hayba-performance-budget

Measure a level budget

When to use

Use for a reported slowdown or an explicit target platform budget. Record resolution, quality settings, map, camera path, and PIE/editor state. Without a stated target, establish a baseline and avoid inventing a pass threshold.

Tool routing

• Read editor_get_performance_stats under repeatable conditions. Use scene_export to locate concentrations of actors and wp_get_cells for streaming distributions. editor_stream_log can expose relevant load or render warnings.
• Choose a hypothesis that matches the measured bottleneck. Instance consolidation helps repeated mesh draw calls but can harm culling granularity; reducing foliage density may help GPU or overdraw but changes art direction. Material and texture work should be routed to their focused tools after identifying the asset.
• Change one factor, then measure again from the same camera and settings. Use validator_run after scene edits and editor_capture_viewport to catch visual regression.
• If the accepted optimization changed the map, save its exact active path with level_save, inspect saved, dirty, and verified status, and read back the changed actors or foliage before reporting the measured improvement. If persistence is uncertain, mark the optimization unsaved and route it to hayba-world-save-integrity.

Evidence

Provide baseline, target, changed factor, after values, and visual/validator findings. Compare like for like. Stop if the tool reports only aggregate metrics that cannot identify the suspected cause; request a narrower profiling run instead of claiming an optimization.