Mobile appliances need to work at a low level of power as well as perform at a high level. As energy consumption and execution speed are in a reciprocal relation, using a combination of a multicore processor system and its controlling software is a key to reducing energy. We are developing an on-chip multicore architecture for embedded systems with high performance at low power and cost. We propose a method of reducing the energy consumed for this system during execution by parallelizing and applying Dynamic Voltage Frequency Scaling (DVFS) to each element processor in accordance with its workload. The program's workload at runtime needs to be accurately estimated, and the most appropriate frequency and voltage for each processor running an application is chosen. We propose and evaluate a method for predicting the progress of an application by inserting checkpoints and getting a minimal number of processors for the least energy.