Power performance analysis of 3-D finite element mesh refinement with tetrahedra by CUDA/MPI on multi-core and GPU architecture

Software power performance tuning handles the critical design constraints at the level of software that run beyond hardware platforms composed of large numbers of power-hungry components. The power dissipation of a Single Instruction Multiple Data (SIMD) program such as FEM mesh refinement is highly dependent on its underlying algorithm and the power-consuming features of hardware Processing Elements (PE). This contribution presents a practical methodology for modeling and analyzing the power performance of parallel 3-D FEM mesh refinement on CUDA/MPI architecture based on detailed software prototypes and power parameters in order to predict the power functionality and runtime behavior of the algorithm, optimize the program design and thus achieve the best power efficiency.