Browse Prior Art Database

Statistical algorithm for evaluation of interconnection design

IP.com Disclosure Number: IPCOM000015417D
Original Publication Date: 2002-Nov-15
Included in the Prior Art Database: 2003-Jun-20

Publishing Venue

IBM

Abstract

A statistical framework is disclosed that assesses the effectiveness of incorporating a pre-route algorithm in the ULSI design process. The main emphasis is on incorporating intervention with custom algorithms in the design process in a seamless manner. The emphasis here is to formulate a way of thinking about this class of problem. Let R i denote the region of influence that physically encloses the new custom interconnections in trial i. Examples of the four types of design routes in the previous trial (i-1) are: the custom interconnections, targeted routes in R i , routes that partially pass through R i , and the remaining routes that do not pass through R i . The total route length of all signal routes in trial i is composed of four separate components: the total length of new custom interconnections contained in R i in trial i; the total length of all custom routes in previous trials 0 to (i-1); the total length of route segments routed by the routing system where at least part of the signal route passes through R i ; and the total length of route segments routed by the routing system where no part of the signal route passes through R i . The wirelength of each signal route in the design is included in one of these four terms. Assignment of the wirelength to the appropriate term occurs according to the following procedure: If the signal route is a custom route in a previous trial, the wirelength of this route is included in the first term; if the signal route is a new custom route in trial i, the wirelength is included in the second term; if any section of the route is located in R i , the wirelength of the entire signal route is included in the third term; if all sections of the route are located outside R i , the signal wirelength is included in the fourth term. The expression for the total number of vias in all the signal routes in trial i is obtained by substituting vias for netlength in each term.