Explain the interaction between POCV and clock reconvergence pessimism removal on a clock path - do they double-count?

Short Answer

No - they correct different artifacts. CRPR (clock reconvergence pessimism removal) removes the double-counted delay on the clock segment shared by launch and capture; POCV's graph-merging pessimism removal separately removes the stochastic-max inflation that occurs where several fan-in distributions converge. A clock path needs both active.

Technical Explanation

• They fix two different kinds of pessimism, not the same one twice. CRPR removes the shared-clock-segment double count that on-chip variation (OCV) and POCV (parametric OCV, a statistical variation model) introduce whenever the launch and capture clock paths overlap - without it, the common segment gets charged two different delays at once, one for launch and one for capture.
• POCV's separate correction. Graph-merging pessimism removal corrects a different problem: where several fan-in paths converge at a gate, the statistical maximum of multiple delay distributions can exceed any single realistic delay through that gate. Removing this pessimism keeps the merged result from overstating the true worst case.
• Both can be active on the same clock path. The common clock segment needs CRPR to reconcile launch against capture. Any reconvergent structure in the clock tree, or converging data paths at the endpoint, separately needs graph-merging pessimism removal.
• Why this isn't double-counting. CRPR is about how one shared segment is treated differently on the launch side versus the capture side. Graph merging is about the statistical operation of combining several distributions into one worst case. Different mechanisms, different targets - fixing one does not touch the other.
• The POCV subtlety worth knowing. Under POCV, the delays CRPR reconciles are themselves distributions with a mean and a standard deviation, so CRPR operates statistically too - the common-path pessimism correction is computed on the distribution, not a single number. report_timing -variation (PT) shows the clock reconvergence pessimism line and the statistical graph pessimism line side by side, as two separate corrections.
• What a candidate must not conflate. Thinking graph-merging removal subsumes CRPR, or the reverse, misses that the two fix orthogonal artifacts and an accurate POCV clock path genuinely needs both.

Follow-up Question & Model Response

If you disabled CRPR but left POCV graph-merging pessimism removal on, what specifically would change in a report_timing -variation report for a path with a long shared clock segment?

Candidate Model Response: The graph-merging pessimism line would look unchanged, since that correction addresses fan-in convergence and doesn't touch the shared clock segment at all. What would change is the clock reconvergence pessimism line disappearing or reading zero, and the path's reported slack getting worse, because the shared segment between launch and capture would now be charged variation twice - once as if it were slow for the launch path and once as if it were fast for the capture path - with no credit given back. On a path with a long common clock segment, that double-charge can be a significant chunk of the reported slack, so the degradation would be visible and specific to paths sharing a lot of clock tree.