Abstract : Energy budget is an intrinsic limitation of mobile electronic appliances such as smartphones or other connected objects. Energy saving is made to compete in modern design against performance optimization. Energy saving mechanisms are usually available (such as power or clock gating, frequency/voltage scaling), but methods to assess their efficiency operate at diverse levels. Either very high, with Excel-like formulas expressing cost functions in systems where applicative dynamic aspects are scarcely present, or already rather low with SystemC transaction-level simulation models in which dynamic abstractions of both actual applicative code and underlying Operating System and instruction-set architecture are needed. In the present paper we describe an intermediate level practical technique for exploiting power/performance ratio information with mathematical solving methods, such as Sat Modulo Theories (SMT) and Constraint Programming. The goal is to get estimation results and optimization proposals at the medium intermediate range, where application use cases have dynamic behaviors, but at a level much coarser than instruction-level code or even algorithmic function block. We present modeling and benchmark results illustrating our approach on a simple big.LITTLE-type architecture.