WebGPU & NodeMaterial: ComputeNode

[sunag]

Description

We are very close to having a robust particle system native of Three.js made in GPGPU.
WebGPU compute feature is amazing, the only problem so far was the 16-bit limitation.

WebGPU & NodeMaterial

• add ComputeNode and StorageBufferNode
• update webgpu_compute example

ShaderNode

• add compute( dispatchCount ) ⇒ ( dispatchCount = particlesCount )
• add timer( timeScale )
• add storage( bufferAttribute )
• add func( code ) ⇒ code is a GLSL or WGSL function code

Creating and calling a WGSL function

const funcNode = func( `( someParam:vec2<f32> ) -> vec2<f32> {

	return someValue;

}` );

const funcCallNode = funcNode.call( { someParam: vec2( 1.0, 1.0 ) } );

See more in this example code

This contribution is funded by Google via Igalia

[LeviPesin]

That's amazing!!!

I think that after this PR will be merged and the PixelNode would be added, I would make a PR to add to nodes/gpgpu WebGL and WebGPUComputationRenderers (draft versions here: https://github.com/LeviPesin/THREE.js-PathTracing-Renderer/tree/webgpu/js/pathtracing/utils, I made them as a part of my still-going porting of the @erichlof's Pathtracing Renderer to the Nodes system).

[LeviPesin]

Also, this PR resolves #23598 and almost resolves #23599.

[LeviPesin]

I think it is better not to hardcode instanceIndex here, but rather use global_invocation_id and similar things (I think something like InvocationIDNode should be created for them).

[sunag]

I like that we have a single node for use common of "call index", but surely we can have both.

[LeviPesin]

I think this can be an argument in the constructor?