Compiler Directed Battery-Aware Implementation of Mobile Applications

Handheld embedded systems are crucial to obtain high performance for execution time as well as efficient battery usage. Unfortunately, most of the compilation techniques to obtain an efficient binary code for complex multimedia processors lack the optimal energy-cycle code. This work describes the methodology of the compilation for next generation handheld devices, which will support data-compute intensive applications at small form factors. We designed and implemented optimization framework which supports conventional code optimization scheme along with additional energy efficient benefits for multimedia applications such as MPEG-2 transcodec, H-264L transcodec. Optimization space is searched with genetic algorithm. Whole scheme reduces energy, on both as per cycle basis and as the total energy used over the lifetime of an application. The optimized G-728 audio codec meets real-time constraints on the Nexperia series of multimedia processor with low energy consumption. Furthermore, the performance improves by a factor of 0.489 and the energy consumption decreases by a factor of 0.203 over the baseline executable code