OpenROAD (The-OpenROAD-Project) errors

5 recurring problems engineers report in OpenROAD (The-OpenROAD-Project), ranked by impact.

#100 by errors trackedVerificationECAD

Recurring OpenROAD (The-OpenROAD-Project) errors, by impact

  1. 1MINLENGTH TWOSIDES end-of-line rule is checked one-sided because the wrong flag is passed.Verification · Agent failure · 1 report66
  2. 2On a chip process with backside power layers, the router counts layers two different ways in one function, so it can pull track spacing from a never-enabled layer and compute the wrong tile size with no error.ECAD · Agent failure · 1 report66
  3. 3Since a change merged Oct 24, detailed_route starts at iteration 3 when DRC markers exist, skipping iterations 0-2 that route nets with guides but no wires; in an ECO flow after repair_design or repair_antennas, ripped-up nets are never routed while the command "reports 0 violations and returns normally".ECAD · Agent failure · 1 report66
  4. 4Two LEF58 end-of-line DRC rule variants are enforced as a plain rule with no warning: the router never reads the TOCONCAVECORNER and TONOTCHLENGTH flags and checks them as an ordinary end-of-line rule with WITHIN 0. Other unmodelled options get a warning, but these two are neither modelled nor warned about.Verification · Agent failure · 1 report66
  5. 5When the via-swap fallback runs out of options it skips updateDRNet, so the checker keeps a stale copy of the net from the last via it tried and writes DRC markers for a via that is not in the design. In the last batch, 21 of 21 nets with a Metal Spacing marker went through that fallback.ECAD · Agent failure · 1 report66

How it fails

  1. Agent failure5

Engineering tools that address these errors

Sourced from public community reports. Costs and delays on individual entries are modeled estimates, not measurements.