Parse shaders with half-precision floats (float16)

[lyricwulf]

When using 16bit float storage and arithmetic extensions:

#extension GL_EXT_shader_16bit_storage : enable
#extension GL_EXT_shader_explicit_arithmetic_types_float16 : enable

Shader compiles to spv without issue, but trying to create a wgpu shader module:

[WARN wgpu_core::device] Failed to parse shader SPIR-V code for Some("shader.comp.spv"): UnsupportedCapability(Float16)

[kvark]

If you want to use them from wgpu, you'd also need a feature inside wgpu to be exposed/enabled first. WebGPU group has it on the radar for the nearest future, but it's not there yet. See gpuweb/gpuweb#658

[TimothyOlsson]

If you want to use them from wgpu, you'd also need a feature inside wgpu to be exposed/enabled first. WebGPU group has it on the radar for the nearest future, but it's not there yet. See gpuweb/gpuweb#658

Two relatively new PRs have added f16 for WebGPU

gpuweb/gpuweb#2696

gpuweb/gpuweb#2647

[teoxoy]

Support for f16 for wgpu has been merged in gfx-rs/wgpu#2646.
Let's track f16 support for WGSL in gfx-rs/wgpu#4384.