Energy-aware parallelization flow and toolset for C code
Lazarescu, Mihai T and Cohen, Albert and Guatto, Adrien and L{\^e}, Nhat Minn and Lavagno, Luciano and Pop, Antoniu and Prieto, Manuel and Terechko, Andrei and Sutii, Alexandru
Abstract
Multicore architectures are increasingly used in embedded systems to achieve higher throughput with lower energy consumption. This trend accentuated the need to convert existing sequential code to eectively exploit the resources of these architectures. We present the work-in-progress of the EU FP7 PHARAON project that aims to develop a complete set of techniques and tools to guide and assist the designer in the development process for heterogeneous parallel architectures. We focus on the legacy C code parallelization ow that includes a performance estimation tool, a parallelization tool, and a streaming-oriented parallelization framework. We demonstrate the eectiveness of the use of the toolset on a use case where we measure the quality and time for parallelization for inexperienced users and the parallelization ow and performance results for the parallelization of a practical example of a stereo vision application.