robtfm

> Do you think that if mesh uses baseColorUV instead of normalUV for normal texture, the mesh will look correct? no, but i think errors in occlusion map uvs or...

one issue with my suggestion is that gltfs might use multiple identical sets of texcoords, in which case it would currently be working but would lose features with this strategy....

> HashSet doesn't keep single value, it keeps all of them. (just wanted to point out that while the hashset has all of them, only one ATTRIBUTE_UV_0 can exist on...

note this also fixes the issues from #11645 but is maybe more contentious

modified to sort shadows by mesh id, since we only batch entities with the same mesh (we still sort by pipeline first)

> I wonder if sorting the opaque pass by mesh id would net us a similar decrease? I guess that would depend on frag shader cost (more overdraw / less...

```rust /// Describes the horizontal alignment of multiple lines of text relative to each other. /// This only affects the internal positioning of the lines of text within a text...

yes this was done to conserve bindgroup entries. pcf shadows will help, but if the current directional quality is acceptable then when pcf lands we may want to reduce that...

a different radius in texel units would be good for Jimenez14. i think we can't do that with the optimisations for Castano13 since it relies on the texel positions for...

[8e3c28b ](https://github.com/bevyengine/bevy/pull/6284/commits/8e3c28b1ca4b62d5a760c1c57442e58914f5a02b) has broken viewports. i tried adding a viewport to the shader_prepass example and got a variety of weird effects. reverted this commit and it all works fine.