Pavel Rojtberg

which asserts do you run into with 16bit indices? you are likely using the ImDrawList API to get there? Can you show some code that triggers the assert? also, would...

> I think just adding a subfolder to the framework would not be as much of a breaking change. sounds good. Feel free to create a PR

focused shadow map has no snapping to texels. Needs to be done first to avoid "swimming" artifacts. https://docs.microsoft.com/en-us/windows/desktop/dxtecharts/common-techniques-to-improve-shadow-depth-maps#techniques-to-improve-shadow-maps

is assimp built by ogre or externally? is the used assimp library built as debug?

sounds like #1303. can you double-check that assimp is built as debug?

yes, Codec_Assimp will be a debug librarary. What about Assimp itself? Where does assimp_DIR in Cmake point to? Does it contain the debug library of assimp? If all above is...

no idea though.. might be MinGW specific. You can try master, where we updated to assimp 5.3.1 to check whether it is caused by assimp

the remaining bottleneck is likely in findVisibleNodes - more profiling needed.

| Ogre | Configuration | Static (ms) | |--------|--------------|-------------| | 1.11.5dev | GL | 39.42 | | 1.11.5dev | GL + #985 | 36.00 | AMD Ryzen 7 2700X Eight-Core...

the issue is that we only consider the caster bbox for focusing the shadow camera:  we could also add the receiver bbox, but that would drastically decrease shadow resolution....