Computing Leakage Current Distributions and Determination of Minimum Leakage Vectors for Combinational Designs

Computing Leakage Current Distributions and Determination of Minimum Leakage Vectors for Combinational Design. (May 2006) Kanupriya Gulati, B.E., University of Delhi Chair of Advisory Committee: Dr. Sunil Khatri Analyzing circuit leakage and minimizing leakage during the standby mode of operation of a circuit are important problems faced during contemporary circuit design. Analysis of the leakage profiles of an implementation would enable a designer to select between several implementations in a leakage optimal way. Once such an implementation is selected, minimizing leakage during standby operation (by finding the minimum leakage state over all input vector states) allows further power reductions. However, both these problems are NP-hard. Since leakage power is currently approaching about half the total circuit power, these two problems are of prime relevance. This thesis addresses these NP-hard problems. An Algebraic Decision Diagram (ADD) based approach to determine and implicitly represent the leakage value for all input vectors of a combinational circuit is presented. In its exact form, this technique can compute the leakage value of each input vector, by storing these leakage values implicitly in an ADD structure. To broaden the applicability of this technique, an approximate version of the algorithm is presented as well. The approximation is done by limiting the total number of discriminant nodes in any ADD. It is experimentally demonstrated that these approximate techniques produce results with quantifiable errors. In particular, it is shown that limiting the number of discriminants to a