IMPLEMENTATION

Close Timing with Physical Evidence

Use each implementation stage to replace estimates with evidence: congestion after placement, skew after CTS, and wire RC after routing. Fix root causes through a controlled loop and preserve proof that every change helped the intended corner.

Inputs

Placed floorplan/PDN and legal cell libraries
Multi-mode/multi-corner SDC and RC assumptions
Clock-tree cell policy, routing layers, and NDR policy

Method

Run global placement and inspect utilization, overflow, congestion, and critical-net topology before committing timing-driven optimization.
Legalize with targeted padding near pin-dense or high-fanout structures, then re-measure congestion instead of applying blanket whitespace.
Build CTS with characterized slew/capacitance limits and an approved buffer list; review skew, insertion delay, and transition at each clock group.
Route globally then in detail; repair violations by correlating failing paths with detours, vias, blockage, and localized congestion.
At every ECO, compare setup, hold, area, power proxy, congestion, and DRC trend across all active scenarios; retain only measured improvements.

Deliverables

Placement/route databases and congestion reports
CTS report with clocks, buffers, skew, and transition
Scenario-tagged timing-closure/ECO log

Pitfalls

Pre-route or ideal-clock slack is a ranking signal, not final closure evidence.
A local setup repair can create hold, congestion, power, or DRC regressions elsewhere; evaluate all active scenarios after each change.

Sources

Global Placement ↗
OpenROAD · Timing-driven placement controls

Clock Tree Synthesis ↗
OpenROAD · CTS characterization and reporting

OpenROAD Developer Guide ↗
OpenROAD · Flow stages and late-stage parasitic context