Multi-Objective Optimization: Runtime Efficiency vs. Energy Efficiency

Heterogeneous hardware systems with different types of compute devices have grown steadily in the field of high performance computing over the past years. Time-efficient distribution of work onto heterogeneous devices is much more challenging than in the homogeneous case. In addition to runtime efficiency, heterogeneous devices raise a second issue - energy efficiency. Optimizing in one direction neglecting the other one is not adequate for such an environment. In this talk we analyze this multi-objective optimization problem in detail and present optimal solutions with respect to runtime efficiency and energy efficiency.