Update ShuffleNative (byte)

src/ImageSharp/Common/Helpers/SimdUtils.HwIntrinsics.cs

@@ -113,7 +113,7 @@ internal static partial class SimdUtils
             [ConstantExpected] byte control)
         {
             if ((Vector512.IsHardwareAccelerated && Vector512_.SupportsShuffleNativeByte) ||
-                (Vector256.IsHardwareAccelerated && Vector256_.SupportsShuffleByte) ||
+                (Vector256.IsHardwareAccelerated && Vector256_.SupportsShuffleNativeByte) ||
                 (Vector128.IsHardwareAccelerated && Vector128_.SupportsShuffleNativeByte))
             {
                 int remainder = 0;
@@ -158,7 +158,7 @@ internal static partial class SimdUtils
             ref Span<byte> destination,
             [ConstantExpected] byte control)
         {
-            if (Vector128.IsHardwareAccelerated && Vector128_.SupportsShuffleNativeByte && Vector128_.SupportsRightAlign)
+            if (Vector128.IsHardwareAccelerated && Vector128_.SupportsShuffleNativeByte && Vector128_.SupportsAlignRight)
             {
                 int remainder = source.Length % (Vector128<byte>.Count * 3);
 
@@ -373,7 +373,7 @@ internal static partial class SimdUtils
                     }
                 }
             }
-            else if (Vector256.IsHardwareAccelerated && Vector256_.SupportsShuffleByte)
+            else if (Vector256.IsHardwareAccelerated && Vector256_.SupportsShuffleNativeByte)
             {
                 Span<byte> temp = stackalloc byte[Vector256<byte>.Count];
                 Shuffle.MMShuffleSpan(ref temp, control);
@@ -445,7 +445,9 @@ internal static partial class SimdUtils
             Span<byte> destination,
             [ConstantExpected] byte control)
         {
-            if (Vector128.IsHardwareAccelerated && Vector128_.SupportsShuffleNativeByte && Vector128_.SupportsRightAlign)
+            if (Vector128.IsHardwareAccelerated &&
+                Vector128_.SupportsShuffleNativeByte &&
+                Vector128_.SupportsAlignRight)
             {
                 Vector128<byte> maskPad4Nx16 = ShuffleMaskPad4Nx16();
                 Vector128<byte> maskSlice4Nx16 = ShuffleMaskSlice4Nx16();
@@ -505,7 +507,10 @@ internal static partial class SimdUtils
             Span<byte> destination,
             [ConstantExpected] byte control)
         {
-            if (Vector128.IsHardwareAccelerated && Vector128_.SupportsShuffleNativeByte && Vector128_.SupportsShiftByte)
+            if (Vector128.IsHardwareAccelerated &&
+                Vector128_.SupportsShuffleNativeByte &&
+                Vector128_.SupportsShiftByte &&
+                Vector128_.SupportsAlignRight)
             {
                 Vector128<byte> maskPad4Nx16 = ShuffleMaskPad4Nx16();
                 Vector128<byte> fill = Vector128.Create(0xff000000ff000000ul).AsByte();
@@ -548,7 +553,10 @@ internal static partial class SimdUtils
             Span<byte> destination,
             [ConstantExpected] byte control)
         {
-            if (Vector128.IsHardwareAccelerated && Vector128_.SupportsShuffleNativeByte && Vector128_.SupportsShiftByte)
+            if (Vector128.IsHardwareAccelerated &&
+                Vector128_.SupportsShuffleNativeByte &&
+                Vector128_.SupportsShiftByte &&
+                Vector128_.SupportsAlignRight)
             {
                 Vector128<byte> maskSlice4Nx16 = ShuffleMaskSlice4Nx16();
                 Vector128<byte> maskE = Vector128_.AlignRight(maskSlice4Nx16, maskSlice4Nx16, 12);

src/ImageSharp/Common/Helpers/Vector128Utilities.cs

@@ -4,6 +4,7 @@
 using System.Diagnostics;
 using System.Diagnostics.CodeAnalysis;
 using System.Runtime.CompilerServices;
+using System.Runtime.InteropServices;
 using System.Runtime.Intrinsics;
 using System.Runtime.Intrinsics.Arm;
 using System.Runtime.Intrinsics.Wasm;
@@ -38,13 +39,26 @@ internal static class Vector128_
     public static bool SupportsShuffleNativeByte
     {
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
-        get => Ssse3.IsSupported || AdvSimd.Arm64.IsSupported || PackedSimd.IsSupported;
+        get
+        {
+            if (Vector128.IsHardwareAccelerated)
+            {
+                if (RuntimeInformation.ProcessArchitecture is Architecture.X86 or Architecture.X64)
+                {
+                    return Ssse3.IsSupported;
+                }
+
+                return true;
+            }
+
+            return false;
+        }
     }
 
     /// <summary>
     /// Gets a value indicating whether right align operations are supported.
     /// </summary>
-    public static bool SupportsRightAlign
+    public static bool SupportsAlignRight
     {
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         get => Ssse3.IsSupported || AdvSimd.IsSupported;
@@ -91,23 +105,16 @@ internal static class Vector128_
     [MethodImpl(MethodImplOptions.AggressiveInlining)]
     public static Vector128<byte> ShuffleNative(Vector128<byte> vector, Vector128<byte> indices)
     {
+        // For x64 we use the SSSE3 shuffle intrinsic to avoid additional instructions. 3 vs 1.
         if (Ssse3.IsSupported)
         {
             return Ssse3.Shuffle(vector, indices);
         }
 
-        if (AdvSimd.Arm64.IsSupported)
+        // For ARM and WASM, codegen will be optimal.
-        {
+        // We don't throw for x86/x64 so we should never use this method without
-            return AdvSimd.Arm64.VectorTableLookup(vector, indices);
+        // checking for support.
-        }
+        return Vector128.Shuffle(vector, indices);
-
-        if (PackedSimd.IsSupported)
-        {
-            return PackedSimd.Swizzle(vector, indices);
-        }
-
-        ThrowUnreachableException();
-        return default;
     }
 
     /// <summary>

src/ImageSharp/Common/Helpers/Vector256Utilities.cs

@@ -33,7 +33,7 @@ internal static class Vector256_
     /// <summary>
     /// Gets a value indicating whether shuffle byte operations are supported.
     /// </summary>
-    public static bool SupportsShuffleByte
+    public static bool SupportsShuffleNativeByte
     {
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         get => Avx2.IsSupported;

src/ImageSharp/Formats/Jpeg/Components/Block8x8F.cs

@@ -5,7 +5,6 @@ using System.Numerics;
 using System.Runtime.CompilerServices;
 using System.Runtime.InteropServices;
 using System.Runtime.Intrinsics;
-using System.Runtime.Intrinsics.X86;
 using System.Text;
 using SixLabors.ImageSharp.Common.Helpers;