How does fix_eco_power recover area and leakage, and what selection modes exist?

Short Answer

fix_eco_power (PT) swaps cells for cheaper ones on paths with positive setup slack, or removes buffers on paths with positive hold slack, and backs out any change that creates or worsens a timing or DRC violation. What counts as cheaper depends on the mode: smallest area by default, a numeric library attribute with -power_attribute, a name or string priority list with -pattern_priority, or PrimePower data with -power_mode. You choose one mode per run and run the command again for another.

Technical Explanation

• It is iterative. It makes targeted changes, re-analyzes timing and DRC, backs out changes that hurt, explores further swaps and repeats until the cost of more progress exceeds the benefit.
• A replacement happens only if it creates or worsens no timing or DRC violation, has the same logic function, meets the eco_alternative_cell_attribute_restrictions (PT) rules, and is preferred under the active mode.
• -pattern_priority {HVT NVT LVT} uses name prefixes, listed lowest-power first. Add -attribute when a user-defined string attribute carries the priority instead of the name.
• -power_mode takes dynamic, leakage or total and uses PrimePower results. It cannot be combined with -power_attribute or -pattern_priority. In total mode, high switching activity favours downsizing and low activity favours a higher Vt.
• PrimePower data needs power_enable_analysis (PT) set to true, activity read in, and update_power (PT). In DMSA you name exactly one -dynamic_scenario and one -leakage_scenario, while timing and DRC in all scenarios are still honoured.
• fix_eco_power -methods {remove_buffer} (PT) removes buffers from paths with positive hold slack.
• It appears twice in the recommended order: first for sizing and buffer removal, and last as a Vt swap that changes no layout.
• -pba_mode path or exhaustive finds slack that graph-based pessimism hides, so more cells qualify for a swap.

What To Check

• Power before and after with the same activity and scenario.
• Setup and hold WNS and TNS unchanged in every scenario.
• Available alternative cells and dont_use marks.
• How many cells changed per group.

Healthy, Suspicious & Hard-stop Results

• Healthy (illustrative): Leakage drops 18% with setup and hold WNS unchanged in every scenario.
• Suspicious (illustrative): Recovery under 2%, which usually means the priority list is in the wrong order or update_power (PT) ran without activity annotated.
• Hard stop: Any new timing or DRC violation after recovery. The command should never create one, so a scenario was probably missing from the command focus.

What The Interviewer Is Testing

• Knowing the selection modes and that they are exclusive per run.
• Explaining why recovery runs first and a leakage-only swap runs last.

Follow-up Question & Model Response

"Why run power recovery before timing fixing at all?"

Candidate Model Response: Early recovery downsizes cells on paths with plenty of setup slack and removes buffers where hold slack is spare. That frees area and placement room that the later DRC and timing fixes can use. It never creates violations, so it does not make the later steps harder. The final leakage pass is different: it is a Vt swap with no layout change, so it can run after timing closes without disturbing placement.