What breaks when an ECO adds cells across voltage areas, and how is it repaired?

Short Answer

An ECO buffer on a net that crosses domains can land in a voltage area whose power domain does not match, get the wrong supply, sit on the wrong side of an isolation cell, or use a single-rail cell where a dual-rail cell is needed. Repair it with fix_mv_design -buffer (ICC2) for buffer trees and fix_mv_design -diode (ICC2) for diodes, and connect the supplies of new ECO cells with eco_update_supply_net (ICC2). Then check again, since the fixes are not guaranteed placement or timing legal.

Technical Explanation

• What breaks: an ECO buffer's power domain may not match the voltage area it sits in, it may be a single-rail cell where the net needs a dual-rail one, it can create an isolation violation in a buffer tree, or its supply pins may never be connected.
• eco_update_supply_net (ICC2) connects supply nets for cells added by size_cell (ICC2) in freeze-silicon mode, add_buffer (ICC2), add_eco_repeater (ICC2), split_fanout (ICC2) and add_buffer_on_route (ICC2). -cells restricts it to the named cells and the ECO buffer trees they belong to.
• fix_mv_design -buffer (ICC2) fixes isolation violations and illegal library cells in buffer trees. It can swap single-rail and dual-rail cells, change backup supplies, change level-shifter input or output supplies (only with -level_shifter, since buffer trees stop at level shifters by default), and fix a domain and voltage area mismatch with a physical or logical feedthrough.
• fix_mv_design -diode (ICC2) fixes power domain and voltage area mismatches for diodes by assigning a matching power domain. Only -report_only and -verbose go with it.
• Limits: fix_mv_design (ICC2) does not insert isolation or level-shifter cells, and its changes do not create new isolation or voltage violations. It commits the UPF first if needed, keeps original placement unless -move_within_x or -move_within_y is given, and uses only supplies available in the voltage area. When it swaps a library cell it removes that cell's PG connections, so reconnect them with connect_pg_net (ICC2).
• Its changes are not guaranteed placement or timing legal, so check legality and timing after it. -report_only shows proposed changes without making them.
• Prevent it at the source: give PrimeTime the voltage areas through physical data so fixes land in the right area, and keep buffer lists to cells legal in each domain.

What To Check

• check_mv_design (ICC2) after the ECO is applied.
• Supply pin connections on every new ECO cell.
• Domain against voltage area for every ECO buffer and diode.
• Legality and timing after fix_mv_design (ICC2).

Healthy, Suspicious & Hard-stop Results

• Healthy (illustrative): check_mv_design (ICC2) clean after the ECO, with every new cell's supply pins connected.
• Suspicious (illustrative): A few ECO buffers flagged for domain mismatch, all fixable by a cell swap.
• Hard stop: Isolation violations the command cannot fix, because it does not insert isolation cells; the netlist or UPF must change.

What The Interviewer Is Testing

• Can name the multivoltage failures an ECO buffer causes.
• Knows what the -buffer and -diode modes of fix_mv_design (ICC2) fix and what the command will not do.
• Checks placement and timing after the repair.

Follow-up Question & Model Response

"Why can fix_mv_design fix a mismatch with a feedthrough instead of moving the cell?"

Candidate Model Response: By default the command keeps the original placement of buffers and inverters. A physical or logical feedthrough lets the buffer belong to a power domain that matches the voltage area it sits in, which fixes the mismatch without disturbing placement. Moving is possible only if you ask for it with -move_within_x or -move_within_y, which shift a buffer into a nearby voltage area so it can stay single-rail, and the target voltage area must allow feedthroughs. If the result is still wrong for timing, re-plan the ECO and run the check again.