performance boost

[dcg1985362]

Hi I am very interested in your engine. do you use GPU instancing draw and batch the d3d12 barriers for better performance?

[adepke]

This is a GPU-driven mesh renderer, where all indirect draw commands are generated by the mesh culling pass. I don't currently support instancing, as that would introduce a large amount of complexity here and I'm more interested in volumetric rendering instead of massive amounts of duplicated meshes right now. What I have currently is plenty capable of many thousands of meshes, see: https://www.youtube.com/watch?v=WQRD-Eds0CU

Reference:

VanguardEngine/VanguardEngine/Source/Rendering/Renderer.cpp

Line 394 in a693287

for (const auto& renderable : renderables)

For barriers - yes they are batched a bit, there's a built-in mechanism where render passes request barriers (explicitly or implicitly via pass read/write dependencies), and these barriers are only then actually emitted into command lists once a CommandList::FlushBarriers() occurs.