Static Timing Analysis Pessimism Reduction via Common Point Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Graph-based static timing analysis (GBA) introduces excessive pessimism due to advanced on-chip variation (AOCV) analysis, leading to underestimated path slack values and increased development and verification efforts, as well as performance and power profile issues, caused by conservative derating factors and merging of timing signals from different common points.
Innovation Solution
The method involves identifying a common point for a subset of timing paths, treating AOCV depth and distance as attributes of timing signals rather than fixed values, and calculating derating factors based on these attributes to reduce pessimism by avoiding the merging of signals from different common points and using worst-case metrics only within the subset.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If AOCV analysis is applied with path-based methodology, then timing analysis accuracy is improved, but excessive pessimism is introduced due to worst-case metrics and merging of timing signals from different common points
Solution Approach 1:
The patent segments the timing analysis by identifying distinct common points in the clock path and treating timing signals originating from different common points as separate entities. This segmentation prevents the erroneous merging of timing signals and allows for more accurate, less pessimistic slack calculations by analyzing each segment independently rather than applying worst-case metrics across all paths uniformly.
Solution Approach 2:
Instead of applying worst-case AOCV metrics to all timing paths and then attempting to reduce pessimism, the patent inverts the approach by first identifying and separating timing signals from different common points, then applying AOCV analysis only within each segmented group. This inversion从根本上 prevents the introduction of excessive pessimism rather than trying to mitigate it afterward.
2Reliability
If global derating factor is applied to all cells and interconnects, then timing variations are accounted for, but too pessimistic estimation results due to increasing influence of on-chip variations
Solution Approach 1:
The patent applies local quality by transitioning from a global derating factor applied uniformly to all cells and interconnects to a localized approach where AOCV metrics are calculated specifically for each timing path segment between common points. This allows the derating to be tailored to the actual on-chip variation characteristics of each local region, reducing overall pessimism while maintaining reliability.
3Device complexity
If timing signals from different common points are merged, then resource requirements are reduced, but pessimism increases due to conservative common point selection
Solution Approach 1:
The patent segments timing signals by their common point origin and maintains separate analysis for each segment. This prevents the merging of signals from different common points, thereby avoiding the introduction of pessimism that would result from conservative common point selection, while still managing resource requirements through efficient processing of segmented data.
4Ease of operation
If worst-case AOCV metrics are applied at each timing arc, then timing analysis is simplified, but merging pessimism is introduced affecting development and verification effort
Solution Approach 1:
The patent segments the application of AOCV metrics by restricting them to timing arcs within the same common point group. This segmentation maintains the simplicity of applying worst-case metrics locally while avoiding the merging pessimism that would result from applying these metrics across all timing arcs regardless of common point, thereby reducing unnecessary development and verification effort.
Data Source
AI summary
A method for performing graph-based static timing analysis comprises reading in a design of an integrated circuit having a subset of timing paths, each timing path of the subset having a common point, wherein the common point is identical for all timing paths of the subset. The method comprises initiating a timing signal at the common point, the timing signal propagating along a plurality of timing arcs of the subset. The timing signal has a plurality of attributes varying with the propagation including a depth value and/or a distance value. The method comprises determining a derating factor for a delay of at least one of the plurality of timing arcs depending on the depth and/or the distance value of the timing signal at a pin of said at least one timing arc, and generating a timing report based on the derating factor.