Merge pull request #3102 from SixLabors/js/blend-perf-fixes

Add working-buffer blending and adjust row APIs

src/ImageSharp/Advanced/IRowOperation{TBuffer}.cs

@@ -15,12 +15,12 @@ public interface IRowOperation<TBuffer>
     /// </summary>
     /// <param name="bounds">The bounds of the operation.</param>
     /// <returns>The required buffer length.</returns>
-    int GetRequiredBufferLength(Rectangle bounds);
+    public int GetRequiredBufferLength(Rectangle bounds);
 
     /// <summary>
     /// Invokes the method passing the row and a buffer.
     /// </summary>
     /// <param name="y">The row y coordinate.</param>
     /// <param name="span">The contiguous region of memory.</param>
-    void Invoke(int y, Span<TBuffer> span);
+    public void Invoke(int y, Span<TBuffer> span);
 }

src/ImageSharp/Advanced/ParallelRowIterator.cs

@@ -68,7 +68,7 @@ public static partial class ParallelRowIterator
         ParallelOptions parallelOptions = new() { MaxDegreeOfParallelism = numOfSteps };
         RowOperationWrapper<T> wrappingOperation = new(top, bottom, verticalStep, in operation);
 
-        Parallel.For(
+        _ = Parallel.For(
             0,
             numOfSteps,
             parallelOptions,
@@ -138,7 +138,7 @@ public static partial class ParallelRowIterator
         ParallelOptions parallelOptions = new() { MaxDegreeOfParallelism = numOfSteps };
         RowOperationWrapper<T, TBuffer> wrappingOperation = new(top, bottom, verticalStep, bufferLength, allocator, in operation);
 
-        Parallel.For(
+        _ = Parallel.For(
             0,
             numOfSteps,
             parallelOptions,
@@ -195,7 +195,7 @@ public static partial class ParallelRowIterator
         ParallelOptions parallelOptions = new() { MaxDegreeOfParallelism = numOfSteps };
         RowIntervalOperationWrapper<T> wrappingOperation = new(top, bottom, verticalStep, in operation);
 
-        Parallel.For(
+        _ = Parallel.For(
             0,
             numOfSteps,
             parallelOptions,
@@ -262,7 +262,7 @@ public static partial class ParallelRowIterator
         ParallelOptions parallelOptions = new() { MaxDegreeOfParallelism = numOfSteps };
         RowIntervalOperationWrapper<T, TBuffer> wrappingOperation = new(top, bottom, verticalStep, bufferLength, allocator, in operation);
 
-        Parallel.For(
+        _ = Parallel.For(
             0,
             numOfSteps,
             parallelOptions,

src/ImageSharp/Formats/Webp/WebpAnimationDecoder.cs

@@ -2,6 +2,7 @@
 // Licensed under the Six Labors Split License.
 
 using System.Buffers;
+using System.Numerics;
 using SixLabors.ImageSharp.Formats.Webp.Chunks;
 using SixLabors.ImageSharp.Formats.Webp.Lossless;
 using SixLabors.ImageSharp.Formats.Webp.Lossy;
@@ -456,15 +457,21 @@ internal class WebpAnimationDecoder : IDisposable
             // The destination frame has already been prepopulated with the pixel data from the previous frame
             // so blending will leave the desired result which takes into consideration restoration to the
             // background color within the restore area.
-            PixelBlender<TPixel> blender =
+            PixelBlender<TPixel> blender = PixelOperations<TPixel>.Instance.GetPixelBlender(
-                PixelOperations<TPixel>.Instance.GetPixelBlender(PixelColorBlendingMode.Normal, PixelAlphaCompositionMode.SrcOver);
+                PixelColorBlendingMode.Normal,
+                PixelAlphaCompositionMode.SrcOver);
+
+            // By using a dedicated vector span we can avoid per-row pool allocations in PixelBlender.Blend
+            // We need 3 Vector4 values per pixel to store the background, foreground, and result pixels for blending.
+            using IMemoryOwner<Vector4> workingBufferOwner = imageFrame.Configuration.MemoryAllocator.Allocate<Vector4>(restoreArea.Width * 3);
+            Span<Vector4> workingBuffer = workingBufferOwner.GetSpan();
 
             for (int y = 0; y < restoreArea.Height; y++)
             {
                 Span<TPixel> framePixelRow = imageFramePixels.DangerousGetRowSpan(y);
                 Span<TPixel> decodedPixelRow = decodedImageFrame.DangerousGetRowSpan(y)[..restoreArea.Width];
 
-                blender.Blend<TPixel>(imageFrame.Configuration, framePixelRow, framePixelRow, decodedPixelRow, 1f);
+                blender.Blend<TPixel>(imageFrame.Configuration, framePixelRow, framePixelRow, decodedPixelRow, 1f, workingBuffer);
             }
 
             return;

src/ImageSharp/PixelFormats/PixelBlender{TPixel}.cs

@@ -3,7 +3,6 @@
 
 using System.Buffers;
 using System.Numerics;
-using SixLabors.ImageSharp.Memory;
 
 namespace SixLabors.ImageSharp.PixelFormats;
 
@@ -52,16 +51,53 @@ public abstract class PixelBlender<TPixel>
         Guard.MustBeBetweenOrEqualTo(amount, 0, 1, nameof(amount));
 
         using IMemoryOwner<Vector4> buffer = configuration.MemoryAllocator.Allocate<Vector4>(maxLength * 3);
-        Span<Vector4> destinationVectors = buffer.Slice(0, maxLength);
+        this.Blend(
-        Span<Vector4> backgroundVectors = buffer.Slice(maxLength, maxLength);
+            configuration,
-        Span<Vector4> sourceVectors = buffer.Slice(maxLength * 2, maxLength);
+            destination,
+            background,
+            source,
+            amount,
+            buffer.Memory.Span[..(maxLength * 3)]);
+    }
+
+    /// <summary>
+    /// Blends 2 rows together using caller-provided temporary vector scratch.
+    /// </summary>
+    /// <typeparam name="TPixelSrc">the pixel format of the source span</typeparam>
+    /// <param name="configuration"><see cref="Configuration"/> to use internally</param>
+    /// <param name="destination">the destination span</param>
+    /// <param name="background">the background span</param>
+    /// <param name="source">the source span</param>
+    /// <param name="amount">
+    /// A value between 0 and 1 indicating the weight of the second source vector.
+    /// At amount = 0, "background" is returned, at amount = 1, "source" is returned.
+    /// </param>
+    /// <param name="workingBuffer">Reusable temporary vector scratch with capacity for at least 3 rows.</param>
+    public void Blend<TPixelSrc>(
+        Configuration configuration,
+        Span<TPixel> destination,
+        ReadOnlySpan<TPixel> background,
+        ReadOnlySpan<TPixelSrc> source,
+        float amount,
+        Span<Vector4> workingBuffer)
+        where TPixelSrc : unmanaged, IPixel<TPixelSrc>
+    {
+        int maxLength = destination.Length;
+        Guard.MustBeGreaterThanOrEqualTo(background.Length, maxLength, nameof(background.Length));
+        Guard.MustBeGreaterThanOrEqualTo(source.Length, maxLength, nameof(source.Length));
+        Guard.MustBeBetweenOrEqualTo(amount, 0, 1, nameof(amount));
+        Guard.MustBeGreaterThanOrEqualTo(workingBuffer.Length, maxLength * 3, nameof(workingBuffer.Length));
+
+        Span<Vector4> destinationVectors = workingBuffer[..maxLength];
+        Span<Vector4> backgroundVectors = workingBuffer.Slice(maxLength, maxLength);
+        Span<Vector4> sourceVectors = workingBuffer.Slice(maxLength * 2, maxLength);
 
         PixelOperations<TPixel>.Instance.ToVector4(configuration, background[..maxLength], backgroundVectors, PixelConversionModifiers.Scale);
         PixelOperations<TPixelSrc>.Instance.ToVector4(configuration, source[..maxLength], sourceVectors, PixelConversionModifiers.Scale);
 
         this.BlendFunction(destinationVectors, backgroundVectors, sourceVectors, amount);
 
-        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors[..maxLength], destination, PixelConversionModifiers.Scale);
+        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors, destination, PixelConversionModifiers.Scale);
     }
 
     /// <summary>
@@ -87,14 +123,48 @@ public abstract class PixelBlender<TPixel>
         Guard.MustBeBetweenOrEqualTo(amount, 0, 1, nameof(amount));
 
         using IMemoryOwner<Vector4> buffer = configuration.MemoryAllocator.Allocate<Vector4>(maxLength * 2);
-        Span<Vector4> destinationVectors = buffer.Slice(0, maxLength);
+        this.Blend(
-        Span<Vector4> backgroundVectors = buffer.Slice(maxLength, maxLength);
+            configuration,
+            destination,
+            background,
+            source,
+            amount,
+            buffer.Memory.Span[..(maxLength * 2)]);
+    }
+
+    /// <summary>
+    /// Blends a row against a constant source color using caller-provided temporary vector scratch.
+    /// </summary>
+    /// <param name="configuration"><see cref="Configuration"/> to use internally</param>
+    /// <param name="destination">the destination span</param>
+    /// <param name="background">the background span</param>
+    /// <param name="source">the source color</param>
+    /// <param name="amount">
+    /// A value between 0 and 1 indicating the weight of the second source vector.
+    /// At amount = 0, "background" is returned, at amount = 1, "source" is returned.
+    /// </param>
+    /// <param name="workingBuffer">Reusable temporary vector scratch with capacity for at least 2 rows.</param>
+    public void Blend(
+        Configuration configuration,
+        Span<TPixel> destination,
+        ReadOnlySpan<TPixel> background,
+        TPixel source,
+        float amount,
+        Span<Vector4> workingBuffer)
+    {
+        int maxLength = destination.Length;
+        Guard.MustBeGreaterThanOrEqualTo(background.Length, maxLength, nameof(background.Length));
+        Guard.MustBeBetweenOrEqualTo(amount, 0, 1, nameof(amount));
+        Guard.MustBeGreaterThanOrEqualTo(workingBuffer.Length, maxLength * 2, nameof(workingBuffer.Length));
+
+        Span<Vector4> destinationVectors = workingBuffer[..maxLength];
+        Span<Vector4> backgroundVectors = workingBuffer.Slice(maxLength, maxLength);
 
         PixelOperations<TPixel>.Instance.ToVector4(configuration, background[..maxLength], backgroundVectors, PixelConversionModifiers.Scale);
 
         this.BlendFunction(destinationVectors, backgroundVectors, source.ToScaledVector4(), amount);
 
-        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors[..maxLength], destination, PixelConversionModifiers.Scale);
+        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors, destination, PixelConversionModifiers.Scale);
     }
 
     /// <summary>
@@ -116,6 +186,27 @@ public abstract class PixelBlender<TPixel>
         ReadOnlySpan<float> amount)
         => this.Blend<TPixel>(configuration, destination, background, source, amount);
 
+    /// <summary>
+    /// Blends 2 rows together using caller-provided temporary vector scratch.
+    /// </summary>
+    /// <param name="configuration"><see cref="Configuration"/> to use internally</param>
+    /// <param name="destination">the destination span</param>
+    /// <param name="background">the background span</param>
+    /// <param name="source">the source span</param>
+    /// <param name="amount">
+    /// A span with values between 0 and 1 indicating the weight of the second source vector.
+    /// At amount = 0, "background" is returned, at amount = 1, "source" is returned.
+    /// </param>
+    /// <param name="workingBuffer">Reusable temporary vector scratch with capacity for at least 3 rows.</param>
+    public void Blend(
+        Configuration configuration,
+        Span<TPixel> destination,
+        ReadOnlySpan<TPixel> background,
+        ReadOnlySpan<TPixel> source,
+        ReadOnlySpan<float> amount,
+        Span<Vector4> workingBuffer)
+        => this.Blend<TPixel>(configuration, destination, background, source, amount, workingBuffer);
+
     /// <summary>
     /// Blends 2 rows together
     /// </summary>
@@ -142,20 +233,89 @@ public abstract class PixelBlender<TPixel>
         Guard.MustBeGreaterThanOrEqualTo(amount.Length, maxLength, nameof(amount.Length));
 
         using IMemoryOwner<Vector4> buffer = configuration.MemoryAllocator.Allocate<Vector4>(maxLength * 3);
-        Span<Vector4> destinationVectors = buffer.Slice(0, maxLength);
+        this.Blend(
-        Span<Vector4> backgroundVectors = buffer.Slice(maxLength, maxLength);
+            configuration,
-        Span<Vector4> sourceVectors = buffer.Slice(maxLength * 2, maxLength);
+            destination,
+            background,
+            source,
+            amount,
+            buffer.Memory.Span[..(maxLength * 3)]);
+    }
+
+    /// <summary>
+    /// Blends a row against a constant source color.
+    /// </summary>
+    /// <param name="configuration"><see cref="Configuration"/> to use internally</param>
+    /// <param name="destination">the destination span</param>
+    /// <param name="background">the background span</param>
+    /// <param name="source">the source color</param>
+    /// <param name="amount">
+    /// A span with values between 0 and 1 indicating the weight of the second source vector.
+    /// At amount = 0, "background" is returned, at amount = 1, "source" is returned.
+    /// </param>
+    public void Blend(
+        Configuration configuration,
+        Span<TPixel> destination,
+        ReadOnlySpan<TPixel> background,
+        TPixel source,
+        ReadOnlySpan<float> amount)
+    {
+        int maxLength = destination.Length;
+        Guard.MustBeGreaterThanOrEqualTo(background.Length, maxLength, nameof(background.Length));
+        Guard.MustBeGreaterThanOrEqualTo(amount.Length, maxLength, nameof(amount.Length));
+
+        using IMemoryOwner<Vector4> buffer = configuration.MemoryAllocator.Allocate<Vector4>(maxLength * 2);
+        this.Blend(
+            configuration,
+            destination,
+            background,
+            source,
+            amount,
+            buffer.Memory.Span[..(maxLength * 2)]);
+    }
+
+    /// <summary>
+    /// Blends 2 rows together using caller-provided temporary vector scratch.
+    /// </summary>
+    /// <typeparam name="TPixelSrc">the pixel format of the source span</typeparam>
+    /// <param name="configuration"><see cref="Configuration"/> to use internally</param>
+    /// <param name="destination">the destination span</param>
+    /// <param name="background">the background span</param>
+    /// <param name="source">the source span</param>
+    /// <param name="amount">
+    /// A span with values between 0 and 1 indicating the weight of the second source vector.
+    /// At amount = 0, "background" is returned, at amount = 1, "source" is returned.
+    /// </param>
+    /// <param name="workingBuffer">Reusable temporary vector scratch with capacity for at least 3 rows.</param>
+    public void Blend<TPixelSrc>(
+        Configuration configuration,
+        Span<TPixel> destination,
+        ReadOnlySpan<TPixel> background,
+        ReadOnlySpan<TPixelSrc> source,
+        ReadOnlySpan<float> amount,
+        Span<Vector4> workingBuffer)
+        where TPixelSrc : unmanaged, IPixel<TPixelSrc>
+    {
+        int maxLength = destination.Length;
+        Guard.MustBeGreaterThanOrEqualTo(background.Length, maxLength, nameof(background.Length));
+        Guard.MustBeGreaterThanOrEqualTo(source.Length, maxLength, nameof(source.Length));
+        Guard.MustBeGreaterThanOrEqualTo(amount.Length, maxLength, nameof(amount.Length));
+        Guard.MustBeGreaterThanOrEqualTo(workingBuffer.Length, maxLength * 3, nameof(workingBuffer.Length));
+
+        Span<Vector4> destinationVectors = workingBuffer[..maxLength];
+        Span<Vector4> backgroundVectors = workingBuffer.Slice(maxLength, maxLength);
+        Span<Vector4> sourceVectors = workingBuffer.Slice(maxLength * 2, maxLength);
 
         PixelOperations<TPixel>.Instance.ToVector4(configuration, background[..maxLength], backgroundVectors, PixelConversionModifiers.Scale);
         PixelOperations<TPixelSrc>.Instance.ToVector4(configuration, source[..maxLength], sourceVectors, PixelConversionModifiers.Scale);
 
         this.BlendFunction(destinationVectors, backgroundVectors, sourceVectors, amount);
 
-        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors[..maxLength], destination, PixelConversionModifiers.Scale);
+        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors, destination, PixelConversionModifiers.Scale);
     }
 
     /// <summary>
-    /// Blends a row against a constant source color.
+    /// Blends a row against a constant source color using caller-provided temporary vector scratch.
     /// </summary>
     /// <param name="configuration"><see cref="Configuration"/> to use internally</param>
     /// <param name="destination">the destination span</param>
@@ -165,26 +325,28 @@ public abstract class PixelBlender<TPixel>
     /// A span with values between 0 and 1 indicating the weight of the second source vector.
     /// At amount = 0, "background" is returned, at amount = 1, "source" is returned.
     /// </param>
+    /// <param name="workingBuffer">Reusable temporary vector scratch with capacity for at least 2 rows.</param>
     public void Blend(
         Configuration configuration,
         Span<TPixel> destination,
         ReadOnlySpan<TPixel> background,
         TPixel source,
-        ReadOnlySpan<float> amount)
+        ReadOnlySpan<float> amount,
+        Span<Vector4> workingBuffer)
     {
         int maxLength = destination.Length;
         Guard.MustBeGreaterThanOrEqualTo(background.Length, maxLength, nameof(background.Length));
         Guard.MustBeGreaterThanOrEqualTo(amount.Length, maxLength, nameof(amount.Length));
+        Guard.MustBeGreaterThanOrEqualTo(workingBuffer.Length, maxLength * 2, nameof(workingBuffer.Length));
 
-        using IMemoryOwner<Vector4> buffer = configuration.MemoryAllocator.Allocate<Vector4>(maxLength * 2);
+        Span<Vector4> destinationVectors = workingBuffer[..maxLength];
-        Span<Vector4> destinationVectors = buffer.Slice(0, maxLength);
+        Span<Vector4> backgroundVectors = workingBuffer.Slice(maxLength, maxLength);
-        Span<Vector4> backgroundVectors = buffer.Slice(maxLength, maxLength);
 
         PixelOperations<TPixel>.Instance.ToVector4(configuration, background[..maxLength], backgroundVectors, PixelConversionModifiers.Scale);
 
         this.BlendFunction(destinationVectors, backgroundVectors, source.ToScaledVector4(), amount);
 
-        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors[..maxLength], destination, PixelConversionModifiers.Scale);
+        PixelOperations<TPixel>.Instance.FromVector4Destructive(configuration, destinationVectors, destination, PixelConversionModifiers.Scale);
     }
 
     /// <summary>

src/ImageSharp/Processing/Processors/Drawing/DrawImageProcessor{TPixelBg,TPixelFg}.cs

@@ -1,6 +1,7 @@
 // Copyright (c) Six Labors.
 // Licensed under the Six Labors Split License.
 
+using System.Numerics;
 using System.Runtime.CompilerServices;
 using SixLabors.ImageSharp.Advanced;
 using SixLabors.ImageSharp.Memory;
@@ -145,7 +146,7 @@ internal class DrawImageProcessor<TPixelBg, TPixelFg> : ImageProcessor<TPixelBg>
                 this.Blender,
                 this.Opacity);
 
-        ParallelRowIterator.IterateRows(
+        ParallelRowIterator.IterateRows<RowOperation, Vector4>(
             configuration,
             new Rectangle(0, 0, foregroundRectangle.Width, foregroundRectangle.Height),
             in operation);
@@ -161,7 +162,7 @@ internal class DrawImageProcessor<TPixelBg, TPixelFg> : ImageProcessor<TPixelBg>
     /// <summary>
     /// A <see langword="struct"/> implementing the draw logic for <see cref="DrawImageProcessor{TPixelBg,TPixelFg}"/>.
     /// </summary>
-    private readonly struct RowOperation : IRowOperation
+    private readonly struct RowOperation : IRowOperation<Vector4>
     {
         private readonly Buffer2D<TPixelBg> background;
         private readonly Buffer2D<TPixelFg> foreground;
@@ -190,13 +191,20 @@ internal class DrawImageProcessor<TPixelBg, TPixelFg> : ImageProcessor<TPixelBg>
             this.opacity = opacity;
         }
 
+        /// <inheritdoc/>
+        public int GetRequiredBufferLength(Rectangle bounds)
+
+            // By using a dedicated vector span we can avoid per-row pool allocations in PixelBlender.Blend
+            // We need 3 Vector4 values per pixel to store the background, foreground, and result pixels for blending.
+            => 3 * bounds.Width;
+
         /// <inheritdoc/>
         [MethodImpl(InliningOptions.ShortMethod)]
-        public void Invoke(int y)
+        public void Invoke(int y, Span<Vector4> span)
         {
             Span<TPixelBg> background = this.background.DangerousGetRowSpan(y + this.backgroundRectangle.Top).Slice(this.backgroundRectangle.Left, this.backgroundRectangle.Width);
             Span<TPixelFg> foreground = this.foreground.DangerousGetRowSpan(y + this.foregroundRectangle.Top).Slice(this.foregroundRectangle.Left, this.foregroundRectangle.Width);
-            this.blender.Blend<TPixelFg>(this.configuration, background, background, foreground, this.opacity);
+            this.blender.Blend<TPixelFg>(this.configuration, background, background, foreground, this.opacity, span);
         }
     }
 }