Hi, we are rendering large terrain tiles at various sizes and resolution. The approx size of the tile’s extent in the picture is 276,992 x 243,967 x 234,252 where the tile mesh is slanted. The picture is the intersection of 4 adjacent tiles. We are importing these tiles from an external software that exports it into fbx, which then gets imported to UE with our custom pipeline. To our knowledge, the vertices are exported as doubles. In FbxMesh.cpp it seems that the original data is being casted from double to floats when filling the FMeshDescription. We are suspecting this is the main reason of the discrepancy we see in the picture, where the vertices are supposed to be exactly the same in world space - but it is not reflected visually. We’ve tried editing the vertex in the Modeling editor mode to the exact location but it seems to automatically be casted to floats and ends up reflecting a different value than what we put in. We are wondering if this is a current UE limitation, if a solution is being worked on to be released in the near future, and what steps are recommended for us to take.
If this is truly a limitation, that the vertices of UStaticMeshes can only be in floats, then we are wondering if a custom vertex factory that splits the source vertices (position,normal,tangent) into 2 vertex buffers and combining them at the beginning of the vertex shader (using LWC) - is a viable workaround. And what are the consequences of that method when using Nanite?