Modernize and optimize pixel format operations across platforms.

src/ImageSharp.ruleset

@@ -1,6 +1,7 @@
 <?xml version="1.0" encoding="utf-8"?>
 <RuleSet Name="ImageSharp" ToolsVersion="17.0">
   <Include Path="..\shared-infrastructure\sixlabors.ruleset" Action="Default" />
-  <Rules AnalyzerId="StyleCop.Analyzers" RuleNamespace="StyleCop.Analyzers">
+  <Rules AnalyzerId="Microsoft.CodeAnalysis.CSharp.Features" RuleNamespace="Microsoft.CodeAnalysis.CSharp.Features">
+    <Rule Id="IDE0290" Action="None" />
   </Rules>
-</RuleSet>
+</RuleSet>

src/ImageSharp/Color/Color.cs

@@ -98,10 +98,8 @@ public static Color FromPixel<TPixel>(TPixel pixel)
         {
             return new(pixel.ToScaledVector4());
         }
-        else
-        {
-            return new(pixel);
-        }
+
+        return new(pixel);
     }
 
     /// <summary>
@@ -250,15 +248,12 @@ public Color WithAlpha(float alpha)
     [MethodImpl(InliningOptions.ShortMethod)]
     public string ToHex()
     {
-        Rgba32 rgba = default;
         if (this.boxedHighPrecisionPixel is not null)
         {
-            this.boxedHighPrecisionPixel.ToRgba32(ref rgba);
-            return rgba.ToHex();
+            return this.boxedHighPrecisionPixel.ToRgba32().ToHex();
         }
 
-        rgba.FromScaledVector4(this.data);
-        return rgba.ToHex();
+        return Rgba32.FromScaledVector4(this.data).ToHex();
     }
 
     /// <inheritdoc />
@@ -280,14 +275,10 @@ public TPixel ToPixel<TPixel>()
 
         if (this.boxedHighPrecisionPixel is null)
         {
-            pixel = default;
-            pixel.FromScaledVector4(this.data);
-            return pixel;
+            return TPixel.FromScaledVector4(this.data);
         }
 
-        pixel = default;
-        pixel.FromScaledVector4(this.boxedHighPrecisionPixel.ToScaledVector4());
-        return pixel;
+        return TPixel.FromScaledVector4(this.boxedHighPrecisionPixel.ToScaledVector4());
     }
 
     /// <summary>

src/ImageSharp/Common/Helpers/ColorNumerics.cs

@@ -41,6 +41,34 @@ public static int GetBT709Luminance(ref Vector4 vector, int luminanceLevels)
     public static byte Get8BitBT709Luminance(byte r, byte g, byte b)
         => (byte)((r * .2126F) + (g * .7152F) + (b * .0722F) + 0.5F);
 
+    /// <summary>
+    /// Gets the luminance from the rgb components using the formula
+    /// as specified by ITU-R Recommendation BT.709.
+    /// </summary>
+    /// <param name="r">The red component.</param>
+    /// <param name="g">The green component.</param>
+    /// <param name="b">The blue component.</param>
+    /// <returns>The <see cref="byte"/>.</returns>
+    [MethodImpl(MethodImplOptions.AggressiveInlining)]
+    public static byte Get8BitBT709Luminance(ushort r, ushort g, ushort b)
+        => (byte)((From16BitTo8Bit(r) * .2126F) +
+                  (From16BitTo8Bit(g) * .7152F) +
+                  (From16BitTo8Bit(b) * .0722F) + 0.5F);
+
+    /// <summary>
+    /// Gets the luminance from the rgb components using the formula as
+    /// specified by ITU-R Recommendation BT.709.
+    /// </summary>
+    /// <param name="r">The red component.</param>
+    /// <param name="g">The green component.</param>
+    /// <param name="b">The blue component.</param>
+    /// <returns>The <see cref="ushort"/>.</returns>
+    [MethodImpl(MethodImplOptions.AggressiveInlining)]
+    public static ushort Get16BitBT709Luminance(byte r, byte g, byte b)
+        => (ushort)((From8BitTo16Bit(r) * .2126F) +
+                    (From8BitTo16Bit(g) * .7152F) +
+                    (From8BitTo16Bit(b) * .0722F) + 0.5F);
+
     /// <summary>
     /// Gets the luminance from the rgb components using the formula as
     /// specified by ITU-R Recommendation BT.709.
@@ -72,8 +100,8 @@ public static ushort Get16BitBT709Luminance(float r, float g, float b)
     /// <param name="component">The 8 bit component value.</param>
     /// <returns>The <see cref="byte"/></returns>
     [MethodImpl(MethodImplOptions.AggressiveInlining)]
-    public static byte DownScaleFrom16BitTo8Bit(ushort component)
-    {
+    public static byte From16BitTo8Bit(ushort component) =>
+
         // To scale to 8 bits From a 16-bit value V the required value (from the PNG specification) is:
         //
         //    (V * 255) / 65535
@@ -102,8 +130,7 @@ public static byte DownScaleFrom16BitTo8Bit(ushort component)
         // An alternative arithmetic calculation which also gives no errors is:
         //
         //    (V * 255 + 32895) >> 16
-        return (byte)(((component * 255) + 32895) >> 16);
-    }
+        (byte)(((component * 255) + 32895) >> 16);
 
     /// <summary>
     /// Scales a value from an 8 bit <see cref="byte"/> to
@@ -112,7 +139,7 @@ public static byte DownScaleFrom16BitTo8Bit(ushort component)
     /// <param name="component">The 8 bit component value.</param>
     /// <returns>The <see cref="ushort"/></returns>
     [MethodImpl(MethodImplOptions.AggressiveInlining)]
-    public static ushort UpscaleFrom8BitTo16Bit(byte component)
+    public static ushort From8BitTo16Bit(byte component)
         => (ushort)(component * 257);
 
     /// <summary>

src/ImageSharp/Common/Helpers/Numerics.cs

@@ -5,7 +5,6 @@
 using System.Runtime.CompilerServices;
 using System.Runtime.InteropServices;
 using System.Runtime.Intrinsics;
-using System.Runtime.Intrinsics.Arm;
 using System.Runtime.Intrinsics.X86;
 
 namespace SixLabors.ImageSharp;
@@ -61,6 +60,12 @@ public static int LeastCommonMultiple(int a, int b)
     [MethodImpl(MethodImplOptions.AggressiveInlining)]
     public static nint Modulo4(nint x) => x & 3;
 
+    /// <summary>
+    /// Calculates <paramref name="x"/> % 4
+    /// </summary>
+    [MethodImpl(MethodImplOptions.AggressiveInlining)]
+    public static nuint Modulo4(nuint x) => x & 3;
+
     /// <summary>
     /// Calculates <paramref name="x"/> % 8
     /// </summary>

src/ImageSharp/Common/Helpers/Shuffle/IComponentShuffle.cs

@@ -1,12 +1,6 @@
 // Copyright (c) Six Labors.
 // Licensed under the Six Labors Split License.
 
-using System.Buffers.Binary;
-using System.Numerics;
-using System.Runtime.CompilerServices;
-using System.Runtime.InteropServices;
-using static SixLabors.ImageSharp.SimdUtils;
-
 // The JIT can detect and optimize rotation idioms ROTL (Rotate Left)
 // and ROTR (Rotate Right) emitting efficient CPU instructions:
 // https://github.com/dotnet/coreclr/pull/1830
@@ -19,190 +13,24 @@ namespace SixLabors.ImageSharp;
 internal interface IComponentShuffle
 {
     /// <summary>
-    /// Shuffles then slices 8-bit integers within 128-bit lanes in <paramref name="source"/>
-    /// using the control and store the results in <paramref name="dest"/>.
+    /// Shuffles then slices 8-bit integers in <paramref name="source"/>
+    /// using a byte control and store the results in <paramref name="destination"/>.
+    /// If successful, this method will reduce the length of <paramref name="source"/> length
+    /// by the shuffle amount.
     /// </summary>
     /// <param name="source">The source span of bytes.</param>
-    /// <param name="dest">The destination span of bytes.</param>
-    void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> dest);
+    /// <param name="destination">The destination span of bytes.</param>
+    void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> destination);
 
     /// <summary>
-    /// Shuffle 8-bit integers within 128-bit lanes in <paramref name="source"/>
-    /// using the control and store the results in <paramref name="dest"/>.
+    /// Shuffle 8-bit integers in <paramref name="source"/>
+    /// using the control and store the results in <paramref name="destination"/>.
     /// </summary>
     /// <param name="source">The source span of bytes.</param>
-    /// <param name="dest">The destination span of bytes.</param>
+    /// <param name="destination">The destination span of bytes.</param>
     /// <remarks>
-    /// Implementation can assume that source.Length is less or equal than dest.Length.
+    /// Implementation can assume that source.Length is less or equal than destination.Length.
     /// Loops should iterate using source.Length.
     /// </remarks>
-    void RunFallbackShuffle(ReadOnlySpan<byte> source, Span<byte> dest);
-}
-
-/// <inheritdoc/>
-internal interface IShuffle4 : IComponentShuffle
-{
-}
-
-internal readonly struct DefaultShuffle4 : IShuffle4
-{
-    public DefaultShuffle4(byte control)
-        => this.Control = control;
-
-    public byte Control { get; }
-
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> dest)
-        => HwIntrinsics.Shuffle4Reduce(ref source, ref dest, this.Control);
-
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void RunFallbackShuffle(ReadOnlySpan<byte> source, Span<byte> dest)
-    {
-        ref byte sBase = ref MemoryMarshal.GetReference(source);
-        ref byte dBase = ref MemoryMarshal.GetReference(dest);
-
-        Shuffle.InverseMMShuffle(this.Control, out uint p3, out uint p2, out uint p1, out uint p0);
-
-        for (nuint i = 0; i < (uint)source.Length; i += 4)
-        {
-            Unsafe.Add(ref dBase, i + 0) = Unsafe.Add(ref sBase, p0 + i);
-            Unsafe.Add(ref dBase, i + 1) = Unsafe.Add(ref sBase, p1 + i);
-            Unsafe.Add(ref dBase, i + 2) = Unsafe.Add(ref sBase, p2 + i);
-            Unsafe.Add(ref dBase, i + 3) = Unsafe.Add(ref sBase, p3 + i);
-        }
-    }
-}
-
-internal readonly struct WXYZShuffle4 : IShuffle4
-{
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> dest)
-        => HwIntrinsics.Shuffle4Reduce(ref source, ref dest, Shuffle.MMShuffle2103);
-
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void RunFallbackShuffle(ReadOnlySpan<byte> source, Span<byte> dest)
-    {
-        ref uint sBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(source));
-        ref uint dBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(dest));
-        uint n = (uint)source.Length / 4;
-
-        for (nuint i = 0; i < n; i++)
-        {
-            uint packed = Unsafe.Add(ref sBase, i);
-
-            // packed          = [W Z Y X]
-            // ROTL(8, packed) = [Z Y X W]
-            Unsafe.Add(ref dBase, i) = (packed << 8) | (packed >> 24);
-        }
-    }
-}
-
-internal readonly struct WZYXShuffle4 : IShuffle4
-{
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> dest)
-        => HwIntrinsics.Shuffle4Reduce(ref source, ref dest, Shuffle.MMShuffle0123);
-
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void RunFallbackShuffle(ReadOnlySpan<byte> source, Span<byte> dest)
-    {
-        ref uint sBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(source));
-        ref uint dBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(dest));
-        uint n = (uint)source.Length / 4;
-
-        for (nuint i = 0; i < n; i++)
-        {
-            uint packed = Unsafe.Add(ref sBase, i);
-
-            // packed              = [W Z Y X]
-            // REVERSE(packedArgb) = [X Y Z W]
-            Unsafe.Add(ref dBase, i) = BinaryPrimitives.ReverseEndianness(packed);
-        }
-    }
-}
-
-internal readonly struct YZWXShuffle4 : IShuffle4
-{
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> dest)
-        => HwIntrinsics.Shuffle4Reduce(ref source, ref dest, Shuffle.MMShuffle0321);
-
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void RunFallbackShuffle(ReadOnlySpan<byte> source, Span<byte> dest)
-    {
-        ref uint sBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(source));
-        ref uint dBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(dest));
-        uint n = (uint)source.Length / 4;
-
-        for (nuint i = 0; i < n; i++)
-        {
-            uint packed = Unsafe.Add(ref sBase, i);
-
-            // packed              = [W Z Y X]
-            // ROTR(8, packedArgb) = [Y Z W X]
-            Unsafe.Add(ref dBase, i) = BitOperations.RotateRight(packed, 8);
-        }
-    }
-}
-
-internal readonly struct ZYXWShuffle4 : IShuffle4
-{
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> dest)
-        => HwIntrinsics.Shuffle4Reduce(ref source, ref dest, Shuffle.MMShuffle3012);
-
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void RunFallbackShuffle(ReadOnlySpan<byte> source, Span<byte> dest)
-    {
-        ref uint sBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(source));
-        ref uint dBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(dest));
-        uint n = (uint)source.Length / 4;
-
-        for (nuint i = 0; i < n; i++)
-        {
-            uint packed = Unsafe.Add(ref sBase, i);
-
-            // packed              = [W Z Y X]
-            // tmp1                = [W 0 Y 0]
-            // tmp2                = [0 Z 0 X]
-            // tmp3=ROTL(16, tmp2) = [0 X 0 Z]
-            // tmp1 + tmp3         = [W X Y Z]
-            uint tmp1 = packed & 0xFF00FF00;
-            uint tmp2 = packed & 0x00FF00FF;
-            uint tmp3 = BitOperations.RotateLeft(tmp2, 16);
-
-            Unsafe.Add(ref dBase, i) = tmp1 + tmp3;
-        }
-    }
-}
-
-internal readonly struct XWZYShuffle4 : IShuffle4
-{
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void ShuffleReduce(ref ReadOnlySpan<byte> source, ref Span<byte> dest)
-        => HwIntrinsics.Shuffle4Reduce(ref source, ref dest, Shuffle.MMShuffle1230);
-
-    [MethodImpl(InliningOptions.ShortMethod)]
-    public void RunFallbackShuffle(ReadOnlySpan<byte> source, Span<byte> dest)
-    {
-        ref uint sBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(source));
-        ref uint dBase = ref Unsafe.As<byte, uint>(ref MemoryMarshal.GetReference(dest));
-        uint n = (uint)source.Length / 4;
-
-        for (nuint i = 0; i < n; i++)
-        {
-            uint packed = Unsafe.Add(ref sBase, i);
-
-            // packed              = [W Z Y X]
-            // tmp1                = [0 Z 0 X]
-            // tmp2                = [W 0 Y 0]
-            // tmp3=ROTL(16, tmp2) = [Y 0 W 0]
-            // tmp1 + tmp3         = [Y Z W X]
-            uint tmp1 = packed & 0x00FF00FF;
-            uint tmp2 = packed & 0xFF00FF00;
-            uint tmp3 = BitOperations.RotateLeft(tmp2, 16);
-
-            Unsafe.Add(ref dBase, i) = tmp1 + tmp3;
-        }
-    }
+    void Shuffle(ReadOnlySpan<byte> source, Span<byte> destination);
 }