A fast and accurate technique to optimize characterization tables for logic synthesis

Cell characterization data is used by synthesis and timing verification tools to compile and validate a cell netlist which meets timing constraints imposed by the designer. Characterization tables contain data for multiple, simple equations representing a cell's behavior and are an alternative to the single, monolithic characteristic equation. Data in the table is fit to a function whose form is fixed by the application, and the cell's response is interpolated from the function. Tables can potentially increase accuracy, but large tables can cause a program to use dramatically more memory and run much slower. The optimization of characterization tables, in which accuracy is maintained but table size is significantly reduced, is important if large programs, such as synthesis, are to complete accurately and in a reasonable runtime. In this paper we address some of the issues involved in characterizing cells and optimizing characterization tables quickly and accurately. Experimental results from the use of these techniques within AMD for a Synopsys cell library is also presented.