Add missing webgpu features

almarklein commented 9 months ago

The features specified by WebGPU (copied from https://gpuweb.github.io/gpuweb/#gpufeaturename):

enum GPUFeatureName {
    "depth-clip-control",
    "depth32float-stencil8",
    "texture-compression-bc",
    "texture-compression-etc2",
    "texture-compression-astc",
    "timestamp-query",
    "indirect-first-instance",
    "shader-f16",
    "rg11b10ufloat-renderable",
    "bgra8unorm-storage",
    "float32-filterable",
};

Features by wgpy-core: https://docs.rs/wgpu-types/0.18.0/wgpu_types/struct.Features.html

This PR adds one missing feature: "bgra8unorm-storage". (The "timestamp-query" was just moved so the order of the list is the same as in the spec.)

The final missing feature "float32-filterable" I cannot add because wgpu-core does not define it yet.

We are interested in having it, since we rely on it in pygfx because it's somewhat crucial in e.g. volume rendering. So I looked a bit into wgpu-core, and found the pr that adds bgra8unorm-storage. Seems like quite a bit of work, so for now I'm going to leave it at leaving the code for this feature commented here. We can keep using TEXTURE_ADAPTER_SPECIFIC_FORMAT_FEATURES (which includes float32-filterable) for now. 😅

almarklein commented 9 months ago

I wonder 🤔 if we can/should fake float32-filterable by enabling TEXTURE_ADAPTER_SPECIFIC_FORMAT_FEATURES instead?

cwfitzgerald commented 9 months ago

FYI bgra8 storage was really hard because it was the first to need to use a special spirv tag for storage textures.

Flost32 filterable should just be some logic checks

almarklein commented 9 months ago

Flost32 filterable should just be some logic checks

Ok, you nerd-sniped me now.

almarklein commented 9 months ago

https://github.com/gfx-rs/wgpu/pull/4759

gfx-rs / wgpu-native

Add missing webgpu features #322