BeginnerQuestion 90 of 95Source: Synopsys PrimeTime User Guide: ECO Flow

What is an ECO, and when do you fix timing in PrimeTime instead of redoing place-and-route?

From PDVerse STA Mentor Guide · pdVerse Mentor Guide

By Tabish Iqbal · 9+ Years VLSI Physical Design Experience

Short Answer

An ECO — engineering change order — is a small, targeted set of netlist changes made late in the flow to fix specific violations, instead of rerunning the full place-and-route flow. A designer reaches for PrimeTime's ECO commands when only a handful of cells need resizing or a few buffers need inserting, because a full re-implementation would cost far more schedule time than the violations are worth.

A before/after diagram: a placed block with six flagged hold-violation flops, then the same block with three small buffers inserted only near those flops, everything else unchanged. — Technical Reference DiagramWhat is an ECO, and when do you fix timing in PrimeTime instead of redoing place-and-route?

Technical Explanation

Late in a project, the cost of a full re-run and the cost of a small surgical fix diverge sharply, and ECO exists for that gap.

fix_eco_timing (PT) fixes setup or hold violations by resizing cells or inserting buffers and inverter pairs, targeting only the specific violations you point it at.
The output is a list of changes, called the ECO itself, that gets written out with write_changes (PT) and applied back into the physical design rather than left only inside PrimeTime.
Redoing full place-and-route reconsiders every cell's placement and every net's route from scratch, which can shift thousands of other paths' timing along with the few you meant to fix.
An ECO instead touches a small, bounded set of cells and nets, keeping everything else in the design exactly as it was, physically as well as logically.
This bounded scope is why ECO is the standard fix late in a schedule, close to tapeout, when there is no time left to absorb a full re-implementation's ripple effects.
Early in a project, before placement is even final, a designer would normally fix a timing problem by adjusting constraints or logic and re-running the full flow instead, since nothing has been committed yet.

Common Mistake

The Trap: treating ECO as a shortcut available at any project stage.

Running an ECO on a design that has not even completed clock tree synthesis wastes the exercise, since a normal re-run would still change too much for the ECO's small fix to hold.
ECO earns its value specifically late, when the alternative — a full re-run — is the more expensive option.

Follow-up Question & Model Response

Once fix_eco_timing has picked a set of cell resizes and buffer insertions, does that fix take effect immediately in the physical layout?

Candidate Model Response: Not by itself. fix_eco_timing (PT) changes PrimeTime's internal netlist model and produces a change list; the physical tool still has to place, route, and legalize those added cells. write_changes (PT) hands that change list over in a format the place-and-route tool can consume. Only after the physical tool applies the changes and the design is re-extracted does the fix exist in silicon, which is why every ECO gets re-verified with a fresh timing run afterward.

Practical Example

Two weeks before tapeout, a block shows six hold violations, each under 30ps, scattered across one clock domain. Instead of re-running place-and-route on the whole block, the team runs fix_eco_timing -type hold (PT) targeted at those six endpoints, inserts small delay buffers, and re-verifies with a fresh signoff run rather than re-placing the entire block.

Complete STA Handbook