So anyway I was slightly sniped by @eniko's block game so now we have blocks too!
-
So anyway I was slightly sniped by @eniko's block game so now we have blocks too! blocks for everyone! My guess here is to have world space divided into these 32x32x32 blocks, and then put those in vram and have a compute shader turn them into mesh shader draws. this is one mesh shader dispatch, with 2 bytes per occupied voxel (5 bits per coordinate), and then the whole dispatch has the per-chunk position / orientation / etc. No culling at the moment so it's 12 tris per voxel, but it should be 6
-
So anyway I was slightly sniped by @eniko's block game so now we have blocks too! blocks for everyone! My guess here is to have world space divided into these 32x32x32 blocks, and then put those in vram and have a compute shader turn them into mesh shader draws. this is one mesh shader dispatch, with 2 bytes per occupied voxel (5 bits per coordinate), and then the whole dispatch has the per-chunk position / orientation / etc. No culling at the moment so it's 12 tris per voxel, but it should be 6
fwiw the stuttering is just OBS struggling to capture at 2880x1800, it's smooth in practice on my laptop here, though I haven't started optimizing it at all.
-
fwiw the stuttering is just OBS struggling to capture at 2880x1800, it's smooth in practice on my laptop here, though I haven't started optimizing it at all.
Plan is to implement roughly the approach in https://research.nvidia.com/labs/rtr/publication/majercik2018raybox/ but I figured I should start with a baseline that's just rendering every damn block.
-
Plan is to implement roughly the approach in https://research.nvidia.com/labs/rtr/publication/majercik2018raybox/ but I figured I should start with a baseline that's just rendering every damn block.
Can't say for certain but i think this shader might be slightly export bound...
-
undefined oblomov@sociale.network shared this topic on