Add add_mul to vector

Author: stijnh

include/kernel_float/prelude.h

@@ -61,14 +61,14 @@ KERNEL_FLOAT_TYPE_ALIAS(f64x, double)
 KERNEL_FLOAT_TYPE_ALIAS(float64x, double)
 
 #if KERNEL_FLOAT_FP16_AVAILABLE
-KERNEL_FLOAT_TYPE_ALIAS(half, __half)
-KERNEL_FLOAT_TYPE_ALIAS(f16x, __half)
-KERNEL_FLOAT_TYPE_ALIAS(float16x, __half)
+KERNEL_FLOAT_TYPE_ALIAS(half, half_t)
+KERNEL_FLOAT_TYPE_ALIAS(f16x, half_t)
+KERNEL_FLOAT_TYPE_ALIAS(float16x, half_t)
 #endif
 
 #if KERNEL_FLOAT_BF16_AVAILABLE
-KERNEL_FLOAT_TYPE_ALIAS(bfloat16x, __bfloat16)
-KERNEL_FLOAT_TYPE_ALIAS(bf16x, __bfloat16)
+KERNEL_FLOAT_TYPE_ALIAS(bfloat16x, bfloat16_t)
+KERNEL_FLOAT_TYPE_ALIAS(bf16x, bfloat16_t)
 #endif
 
 #if KERNEL_FLOAT_BF8_AVAILABLE

include/kernel_float/vector.h

@@ -287,11 +287,20 @@ struct vector: public S {
     }
 
     /**
-     * Returns the result of `*this + lhs * rhs`.
+     * Returns the result of `this + lhs * rhs`.
      *
      * The operation is performed using a single `kernel_float::fma` call, which may be faster then perform
      * the addition and multiplication separately.
      */
+    template<
+        typename L,
+        typename R,
+        typename T2 = promote_t<T, vector_value_type<L>, vector_value_type<R>>,
+        typename E2 = broadcast_extent<E, vector_extent_type<L>, vector_extent_type<R>>>
+    KERNEL_FLOAT_INLINE vector<T2, E2> add_mul(const L& lhs, const R& rhs) const {
+        return ::kernel_float::fma(lhs, rhs, *this);
+    }
+
     template<
         typename L,
         typename R,