chip-8 VS BunnyLOD

Source: https://github.com/gargakshit/chip-8 Libraries: GLFW with OpenGL, Dear ImGui

If you reuse the same index buffer as proposed then you won't be able to share vertices, which can be an important optimization. In my case I use a 323 cache of vert indices and just insert an index rather than a vertex if the vertex already exists and has the same material. I generate normals in the pixel shader to increase what can be shared, and then use a variant of Stan Melax's edge collapse polygon reduction algorithm to combine edges and reduce triangle count.

You might want to take a look at Stan Melax's easy mesh simplification technique with edge collapse at http://www.melax.com/polychop (read the full article link) - I've ported the code so it runs cross platform on https://github.com/dougbinks/BunnyLOD

I use a variant of progressive meshing, but constrained to preserve edges, normals and ambient occulusion in Avoyd based on Stan Melax's work: https://github.com/dougbinks/BunnyLOD

My triangle reduction in Avoyd, loosely based on Stan Melax's progressive mesh approach checks that the vertex being removed is internal to a set of triangles by checking that the internal angles add up to roughly 2Pi. This also ensures only flat surfaces are reduced, although I also perform normal checks and that vertex ambient occlusion gradients are preserved.