Due to the power constraints of the current semiconductor technology, energy consumption has become an important factor for computer systems. Reducing energy consumption can mean more battery life for mobile devices or reduction of financial costs for data centers. One of the energy bottlenecks of computer systems is the information traffic between the processor and memory hierarchy. In this paper we evaluate the energy reduction of our new spill code minimization technique called color flipping in comparison with classical approaches. We implemented the Briggs' register allocator in the LLVM compiler framework with and without color flipping strategy and we ran some SPEC CPU 2006 benchmarks in a modified gem5 simulator for Cortex-A9. Then the energy consumption was estimated using the McPAT framework. Experimental results showed that our technique can reduce about 1% of the energy consumption of integer programs.