If the server is also a client, I can use compute shaders, but if it's a dedicated server, I'll need to maintain a CPU version of my generation algorithm to do server-side checks of client actions 😬
In any case, I'm glad you don't notice any apparent delay.

Thanks for your answer. I asked this question because I'm hesitant to transfer all my voxel generation and meshing to the GPU. But this has many drawbacks: complex pipelines compared to the CPU version, the physics needs readbacks, and I have to make it compatible with networking.

Do you notice any latency due to these readbacks?

I think @acerola.gg has something to say about this :D

It was fun to get it working properly 😅

Wow, 1 draw call is impressive. Good job!

Awesome! I assume you're using 6 quadtrees? Do you use big_space to handle this scale or is it custom made?

This is the tool he uses: github.com/noio/Pixel-U...