GS Volume Culling

[CedricGuillemet]

This is the 1st step for performance improvements with GS. AABB Volumes used are coming from Compressed .PLY quantized chunks. It's far for being efficient. But goal here is to have everything in place to try different algorithms while not increasing complexity too much at once.

Batch of 256 splats are culled. visible ones are sorted and that subset is used for instance count. It should decrease the pressure on GPU/CPU a little.

Next:

• Better partitioning (octree)
• wasm sorting
• radix sort?
• per splat culling
• per volume sorting of splats and volume sorting

EDIT: there should be 1 noticeable issue when camera moves fast with huge number of splats. Latency induced by sorting will make visible splat being not culled late and popping might be visible.

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s). To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[bjsplat]

Please make sure to label your PR with "bug", "new feature" or "breaking change" label(s). To prevent this PR from going to the changelog marked it with the "skip changelog" label.

[bjsplat]

Snapshot stored with reference name: refs/pull/15705/merge

Test environment: https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15705/merge/index.html

To test a playground add it to the URL, for example:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15705/merge/index.html#WGZLGJ#4600

Links to test babylon tools with this snapshot:

https://playground.babylonjs.com/?snapshot=refs/pull/15705/merge https://sandbox.babylonjs.com/?snapshot=refs/pull/15705/merge https://gui.babylonjs.com/?snapshot=refs/pull/15705/merge https://nme.babylonjs.com/?snapshot=refs/pull/15705/merge

To test the snapshot in the playground with a playground ID add it after the snapshot query string:

https://playground.babylonjs.com/?snapshot=refs/pull/15705/merge#BCU1XR#0

[bjsplat]

WebGL2 visualization test reporter:

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15705/merge/testResults/webgl2playwright/index.html

[bjsplat]

Visualization tests for WebGPU (Experimental)

https://snapshots-cvgtc2eugrd3cgfd.z01.azurefd.net/refs/pull/15705/merge/testResults/webgpuplaywright/index.html

[CedricGuillemet]

Switching back to draft, I'm not sure CPU culling on a webworker is a good way to improve perf for high splat count.

[CedricGuillemet]

Culling in webworker adds too much latency in visibility. Closing.