Model for Embedded Systems

To optimize the energy consumption embedded systems, the estimation of energy consumption of the embedded applications are very important. This paper proposes a simple but effective instruction-level energy estimation model for embedded systems. For case study purposes, the model parameters were determined for a commonly used ARM9TDMI-based microcontroller. The total energy consists of the energy consumed by the processor core, flash memory, memory controller, SRAM etc. The model parameters that are determined includes op-code of instructions, number of shift operations, register bank bit flips, instructions weight and their Hamming distance, different types of memory accesses, the effect of pipeline stalls etc. To validate the proposed model, a physical hardware platform was developed which is having energy measurement capabilities. For several experiments conducted on various embedded applications from MiBench benchmark suite and less than 6 % error in the energy consumption estimation was shown. Also an energy profiler tool was developed for the systems that use ARM9TDMI processors which provides valuable information and guidelines for software energy optimization