What is an ECO in physical design?

Short Answer

An ECO, or engineering change order, is a small, controlled change to a design that is already placed and routed, made instead of running the flow again. You edit the netlist, place only the changed cells, reroute only the affected nets and re-check. Timing ECOs fix timing or design rule violations, while functional ECOs change the logic itself.

Technical Explanation

• The block keeps its placement, clock tree and routing. Only the edited cells and the nets that touch them change, which protects the timing, IR and routing work already done in the rest of the block.
• In ICC2 the generic flow for timing or functional changes is to update the netlist with eco_netlist (ICC2) or with netlist editing commands, place the changed cells with place_eco_cells (ICC2), fill the gaps with filler cells, then reconnect the routing with route_eco (ICC2).
• Many timing ECOs start in PrimeTime. fix_eco_drc (PT), fix_eco_timing (PT) and fix_eco_power (PT) edit the design inside PT, and write_changes (PT) writes those edits as a script that ICC2 runs.
• There are two families. A timing ECO sizes cells, swaps threshold voltage or adds buffers and must not change function. A functional ECO changes logic, for example an RTL bug fix, and must match the new RTL.
• Before tapeout you can add cells and move existing ones; ICC2 calls this the unconstrained, or generic, ECO flow. After the base layers are fixed, the freeze silicon flow allows only metal and via changes, so every new gate must map onto a spare cell already on the die.
• Every ECO ends with the same checks: extraction, PrimeTime timing, logic equivalence and physical verification of the changed area. An ECO that fixes one thing and breaks another is common, so the re-check is part of the change.
• ECOs stop paying off when the change list grows very large. Moving many cells and rerouting large parts of the block then costs more, and is harder to verify, than re-running optimization from an earlier stage.
• ICC2 keeps track of what changed. Each edited cell carries an ECO change status such as create_cell, size_cell or add_buffer, and place_eco_cells -eco_changed_cells (ICC2) uses it to place only those cells, leaving every other cell where it was.

Follow-up Question & Model Response

"Why not re-run postroute optimization instead of doing an ECO?"

Candidate Model Response: Re-running optimization touches much more of the block than the problem needs, so clean paths can move and new violations can appear. It also takes far longer and restarts the correlation work with PrimeTime. An ECO keeps the proven parts frozen and changes a few dozen cells, which is easy to review, easy to check for equivalence and quick to re-time. Late in a schedule that predictability matters more than the extra slack a full re-run might find. A full re-run is still the right call when the violations are spread across the whole block, because then an ECO would touch nearly everything anyway.