VReaperV

The original quake 3 code does the same thing: https://github.com/id-Software/Quake-III-Arena/blob/dbe4ddb10315479fc00086f08e25d968b4b43c49/code/renderer/tr_backend.c#L397. Although the result might be different for one reason or another.

If we really want to compress thousands of triangles for UI we can make each one use a lot less data since you don't need the full fp32 precision for...

Yeah, merging the code between material/non-material would be great. I've thought about merging the `UpdateSurfaceData*` functions with `Render_*` in `tr_shade`, though it would be a bit cumbersome with a bunch...

If we're being pedantic, it's `workaround.driver.*` since the issue is not within the API spec, but with the specific implementation.

Inconsistent number of surfaces might be because `R_AddWorldSurfaces()` is not adding a few surfaces for some reason... I thought I fixed all causes of that, but I might've missed some....

Thanks, I'll take a look at it later.

So far I'm unable to reproduce either of these bugs and I get consistent 989 surfaces.

> I'm using all default cvars except `r_materialSystem` and `r_arb_bindless_texture`. I tried to test with Intel to see if it's driver-specific, but that [didn't work out](https://github.com/DaemonEngine/Daemon/issues/1260). Anyway dumping the GPU...

After asking to test with `r_gpuOcclusionCulling on` on `forlorn` for a different issue with a lower resolution, it was discovered that some surfaces were also missing following that. So most...

There've been some fixes for missing surfaces that have fixed similar issues on other maps, so it's possible that this was fixed too.