using System;
using System.Numerics;
using System.Runtime.CompilerServices;
using System.Runtime.InteropServices;
using System.Runtime.Intrinsics;
using System.Runtime.Intrinsics.X86;

namespace Intrinsics
{
    [StructLayout(LayoutKind.Sequential)]
    public struct SrVector
    {
        public readonly float X;
        public readonly float Y;
        public readonly float Z;

        private Vector128<float> _m128;

        public SrVector(float x, float y, float z)
        {
            X = x;
            Y = y;
            Z = z;
            _m128 = Vector128.Create(X, Y, Z, 0f);
        }

        [MethodImpl(MethodImplOptions.AggressiveInlining)]
        public static float Dot(SrVector a, SrVector b) => Sse41.DotProduct(a._m128, b._m128, 0x7F).ToScalar();
    }

    class Program
    {
        static void Main(string[] args)
        {
            // Do something with the results to ensure the compiler does not optimize the computations away in Release build.
            float sum = 0;

            Vector3 v4a = new Vector3(1f, 0f, 3.2f);
            Vector3 v4b = new Vector3(2f, 2f, 0f);
            var vdot = Vector3.Dot(v4a, v4b);
            sum += vdot;

            SrVector sra = new SrVector(1f, 0f, 3.2f);
            SrVector srb = new SrVector(2f, 2f, 0f);
            var srdot = SrVector.Dot(sra, srb);
            sum += srdot;

            Console.WriteLine(sum);
        }
    }
}