The mesh shader in this scenario is the bottleneck as it cannot feed the rasterizer fast enough. This is expected as it is running the SN algorithm on a per-frame basis. This could be very interesting when coupled with some sort of fluid simulation, as this can allow you to convert fluid simulation particles (in the case of SPH) directly to a mesh that the rasterizer can render.
It is also worth noting that in the case of static geometry or geometry that does not update on a per-frame basis, you could run the SN algorithm prior to this and cache the results (save meshlet vertex/triangle in a buffer), and render those using Mesh Shaders / Task Shaders for culling.
The advantage of running the algorithm in the mesh shader like this is that LODs & resolution selection become trivial, as you can scale the meshlets and the chunks however you like without having to calculate octrees or quad-trees (but of course in my implementation I did not implement anything like skirts to hide the gaps between chunks of multiple LOD levels).
Nesta página do site você pode assistir ao vídeo on-line Surface Nets experiments using Task / Mesh shaders duração hora minuto segundo em boa qualidade , que foi baixado pelo usuário Jed037 01 Junho 2026, compartilhe o link com seus amigos e conhecidos, no youtube este vídeo já foi visto 214 vezes e gostou 9 espectadores. Boa visualização!