diff options
| author | Jacob Pratt <jacob@jhpratt.dev> | 2025-01-08 00:52:49 -0500 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-01-08 00:52:49 -0500 |
| commit | 57eb95ca6fe7cdacdf1c272ae106846ae3117f91 (patch) | |
| tree | 3c60673371f6b8d871429f7b10927e3cdb7dff50 | |
| parent | 45250f367ffa6c4325c05f10a87071742279c07f (diff) | |
| parent | 427abb69bf4c788504fa454fa03559e815c682df (diff) | |
| download | rust-57eb95ca6fe7cdacdf1c272ae106846ae3117f91.tar.gz rust-57eb95ca6fe7cdacdf1c272ae106846ae3117f91.zip | |
Rollup merge of #135203 - RalfJung:arm-soft-float, r=workingjubilee
arm: add unstable soft-float target feature This has an actual usecase as mentioned [here](https://github.com/rust-lang/rust/issues/116344#issuecomment-2575324988), and with my recent ARM float ABI changes there shouldn't be any soundness concerns any more. We will reject enabling this feature on `hf` targets, but disabling it on non-`hf` targets is entirely fine -- the target feature refers to whether softfloat emulation is used for float instructions, and is independent of the ABI which we set separately via `llvm_floatabi`. Cc ``@workingjubilee``
| -rw-r--r-- | compiler/rustc_target/src/target_features.rs | 8 | ||||
| -rw-r--r-- | tests/ui/check-cfg/target_feature.stderr | 1 |
2 files changed, 9 insertions, 0 deletions
diff --git a/compiler/rustc_target/src/target_features.rs b/compiler/rustc_target/src/target_features.rs index f594d20f928..9fd07c8634a 100644 --- a/compiler/rustc_target/src/target_features.rs +++ b/compiler/rustc_target/src/target_features.rs @@ -148,6 +148,11 @@ const ARM_FEATURES: &[(&str, Stability, ImpliedFeatures)] = &[ ("neon", Unstable(sym::arm_target_feature), &["vfp3"]), ("rclass", Unstable(sym::arm_target_feature), &[]), ("sha2", Unstable(sym::arm_target_feature), &["neon"]), + // This can be *disabled* on non-`hf` targets to enable the use + // of hardfloats while keeping the softfloat ABI. + // FIXME before stabilization: Should we expose this as a `hard-float` target feature instead of + // matching the odd negative feature LLVM uses? + ("soft-float", Unstable(sym::arm_target_feature), &[]), // This is needed for inline assembly, but shouldn't be stabilized as-is // since it should be enabled per-function using #[instruction_set], not // #[target_feature]. @@ -790,6 +795,9 @@ impl Target { match self.llvm_floatabi.unwrap() { FloatAbi::Soft => { // Nothing special required, will use soft-float ABI throughout. + // We can even allow `-soft-float` here; in fact that is useful as it lets + // people use FPU instructions with a softfloat ABI (corresponds to + // `-mfloat-abi=softfp` in GCC/clang). NOTHING } FloatAbi::Hard => { diff --git a/tests/ui/check-cfg/target_feature.stderr b/tests/ui/check-cfg/target_feature.stderr index 70fec8a350a..bf54d17f6ec 100644 --- a/tests/ui/check-cfg/target_feature.stderr +++ b/tests/ui/check-cfg/target_feature.stderr @@ -202,6 +202,7 @@ LL | cfg!(target_feature = "_UNEXPECTED_VALUE"); `sme-lutv2` `sme2` `sme2p1` +`soft-float` `spe` `ssbs` `sse` |