A Markov chain sequence generator for power macromodeling

In macromodeling-based power estimation, circuit macromodels are created from simulations of synthetic input vector sequences. Fast generation of these sequences with all possible statistics is crucial for ensuring the accuracy and speed of macromodeling. In this paper, we present a novel sequence generator based on a Markov chain model. Specifically, we formulate the problem of generating a sequence of vectors with given statistics as a transition matrix computation problem, in which the matrix elements are subject to constraints derived from the specified statistics. We also present a practical heuristic that computes such a matrix and generates a sequence of  ¢¡ -bit vectors in £¥¤ ¡¦ ¨§©¡ time. Our generator is guaranteed to yield vector sequences with a given average input probability  , average transition density  , and spatial correlation  , or reports that the specified sequence type does not exist. Derived from a strongly mixing Markov chain, it generat...

• Type: Conference
• Year: 2002
• Where: ICCAD
• Authors: Xun Liu, Marios C. Papaefthymiou
• Keywords: Binary Vector Sequences, Hardware, ICCAD 2002, Input Vector Sequences, Transition Matrix Computation