ASIP Instruction Encoding for Energy and Area Reduction

Application-specific VLIW processors provide an energy and area efficient solution for high-performance embedded applications. One significant design issue is that the long instruction word required to express the instruction parallelism represents a significant cause of energy dissipation. We present an application- tailored instruction encoding solution that modifies the instruction architecture to minimize the instruction word width. We demonstrate the effectiveness of our solution across a range of benchmarks, resulting in average energy savings of 20% and an average area reduction of 18%, with no performance penalty.