Performance Shift(s): `8aa57d74`

[github-actions]

Benchmark comparison has identified performance shifts at

• commit 8aa57d7 (Retire use of the New: labels. #4801).

Please review the report below and take corrective/congratulatory action as appropriate 🙂

Performance shift report

       before           after         ratio
     [4fa3f06c]       [8aa57d74]
     <main~1>         <main>    
+        2.421875         2.921875     1.21  experimental.ugrid.regions_combine.CombineRegionsCreateCube.track_addedmem_create_combined_cube(500)

Generated by GHA run 2505625678

[trexfeathers]

Noise. I don't think we have any bright ideas on track_addedmem do we @pp-mo?

[pp-mo]

Noise. I don't think we have any bright ideas on track_addedmem do we @pp-mo?

The only idea left that I can think of is maybe to add an offset to the value, which would reduce sensitivity since the test is a ratio (like the times).
Unless we can control pass sensitivity specifically for these tests, somehow ?

I think we discussed all this before, and didn't rate any of our proposed solutions much. But I agree, it keeps happening so perhaps we should think again.

[pp-mo]

Unless we can control pass sensitivity

@trexfeathers I can't remember where that is controlled, can you remind me?

[trexfeathers]

Unless we can control pass sensitivity

@trexfeathers I can't remember where that is controlled, can you remind me?

airspeed-velocity/asv#1208

[pp-mo]

Ok, re-remembered + found it here

IMHO our frameworks are still over-complicated !
But at least benchmarks don't use Cirrus.
gha --> nox --> asv