fesens/opentpu/4-ppa-checker

Post-Route Timing Check (OpenROAD)

Used inside a macro placement exploration loop. After routing completes in OpenROAD, this skill reads the design's timing health and decides whether the loop should stop or continue with another placement candidate.

When to run

Run this skill only after detailed routing has finished for a placement candidate. Pre-route or post-CTS slack is based on estimated parasitics and will mislead the exploration loop — drop those numbers, don't act on them.

Commands

In the OpenROAD shell (or .tcl script), run both, in this order:

report_tns
report_wns

• WNS (Worst Negative Slack) — the single worst path. Tells you if any path is failing.
• TNS (Total Negative Slack) — the sum of negative slack across all failing endpoints. Tells you how much failure there is in aggregate.

Both are reported in the design's time unit (typically nanoseconds). A clean design reports WNS = 0 and TNS = 0. Negative values indicate failing paths.

Stop criteria

Stop the macro placement exploration when either condition is true:

1. Timing goal met — WNS >= 0 AND TNS >= 0 (fully clean timing, no negative slack anywhere).
2. Try limit exhausted — the exploration has used its budget of placement attempts.

If neither condition is true, return control to the placement loop to try the next candidate.

Why both conditions matter: without the try limit, a design that cannot close timing under any placement will loop forever. Without the goal check, you'll burn the full budget even on a placement that already closes on iteration 2.

Reporting

After each check, report back in this exact format so the loop driver can parse it:

Iteration: <N> / <try_limit>
WNS: <value> ns   (goal: >= 0)
TNS: <value> ns   (goal: >= 0)
Decision: STOP (goal met) | STOP (try limit exhausted) | CONTINUE

Keep the output terse — this skill runs inside a loop and verbose output clutters the exploration log.