Examples - Update Compute Rain to use computeIndirect

[cmhhelgeson]

Related issue: #31488

Description

Updates webgpu_compute_particles_rain to use indirect compute functionality. Instead of always dispatching the maximum number of particles, we only dispatch enough workgroups to handle the number of rendered particles.

Currently, I'm updating the indirect buffer with a uniform, but let me know if there's a better method.

[github-actions]

📦 Bundle size

Full ESM build, minified and gzipped.

	Before	After	Diff
WebGL	355.14 86.33	355.14 86.33	+0 B +0 B
WebGPU	609.53 170.95	609.53 170.95	+0 B +0 B
WebGPU Nodes	608.14 170.68	608.14 170.68	+0 B +0 B

🌳 Bundle size after tree-shaking

Minimal build including a renderer, camera, empty scene, and dependencies.

	Before	After	Diff
WebGL	486.76 121.07	486.76 121.07	+0 B +0 B
WebGPU	678.47 186.28	678.47 186.28	+0 B +0 B
WebGPU Nodes	620.46 169.54	620.46 169.54	+0 B +0 B

[Mugen87]

This example used to work in Firefox since it worked with WebGL as well. This is not true anymore, the example now badly flickers.

I would prefer to keep some of the compute examples compatible for WebGL as well. A lot of them are already WebGPU exclusive.

[cmhhelgeson]

This example used to work in Firefox since it worked with WebGL as well. This is not true anymore, the example now badly flickers.

I would prefer to keep some of the compute examples compatible for WebGL as well. A lot of them are already WebGPU exclusive.

Resolved by putting the count variable back and explaining why count still needs to be provided even when indirect compute is used.

[cmhhelgeson]

Perhaps the WebGLBackend code should be modified to treat the compute call as if dispatchSize was set to null when an IndirectStorageBufferAttribute is passed in so it will look to the compuetNode.count instead rather than having to switch based on the backend in the front-facing code.

Nevermind the warning already accounts for this behavior. Perhaps the warning can be modified to indicate to the user that the code will default to the computeNode.count value.

[Mugen87]

IMO, it is not required to include IndirectStorageBufferAttribute since the concept of workgroups or grids is advanced and maybe too complicated for the average user. I would prefer if the majority of our compute samples (at least the ones that work with WebGL and WebGPU) are implement in the simplest possible way in order to keep the demos approachable. Otherwise I fear they get too complicated.

Edit: I'm okay with #32176 (since it's already WebGPU only) but I would leave the rain example as it is.

[cmhhelgeson]

IMO, it is not required to include IndirectStorageBufferAttribute since the concept of workgroups or grids is advanced and maybe too complicated for the average user. I would prefer if the majority of our compute samples (at least the ones that work with WebGL and WebGPU) are implement in the simplest possible way in order to keep the demos approachable. Otherwise I fear they get too complicated.

Edit: I'm okay with #32176 (since it's already WebGPU only) but I would leave the rain example as it is.

Was just about to ask about fluid, so that's fine by me. Will close this PR.