Low Power Bus Binding Exploiting Optimal Substructure

The earlier the stage where we perform low power design, the higher the dynamic power reduction we achieve. In this paper, we focus on reducing switching activity in high-level synthesis, especially, in the problem of functional module binding, bus binding or register binding. We propose an effective low power bus binding algorithm based on the table decomposition method, to reduce switching activity. The proposed algorithm is based on the decomposition of the original problem into sub-problems by exploiting the optimal substructure. As a result, it finds an optimal or close-to-optimal binding solution with less computation time. Experimental results show the proposed method obtains a solution 2.3–22.2% closer to optimal solution than one with a conventional heuristic method, 8.0–479.2 times faster than the optimal one (at a threshold value of 1.0E+9). key words: low power, bus binding, switching activity, table decomposition, optimal substructure