Support `vfp5` target feature

ARM Cortex-M7 has `vfp5`, according to documentation - [1][doc-0] and [2][doc-1].

LLVM already has it, see https://github.com/llvm/llvm-project/issues/95053 (`FPUVersion::VFPV5`).

Seems to `rustc` doesn't supports it, throwing message
```
'+vfp5' is not a recognized feature for this target (ignoring feature)
```

Could be great to support `vfp5` in the `rustc` ❤️‍🔥


[doc-1]: https://embeddedartistry.com/blog/2017/10/11/demystifying-arm-floating-point-compiler-options/
[doc-0]: https://developer.arm.com/documentation/ddi0489/f/floating-point-unit/about-the-fpu


- - -

Oh I see that [there](https://github.com/rust-lang/rust/blob/master/compiler/rustc_target/src/spec/targets/thumbv7em_none_eabihf.rs#L29-L36) is `vfp4d16sp` feature. Am I right? Maybe could be good to create alias for it as `vfp5`?
https://github.com/rust-lang/rust/blob/35e6e4c5e9d3936df098d30750f28716740ea71b/compiler/rustc_target/src/spec/targets/thumbv7em_none_eabihf.rs#L29-L36

Also I see features
- `vfp4d16` - VFP4 instructions with only 16 d-registers
- `vfp4d16sp` - VFP4 instructions with only 16 d-registers and no double precision
- `vfp4sp` -  VFP4 instructions with no double precision

	// Cortex-M7 (vfp5).
	// Both the Cortex-M4 and the Cortex-M7 only have 16 double-precision registers
	// available, and the Cortex-M4 only supports single-precision floating point operations
	// whereas in the Cortex-M7 double-precision is optional.
	//
	// Reference:
	// ARMv7-M Architecture Reference Manual - A2.5 The optional floating-point extension
	features: "+vfp4d16sp".into(),

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Support `vfp5` target feature #127449

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Support vfp5 target feature #127449

Description