performace optimize: Avoid cast float to unsigned in Bgra4444.Pack

[7aGiven]

• I confirm that I am the author of this code and release it to the FNA project under the Ms-PL license. This contribution does not contain code from other sources, including code generated by a Large Language Model ("AI").

• I confirm that no changes being submitted have been derived from decompiled software, be it from Microsoft XNA or other proprietary software.

JPG emample from: https://file-examples.com/index.php/sample-images-download/sample-jpg-download/

Benchmark:

using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Running;
using Microsoft.Xna.Framework;
using Microsoft.Xna.Framework.Graphics;
using System;
using System.IO;

namespace Benchmark
{
    public static class MathHelper
    {
        public static float Clamp(float value, float min, float max)
        {
            if (value > max)
            {
                value = max;
            }
            else if (value < min)
            {
                value = min;
            }
            return value;
        }
    }
}

public struct Bgra4444
{
    internal static ushort Pack1(float x, float y, float z, float w)
    {
        return (ushort) (
            (((ushort) Math.Round(Benchmark.MathHelper.Clamp(x, 0, 1) * 15.0f)) << 8) |
            (((ushort) Math.Round(Benchmark.MathHelper.Clamp(y, 0, 1) * 15.0f)) << 4) |
            ((ushort) Math.Round(Benchmark.MathHelper.Clamp(z, 0, 1) * 15.0f)) |
            (((ushort) Math.Round(Benchmark.MathHelper.Clamp(w, 0, 1) * 15.0f)) << 12)
        );
    }
    internal static ushort Pack2(float x, float y, float z, float w)
    {
        return (ushort) (
            (((int) Math.Round(Benchmark.MathHelper.Clamp(x, 0, 1) * 15.0f)) << 8) |
            (((int) Math.Round(Benchmark.MathHelper.Clamp(y, 0, 1) * 15.0f)) << 4) |
            ((int) Math.Round(Benchmark.MathHelper.Clamp(z, 0, 1) * 15.0f)) |
            (((int) Math.Round(Benchmark.MathHelper.Clamp(w, 0, 1) * 15.0f)) << 12)
        );
    }
}

[DisassemblyDiagnoser]
public class MyBenchmark
{
    private static Vector4[] colors;
    private static ushort[] bgras;
    static MyBenchmark()
    {
        Game game = new Game();
        PresentationParameters pp = new PresentationParameters() { DeviceWindowHandle = game.Window.Handle, IsFullScreen = false };
        GraphicsDevice gd = new GraphicsDevice(GraphicsAdapter.DefaultAdapter, GraphicsProfile.HiDef, pp);
        Texture2D texture = Texture2D.FromStream(gd, File.OpenRead(@"C:\tmp\file_example_JPG_500kB.jpg"));
        Color[] colors = new Color[texture.Width * texture.Height];
        texture.GetData(colors);
        MyBenchmark.colors = new Vector4[colors.Length];
        bgras = new ushort[colors.Length];
        for (int i = 0; i < colors.Length; i++)
        {
            MyBenchmark.colors[i] = colors[i].ToVector4();
        }
    }

    [Benchmark]
    public void Pack1()
    {
        Vector4[] colors = MyBenchmark.colors;
        ushort[] bgras = MyBenchmark.bgras;
        for (int i = 0; i < colors.Length; i++)
        {
            bgras[i] = Bgra4444.Pack1(colors[i].X, colors[i].Y, colors[i].Z, colors[i].W);
        }
    }
    [Benchmark]
    public void Pack2()
    {
        Vector4[] colors = MyBenchmark.colors;
        ushort[] bgras = MyBenchmark.bgras;
        for (int i = 0; i < colors.Length; i++)
        {
            bgras[i] = Bgra4444.Pack2(colors[i].X, colors[i].Y, colors[i].Z, colors[i].W);
        }
    }
}

static class Program
{
    [STAThread]
    static void Main()
    {
        BenchmarkRunner.Run<MyBenchmark>();
    }
}

.NET Framework 4.8.1 (4.8.9363.0), X86 LegacyJIT (Job: DefaultJob)

; Bgra4444.Pack1(Single, Single, Single, Single)
       push      ebp
       mov       ebp,esp
       push      esi
       sub       esp,0C
       push      dword ptr [ebp+14]
       push      0
       push      3F800000
       call      dword ptr ds:[104989FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[28C7B48]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       esi,[ebp-8]
       and       esi,0FFFF
       shl       esi,8
       push      dword ptr [ebp+10]
       push      0
       push      3F800000
       call      dword ptr ds:[104989FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[28C7B50]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       eax,[ebp-8]
       and       eax,0FFFF
       shl       eax,4
       or        eax,esi
       mov       esi,eax
       push      dword ptr [ebp+0C]
       push      0
       push      3F800000
       call      dword ptr ds:[104989FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[28C7B58]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       eax,[ebp-8]
       and       eax,0FFFF
       or        eax,esi
       mov       esi,eax
       push      dword ptr [ebp+8]
       push      0
       push      3F800000
       call      dword ptr ds:[104989FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[28C7B60]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       eax,[ebp-8]
       and       eax,0FFFF
       shl       eax,0C
       or        eax,esi
       and       eax,0FFFF
       lea       esp,[ebp-4]
       pop       esi
       pop       ebp
       ret       10
; Total bytes of code 196

.NET Framework 4.8.1 (4.8.9363.0), X86 LegacyJIT (Job: DefaultJob)

; Bgra4444.Pack2(Single, Single, Single, Single)
       push      ebp
       mov       ebp,esp
       push      esi
       sub       esp,0C
       push      dword ptr [ebp+14]
       push      0
       push      3F800000
       call      dword ptr ds:[104B89FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[49C7B30]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       esi,[ebp-8]
       shl       esi,8
       push      dword ptr [ebp+10]
       push      0
       push      3F800000
       call      dword ptr ds:[104B89FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[49C7B38]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       eax,[ebp-8]
       shl       eax,4
       or        eax,esi
       mov       esi,eax
       push      dword ptr [ebp+0C]
       push      0
       push      3F800000
       call      dword ptr ds:[104B89FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[49C7B40]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       eax,[ebp-8]
       or        eax,esi
       mov       esi,eax
       push      dword ptr [ebp+8]
       push      0
       push      3F800000
       call      dword ptr ds:[104B89FC]; Benchmark.MathHelper.Clamp(Single, Single, Single)
       fmul      dword ptr ds:[49C7B48]
       fstp      qword ptr [ebp-10]
       fld       qword ptr [ebp-10]
       fistp     dword ptr [ebp-8]
       mov       eax,[ebp-8]
       shl       eax,0C
       or        eax,esi
       and       eax,0FFFF
       lea       esp,[ebp-4]
       pop       esi
       pop       ebp
       ret       10
; Total bytes of code 175

Method	Mean	Error	StdDev	Code Size
Pack1	76.86 ms	0.842 ms	0.787 ms	315 B
Pack2	77.26 ms	1.261 ms	1.239 ms	294 B

[7aGiven]

link #862

[7aGiven]

Game rarely use PackedVecotr, So do benchmark to show assembly.

[7aGiven]

The asm difference is 4 instruction of and eax,0FFFF

So performance is rarely different, merge or not merge both OK.