|
|
Article: 4 System-level statistical methodologies.(Statistical Performance Modeling and Optimization)
- Article from:
- Foundations and Trends in Electronic Design Automation
- Article date:
- December 1, 2006
- Author:
CopyrightCOPYRIGHT 2006 Now Publishers, Inc. This material is published under license from the publisher through the Gale Group, Farmington Hills, Michigan. All inquiries regarding rights should be directed to the Gale Group. (Hide copyright information)
|
As we move from transistor level toward system level, one major challenge for statistical analysis and optimization stems from the underlying large problem size that has a twofold meaning. First, a full integrated system typically contains millions of transistors and, hence, the circuit size is huge. Second, to capture the intra-die variations of the entire system, a large number of random variables must be utilized. For example, if we attempt to model the per-transistor random doping variation that is expected to be the dominant variation component at 45 nm technologies and beyond [4], the total number of random variables may reach [10.sup.3] ~ [10.sup.6] for large-scale ...