A Linear Programming Approach for the Estimation of an Upper Bound on the Maximum Power of CMOS Circuits

|Maximum instantaneous power in VLSI circuits is of great importance to the design of power and ground lines in a VLSI circuit. Underestimating the maximum instantaneous power greatly reduces the circuits reliability. An accurate estimate of the maximum instantaneous power is therefore needed. Unfortunately , nding the input vectors which give the maximum power consumption is a combinational problem. An exhaustive search is very CPU-time intensive even for small circuits with few primary inputs. We therefore propose a novel linear programming approach, using variables denoting switching scenarios, instead of logic levels between possible switching events, to estimate an upper bound on the maximum power consumption in a CMOS circuit under an input vector change. We present results demonstrating the approach and discuss the beneets and limitations of three linear programming solvers for our type of linear programming problem.