Two reports of the same design disagree. Before anyone argues about which is right, the disagreement has to be localised. This topic gives a staged comparison that finds the first point of divergence, two cheap sanity checks for libraries and parasitics, a heuristic for suspicious constraint changes, and a triage checklist driven by the distribution of the differences.
Lesson | After it you can... |
|---|---|
Find the first stage at which two flows diverge. | |
Detect unit errors and swapped corners. | |
Flag a suspicious improvement from a constraint change. | |
Classify a difference from its distribution. |
This topic extends Signoff vs Implementation Views from measuring correlation to localising a break.
If the cause is a global setting, every path is affected the same way and debugging one path rediscovers it a thousand times; locating the stage and distribution first tells you what kind of problem it is.