NightlyTestManual May 25 Run Failures

[AmirhosseinPoolad]

Latest automated weekly test failed. Documenting the issues below:

• vtr_reg_nightly_test1:
  • One failure
  • Small circuit (mult119.v) with <100 blocks
  • Minimum channel width changed to 42 (Golden result = 24)
  • Everything else stayed more or less the same
  • Same number of clustered blocks
  • Placement went for way longer and never entered the second stage
• vtr_reg_nightly_test3:
  • One failure
  • Small circuit (sha.v) with ~200 blocks
  • Minimum channel width changed to 68 (Golden result = 52)
  • CPD actually improved by a negligible amount (10.757ns vs 10.9834ns)
• vtr_reg_nightly_test4:
  • One failure
  • Medium/large circuit (titan minres) with ~14k blocks
  • TWL actually improved by a negligible amount (1633788 vs 1655917)
  • CPD changed from 4.851ns to 6.367ns
    • Circuit has 4 clock domains, other 3 were either better or not noticably worse

The failures started from bc4a791.

[AlexandreSinger]

Yeah, it looks like your recent changes to the seed selector worked fine on its own, but when it merged with Master it interacted with another PR to cause this change. This can happen. Who will be on the hook for regenerating these golden results. It really shouldn't take long to regenerate these.

[AlexandreSinger]

FYI, the NightlyTests failed again yesterday: https://github.com/verilog-to-routing/vtr-verilog-to-routing/actions/runs/15371421194

It looks to be the same QoR failures as before.

[AmirhosseinPoolad]

Fixed by #3115