Operation shuffling for low energy L0 cluster generation on heterogeneous VLIW processors

Abstract

Clustering L0 buffers is effective for reduction of energy consumption in the instruction memory hierarchy of embedded VLIW processors. However, efficiency of the clustering depends on schedule and assignment of a target application. This paper proposes a tool flow to explore operation shuffling for improving generation of L0 clusters. The tools explore assignment of operations for each cycle and generate various schedules. This approach makes it possible to reduce energy consumption for various processor architectures, however, the computational complexity is large because of huge exploration space. Therefore, some heuristics are also developed, which reduce the size of exploration space while the quality of solution remains reasonable. The experimental results indicate potential gain of up to 22.0% in energy of the operation shuffling for various heterogeneous processor architectures. Furthermore, the proposed heuristics drastically reduce the exploration search space by about 90%, while the results are comparable to full search, differences of up to 1%.