Enable f16 in assembly on aarch64 platforms that support it

Signed-off-by: rongfu.leng <[email protected]>

Author: lengrongfu

compiler/rustc_hir_analysis/src/check/intrinsicck.rs

@@ -62,6 +62,7 @@ impl<'a, 'tcx> InlineAsmCtxt<'a, 'tcx> {
             ty::Int(IntTy::I64) | ty::Uint(UintTy::U64) => Some(InlineAsmType::I64),
             ty::Int(IntTy::I128) | ty::Uint(UintTy::U128) => Some(InlineAsmType::I128),
             ty::Int(IntTy::Isize) | ty::Uint(UintTy::Usize) => Some(asm_ty_isize),
+            ty::Float(FloatTy::F16) => Some(InlineAsmType::F16),
             ty::Float(FloatTy::F32) => Some(InlineAsmType::F32),
             ty::Float(FloatTy::F64) => Some(InlineAsmType::F64),
             ty::FnPtr(_) => Some(asm_ty_isize),

compiler/rustc_target/src/asm/aarch64.rs

@@ -59,9 +59,9 @@ impl AArch64InlineAsmRegClass {
         match self {
             Self::reg => types! { _: I8, I16, I32, I64, F32, F64; },
             Self::vreg | Self::vreg_low16 => types! {
-                neon: I8, I16, I32, I64, F32, F64,
+                neon: I8, I16, I32, I64, F16, F32, F64,
                     VecI8(8), VecI16(4), VecI32(2), VecI64(1), VecF32(2), VecF64(1),
-                    VecI8(16), VecI16(8), VecI32(4), VecI64(2), VecF32(4), VecF64(2);
+                    VecI8(16), VecI16(8), VecI32(4), VecI64(2), VecF16(4),VecF16(8),VecF32(4), VecF64(2);
             },
             Self::preg => &[],
         }

compiler/rustc_target/src/asm/mod.rs

@@ -707,13 +707,15 @@ pub enum InlineAsmType {
     I32,
     I64,
     I128,
+    F16,
     F32,
     F64,
     VecI8(u64),
     VecI16(u64),
     VecI32(u64),
     VecI64(u64),
     VecI128(u64),
+    VecF16(u64),
     VecF32(u64),
     VecF64(u64),
 }
@@ -730,13 +732,15 @@ impl InlineAsmType {
             Self::I32 => 4,
             Self::I64 => 8,
             Self::I128 => 16,
+            Self::F16 => 2,
             Self::F32 => 4,
             Self::F64 => 8,
             Self::VecI8(n) => n * 1,
             Self::VecI16(n) => n * 2,
             Self::VecI32(n) => n * 4,
             Self::VecI64(n) => n * 8,
             Self::VecI128(n) => n * 16,
+            Self::VecF16(n) => n * 2,
             Self::VecF32(n) => n * 4,
             Self::VecF64(n) => n * 8,
         })
@@ -751,13 +755,15 @@ impl fmt::Display for InlineAsmType {
             Self::I32 => f.write_str("i32"),
             Self::I64 => f.write_str("i64"),
             Self::I128 => f.write_str("i128"),
+            Self::F16 => f.write_str("f16"),
             Self::F32 => f.write_str("f32"),
             Self::F64 => f.write_str("f64"),
             Self::VecI8(n) => write!(f, "i8x{n}"),
             Self::VecI16(n) => write!(f, "i16x{n}"),
             Self::VecI32(n) => write!(f, "i32x{n}"),
             Self::VecI64(n) => write!(f, "i64x{n}"),
             Self::VecI128(n) => write!(f, "i128x{n}"),
+            Self::VecF16(n) => write!(f, "f16x{n}"),
             Self::VecF32(n) => write!(f, "f32x{n}"),
             Self::VecF64(n) => write!(f, "f64x{n}"),
         }

tests/ui/asm/aarch64/type-check-3.stderr

@@ -111,7 +111,7 @@ error: type `Simd256bit` cannot be used with this register class
 LL |         asm!("{}", in(vreg) f64x4);
    |                             ^^^^^
    |
-   = note: register class `vreg` supports these types: i8, i16, i32, i64, f32, f64, i8x8, i16x4, i32x2, i64x1, f32x2, f64x1, i8x16, i16x8, i32x4, i64x2, f32x4, f64x2
+   = note: register class `vreg` supports these types: i8, i16, i32, i64, f16, f32, f64, i8x8, i16x4, i32x2, i64x1, f32x2, f64x1, i8x16, i16x8, i32x4, i64x2, f16x4, f16x8, f32x4, f64x2
 
 error: incompatible types for asm inout argument
   --> $DIR/type-check-3.rs:88:33

tests/ui/asm/aarch64/type-f16.rs

@@ -0,0 +1,21 @@
+//@ only-aarch64
+//@ run-pass
+
+#![feature(f16, f128)]
+use std::arch::asm;
+#[inline(never)]
+pub fn f32_to_f16_asm(a: f32) -> f16 {
+    let ret: f16;
+    unsafe {
+        asm!(
+        "fcvt    {ret:h}, {a:s}",
+        a = in(vreg) a,
+        ret = lateout(vreg) ret,
+        options(nomem, nostack),
+        );
+    }
+    ret
+}
+fn main() {
+    f32_to_f16_asm(1.0 as f32);
+}