Search Issue Tracker
By Design
Votes
0
Found in
2019.3.0a1
2019.3.0f2
Issue ID
1203358
Regression
Yes
[Performance Regression] Prefab PerformanceTest: MeasureSaveTimeForPrefab(20,1,1,1000) is significantly slower
The PrefabSaveTests test: Test: MeasureSaveTimeForPrefab(20,1,1,1000) is significantly slower in 2019.3 than 2019.2 or 18.4
It looks like the regression was introduced with the change-set: 2019.3.0a1 (25828b1b312)
There is a second, even worse regression introduced in 2020.1.a5 (1e4819d07b04)
Platform observed on - MacOS
To observe the issue:
https://observer.cds.internal.unity3d.com/afb89201c7d5c267b5743e5f7de3f8c2fd091ba0
Add comment
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
Latest issues
- [Android] PointerEventData.pointerId value is incorrect when pressing a finger on the screen
- Standard Unity Materials and Shaders become corrupted after importing specific Asset Packages
- [Linux][OpenGL][Vulkan] Draw calls are not shown in the Event List when taking a capture of a frame with RenderDoc
- Inaccurate collision detections when Rigidbody Collision Detection is set to "Continuous" or "Continuous Dynamic"
- Crash on Object::IncrementPersistentDirtyIndex when upgrading project version
Resolution Note (2019.3.X):
The regression with performance is due to the performance tests are leaking objects. If you run the performance test just one, v1 and v2 performance is similar. Asset db v1 runs a garbage collection during import and with v2 the garbage collection has been removed. In v2 we want the import to leave the loaded objects state the same before and after imports. By added a single clean up of objects (EditorUtility.UnloadUnusedAssetsImmediate();) in the test, the performance test runs much fast with v2.