[dsengupta0628]

Summary

User activity annotations (set_power_activity -pins/-input_ports, read_saif, read_vcd) were lost or corrupted by netlist edits:

• scan_replace erased every pin annotation in the design, including pins it never touched.
• Pins deleted by cluster_flops and other edits left stale entries behind:
  • report_activity_annotation counted them;
  • -report_annotated could crash on them;
  • new pins created later could inherit their activity.
• The pins that replace a cell (the new scan flop, the multi-bit tray) started without the annotation of the pins they replaced, so their power fell back to the propagated estimate.

The OpenSTA side is already in upstream OpenSTA and in src/sta on master: Power keeps pin annotations across networkChangedNonSdc(), drops an annotation when its pin is deleted, and Power::hasUserActivity()/userActivity() are public.

This PR adds dbSta::copyUserActivity(from, to) and calls it before the old cell is destroyed:

• dft utils::ReplaceCell (scan_replace and its rollback): each old pin's annotation is copied to the new pin given by the port mapping. ReplaceCell now takes the dbSta.
• gpl MBFF::ModifyPinConnections (cluster_flops): each flop's D and Q/QN annotation is copied to the tray pin it is reconnected to.

New pins without a counterpart (scan in/enable) and shared tray pins (clock, scan enable, preset, clear) keep the default activity, as before.

Why copy the annotation

The new D/Q/QN pins carry the same signal on the same net as the pins they replace, so the measured activity still applies. Resizing (swapMaster) already keeps annotations, because the pins stay the same. After cluster_flops the original instance names no longer exist, so re-reading the SAIF cannot re-annotate the tray. Copying is the only way to keep the data.

Effect on the reported case

Step	Before	After
scan_replace	All annotations erased	Untouched pins keep theirs; the replaced flop's pins get theirs copied
cluster_flops	Tray outputs fall back to the propagated estimate; annotation count includes deleted pins	Tray D/Q/QN pins get the flops' annotations; count is correct
New cells after clustering	New pins inherit stale activity from deleted pins	New pins get their normal propagated activity
report_activity_annotation -report_annotated	Could crash after a delete	Lists only existing pins

Testing

• New dft/test/scan_replace_activity_sky130:
  • annotates ff1/Q, runs scan_replace;
  • checks that ff1/Q on the new scan cell keeps 0.05 with origin user.
• New gpl/test/mbff_activity:
  • annotates D and QN of four flops with distinct values, runs cluster_flops into a 4-bit tray;
  • checks that every tray D/QN pin carries the value of the flop placed in that slot.

Not covered

• A deleted top-level port (dbStaCbk::inDbBTermDestroy only disconnects it) still leaves a stale annotation.
• When rsz rejects a move and restores the netlist, re-created pins do not get their annotations back.
• Other transforms that change logic, nets or polarity (e.g. buffer to inverter pairs, restructuring) intentionally do not copy annotations.
• SDC constraints on clustered or replaced flops are not carried over (exceptions drop the deleted pins).

[gemini-code-assist]

Code Review

This pull request introduces functionality to preserve user, SAIF, and VCD activity annotations on pins during netlist edits, such as cell replacement and flop clustering. It implements copyUserActivity in dbSta and integrates it into the DFT cell replacement and GPL multi-bit flip-flop clustering flows, accompanied by new tests. The review feedback recommends adding defensive null checks for the input terms and their corresponding STA pins in copyUserActivity, as well as checking if the sta pointer is null in ReplaceCell before dereferencing it.

[dsengupta0628]

The OpenSTA change will come from upstream (I merged there), so needs to be pulled in 1st before this can be merged