feat: implement F16 support in shaders

[FL33TW00D]

Connections

• Hard dependency on WGSL: Add a skeleton for enable-extensions/enable …; that reports nice errors #6424.
• Resolves WGSL: Support the enable directive #5476.
• Resolves [meta] f16 support #4384.

Checklist for merge-ability

• Test multiple redundant directives (I think there may be a cleaner way to do this: https://github.com/gfx-rs/wgpu/pull/5701/files#diff-9ef835ffc5f7d7313243db5bfa371d3115b82c0705d4f1df75c98dfe07d927a1R2175)
• MSL out
• HLSL out
• SPV out
• Arbitrary for F16 on half-rs: feature: Arbitrary VoidStarKat/half-rs#111. Currently we are using a git dep on my fork which is causing CI to fail.
• hexf_parse f16 support: https://github.com/lifthrasiir/hexf. This one won't be fun. Did some work on it here: lifthrasiir/hexf@master...FL33TW00D:hexf:feature/f16 . Deferred to Implement Parsing of Hexadecimal F16 #7046
• We should make sure that we are properly validating As casts of matrices, because that code currently assumes that all matrices have the same kind (Naga's typechecker ignores scalar kind when casting matrices #6441).
• Validate out f16 inside push constants.
• Continue tests for f16 layout.
• wgsl-out needs to enable f16
• Fix spirv output of matrix casts
• Aliases for f16 types still don't seem to work
• Split out out-of-scope commits

Additional context

F16 is available in >=SM6.2: https://github.com/microsoft/DirectXShaderCompiler/wiki/16-Bit-Scalar-Types

[FL33TW00D]

I've marked this as ready for review, as the wgpu specific logic is implemented and would be great to start iterating on it.

2 main blockers are:

1. Merging of the PR on half-rs.
2. hexf-parse improved implementation.

[FL33TW00D]

Dogfooding works! Fixed a few bugs in 30e12b5. Still waiting on upstream.

[teoxoy]

One thing I'm concerned is that the polyfills we have in some of the backends might not support f16. It would be worth adding more tests covering the built-in functions and operations that support f16. Execution tests would be great as well, they live in tests/tests/shader.

[ErichDonGubler]

@raphamorim: Can you please post the Git history for half-2, presumably in a fork under your user? We noticed that the commit hash 367f64f7a23e8763d1662e7da6731d74422133b3 in half-2's .cargo_vcs_info.json is not accessible from the starkat99 or FL33TWOOD repos.

[raphamorim]

@ErichDonGubler I apologize, I haven't updated the fork on GH. Just ran a git push 🙏

Regarding that hash I confess I did the fork and published as a test to if it was something worth to maintain and use for wgpu, but in case we plan to use half fork then I just updated the tree to only contain @FL33TW00D changes (along with Cargo.toml)

Also please bump to 2.4.2 https://crates.io/crates/half-2

[cwfitzgerald]

All comments should be addressed.

[FrankenApps]

Btw.: half was recently updated with the PR on which this depends.

[cwfitzgerald]

Yup! I also updated it to use half 2.5 again

[ErichDonGubler]

@gfx-rs/wgpu: I will take on Mozilla's audit workload for half.

[ErichDonGubler]

Teo's out, we've handled his previous feedback, and we can iterate on his feedback in follow-up PRs.

[ErichDonGubler]

Resolved conflicts with the extraction of the ir module in naga, and added efabbf6 for a bit of clarity with the num-traits upgrade.

[raphamorim]

LET'S GOOOO, great job folks 🙏

[cwfitzgerald]

Wooooooooooooo!