Search Issue Tracker
AssetBundle allocates more memory when loading it a second time
How to reproduce:
1. Open the attached Project "AssetBundlesLeakRepro"
2. In the Menu Bar, select "Build" -> "CompileBundle"
3. Build and Run the Project for iOS with Development Build checked
4. On a macOS device, run Instruments -> Allocations and debug the running Application on the mobile device
5. In the Build, tap the "Load bundle" -> "Unload bundle" -> "Load bundle" -> "Unload bundle" buttons with some time for the memory usage to appear in Instruments
6. Observe the Memory usage in Instruments
Expected result: Loading the AssetBundle a second time does not make the memory usage increase
Actual result: Loading the AssetBundle a second time uses more memory than the first time and the memory after unloading a bundle does not go down to the initial value
Reproducible with: 2020.3.45f1, 2021.3.18f1, 2022.2.6f1, 2023.1.0b3, 2023.2.0a1
- Android (user-supplied information)
- MacOS Standalone (user-supplied information)
- Windows Standalone (user-supplied information)
- The project is currently set up to build bundles for iOS
- According to the user, iOS shows the largest increase in Allocations and as such, it's easier to see the issue
All about bugs
View bugs we have successfully reproduced, and vote for the bugs you want to see fixed most urgently.
- Texture Importer: Swaps red and blue channels
- Parallax mapping with UVs set other than UV0 plugged into sample texture 2d produces an error
- Objects with single letter name won't show up in hierarchy search results
- The GameObject is not transparent when setting CameraDepthTexture as a depth render target
- SRP Lens Flares with Occlusion enabled inconsistently shine through objects
Memory increases during the second load because of the expected behavior of an internal data structure (hashmap) and the huge number of objects loaded in the assetbundle. The first load reaches close to a capacity tipping point (45%) but then the second load happens to surpass that limit, hence the visibly larger allocation. The memory profiler snapshots are also part of the memory usage reported.