I’m curious whether Nanite uses multi-draw indirect, or if multi-draw indirect is used in D3D11. Currently, with custom vertex factories and scene proxy, how should they be used? My idea is to omit the process of automatically merging draw calls and traversing BatchElement in multi-mesh instanced rendering, combining GPU virtual addressing and bindless techniques to index all virtual textures and virtual mesh resources, achieving true shader-level sharing and on-demand loading. This should provide some performance improvement. It would be even better if there are examples in the source code.