This paper describes a dynamic VDD switching technique to reduce energy dissipation of Dynamically Reconfigurable Processors. Either high or low supply is dynamically selected at each PE at the context-by-context basis. We designed a part of a PE array and applied this technique. A test chip fabricated in 65nm technology operated successfully. Detailed simulations revealed that energy reduction is hindered by energy overhead due to supply switching when we use even lower VDD. We propose a mapping optimization algorithm "PFCM" to minimize the overhead. PFCM reduced energy overhead by 90.8% and thereby the dynamic VDD switching technique reduced energy dissipation by up to 12.5% when running sepia filter, alpha blender and Laplacian filter programs.