Stuart Carnie

> > Most likely then the change to the shader has exposed a driver bug with your GPU > > Yep looks like it, Managed to reproduce it on Windows...

Latest push * Reimplements debug draw support to close #98005 as demonstrated in [this comment](https://github.com/godotengine/godot/issues/98005#issuecomment-2405925243) * Includes additional optimisations to remove the uniform cache, and uses the `LRUCache` to limit...

Thanks @Calinou – I'll look at your third test case now

@Calinou Curiously, I am seeing the opposite, that this PR improves over the official dev 4.4 build (prior to this PR). However, it does appear that #98212 is a significant...

@Calinou / @clayjohn regarding **Sprite rendering (1 unique texture per sprite)** test: https://github.com/godotengine/godot/pull/97340#pullrequestreview-2374632529 The problem is that the uniform set cache is being thrashed (I knew that would be a...

Whilst investigating, I've identified an issue that if the texture is changed: https://github.com/godotengine/godot/blob/e2c6daf7eff6e0b7e2e8d967e95a9ad56e948231/servers/rendering/renderer_rd/storage_rd/texture_storage.cpp#L581-L597 `TextureStorageRD` invalidate caches; however, it won't invalidate the cache in the 2D renderer. Another issue is that...

I can confirm the cache is invalidated when textures are freed, so we just need to address the _changed_ issue.

Another pathological case is that if the ordering of nodes changes regularly, and the number of textures is high, it will result in uniform set churn. Not sure if this...

@akien-mga I have not tested C#, but I suspect you are right, so disabling for now is probably a good idea. I can explore that next.

I had a look in `modules/mono`, and I saw `APPLE_EMBEDDED_ENABLED`, so perhaps it does Just Work™? The architectures are both arm64.