How do you debug a congestion hotspot after global routing?

Short Answer

First find it and measure it: which GCells, which layers, which direction and by how much. Then work out why: too many pins in a small area, a narrow channel between macros, too much wrong-way demand, or tracks lost to blockages or PG. The fix depends on the cause, so the diagnosis is most of the work.

Technical Explanation

• Start with the congestion map and report_congestion to locate the hot spot and see whether it is horizontal, vertical or both. report_congestion -rerun_global_router refreshes the map with a map-only global route, and -boundary and -layers limit the report to the hot spot.
• Pin density hot spots sit over clusters of small, pin-heavy cells. Demand is local and high on the lower layers. The fix is spreading cells with padding or lower local density.
• Macro channel hot spots sit between macros, where many nets must pass through a narrow gap. The fix is widening the channel, reorienting macros so pins face the channel less, or letting routes pass over the macros.
• Wrong-way demand shows as overflow in one direction on layers that mostly run the other way. Routing guides that limit non-preferred edges, or layer rebalancing, help here.
• Resource loss shows as low capacity rather than high demand: GCells in the "Blocked" bin, or capacity cut by PG straps and routing blockages. The fix is in the PG plan or blockage definitions.
• A track utilization routing guide, for example -horizontal_track_utilization 50, limits how much of an area the router fills, which spreads routes around a sensitive region. It is a soft constraint and does not change global routing capacity, so its effect shows up in detail routing, not in the congestion map.
• After any fix, rebuild the map with route_global -congestion_map_only true and compare. Congestion fixes often move the problem next door, so look at the neighbouring GCells too.

What To Check

• Location, direction and layers of the overflowing GCells.
• Demand against capacity on the worst edges, for example "18/9".
• Cell and pin density under the hot spot.
• Blockages, PG straps and macro edges nearby that remove capacity.
• The same area after the fix, plus its neighbours.

Healthy, Suspicious & Hard-stop Results

• Healthy (illustrative): The hot spot clears to zero overflow and the neighbouring GCells stay below capacity.
• Suspicious (illustrative): Overflow moves from one channel to the next one over, which means demand was pushed, not reduced.
• Hard stop: GCells with zero capacity where signal pins sit, or overflow driven by a macro placement that routing guides cannot solve.

What The Interviewer Is Testing

• Whether you diagnose the cause before choosing a fix.
• Can you tell demand problems from capacity problems?
• A strong answer shows you re-check neighbours after the fix.

Follow-up Question & Model Response

"How can you tell a pin density hot spot from a channel hot spot on the map?"

Candidate Model Response: A pin density hot spot is usually a compact blob over standard cells, strongest on the lower layers, and its overflow drops quickly as you move away. A channel hot spot follows the shape of the gap between macros, often a long thin band, and affects the layers the through-nets use, often middle layers. Overlaying cell density and macro outlines on the congestion map usually makes it obvious in a minute.