f32x4::dot

Author: yury

cidre/src/simd.rs

@@ -370,6 +370,14 @@ impl f32x4 {
         Self(unsafe { std::arch::aarch64::vdupq_n_f32(val) })
     }
 
+    #[inline]
+    pub fn dot(&self, other: &Self) -> f32 {
+        unsafe {
+            let mul = std::arch::aarch64::vmulq_f32(self.0, other.0);
+            std::arch::aarch64::vaddvq_f32(mul)
+        }
+    }
+
     pub fn to_bits(&self) -> u128 {
         unsafe { std::mem::transmute(*self) }
     }
@@ -1318,6 +1326,13 @@ mod tests {
         assert_eq!(o1, o2);
     }
 
+    #[test]
+    fn f32x4_dot() {
+        let a = f32x4::with_xyzw(1.0, 2.0, 3.0, 4.0);
+        let b = f32x4::with_xyzw(5.0, 6.0, 7.0, 8.0);
+        assert_eq!(a.dot(&b), 70.0);
+    }
+
     #[cfg(feature = "half")]
     #[test]
     fn f16quat() {

cidre/src/simd/vector_types.rs

@@ -254,6 +254,14 @@ impl Simd<f32, 4, 4> {
     pub const fn with_rgba_f32(r: f32, g: f32, b: f32, a: f32) -> Self {
         Self([r, g, b, a])
     }
+
+    #[inline]
+    pub fn dot(&self, other: &Self) -> f32 {
+        self.0[0] * other.0[0]
+            + self.0[1] * other.0[1]
+            + self.0[2] * other.0[2]
+            + self.0[3] * other.0[3]
+    }
 }
 
 #[cfg(feature = "half")]