Skip to content

Pref Regressions for "Use NonNull in UnsafeWorldCell (#25632)" #25839

Description

@hymm

Bevy version

Regressions in Bevy Metrics

https://metrics.bevy.org/benchmarks_z.html#1559ad0ea5bf36ed569d853f07a59fca1b8dd8bd

Sees regressions:

  • query_get_many_*
  • get_entity_mut_slice_*
  • ecs_resources_insert_remove
  • despawn_world_recursive*
  • event_propagation_four_event_types

Sees improvements:

  • world_entity_50000_entities
  • observe_observer_custom*

I manually checked some of these as the metrics change encompasses multiple prs.

  • query_get_many: I actually see an improvement of 20%. Might be a difference between windows and linux.
  • get_entity_mut_slice: I can confirm the 20% regression here.
  • observe_observer_custom: I can confirm a 5% improvement.

I'm not 100% sure if we should revert or not, since there's both improments and regressions. We should probably look at the assembly and see if we can understand what is going on better.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    A-ECSEntities, components, systems, and eventsC-BugAn unexpected or incorrect behaviorC-PerformanceA change motivated by improving speed, memory usage or compile timesS-Needs-InvestigationThis issue requires detective work to figure out what's going wrong

    Type

    No type

    Projects

    • Status
      Needs SME Triage

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions