Merge pull request #1170 from SixLabors/js/faster-deflate

Faster Deflate

Directory.Build.props

@@ -31,21 +31,21 @@
 
   <!--
     https://apisof.net/
-    +===================+=======+==========+=====================+=============+=================+====================+==============+
-    | SUPPORTS          | MATHF | HASHCODE | EXTENDED_INTRINSICS | SPAN_STREAM | ENCODING_STRING | RUNTIME_INTRINSICS | CODECOVERAGE |
-    +===================+=======+==========+=====================+=============+=================+====================+==============+
-    | netcoreapp3.1     |   Y   |    Y     |         Y           |      Y      |        Y        |        Y           |      Y       |
-    | netcoreapp2.1     |   Y   |    Y     |         Y           |      Y      |        Y        |        N           |      Y       |
-    | netcoreapp2.0     |   Y   |    N     |         N           |      N      |        N        |        N           |      Y       |
-    | netstandard2.1    |   Y   |    Y     |         N           |      Y      |        Y        |        N           |      Y       |
-    | netstandard2.0    |   N   |    N     |         N           |      N      |        N        |        N           |      Y       |
-    | netstandard1.3    |   N   |    N     |         N           |      N      |        N        |        N           |      N       |
-    | net472            |   N   |    N     |         Y           |      N      |        N        |        N           |      Y       |
-    +===================+=======+==========+=====================+=============+=================+====================+==============+
+    +===================+=======+==========+=====================+=============+=================+====================+==============+=========+
+    | SUPPORTS          | MATHF | HASHCODE | EXTENDED_INTRINSICS | SPAN_STREAM | ENCODING_STRING | RUNTIME_INTRINSICS | CODECOVERAGE | HOTPATH |
+    +===================+=======+==========+=====================+=============+=================+====================+==============+=========|
+    | netcoreapp3.1     |   Y   |    Y     |         Y           |      Y      |        Y        |        Y           |      Y       |    Y    |
+    | netcoreapp2.1     |   Y   |    Y     |         Y           |      Y      |        Y        |        N           |      Y       |    N    |
+    | netcoreapp2.0     |   Y   |    N     |         N           |      N      |        N        |        N           |      Y       |    N    |
+    | netstandard2.1    |   Y   |    Y     |         N           |      Y      |        Y        |        N           |      Y       |    N    |
+    | netstandard2.0    |   N   |    N     |         N           |      N      |        N        |        N           |      Y       |    N    |
+    | netstandard1.3    |   N   |    N     |         N           |      N      |        N        |        N           |      N       |    N    |
+    | net472            |   N   |    N     |         Y           |      N      |        N        |        N           |      Y       |    N    |
+    +===================+=======+==========+=====================+=============+=================+====================+==============+=========|
     -->
 
   <PropertyGroup Condition="'$(TargetFramework)' == 'netcoreapp3.1'">
-    <DefineConstants>$(DefineConstants);SUPPORTS_MATHF;SUPPORTS_HASHCODE;SUPPORTS_EXTENDED_INTRINSICS;SUPPORTS_SPAN_STREAM;SUPPORTS_ENCODING_STRING;SUPPORTS_RUNTIME_INTRINSICS;SUPPORTS_CODECOVERAGE</DefineConstants>
+    <DefineConstants>$(DefineConstants);SUPPORTS_MATHF;SUPPORTS_HASHCODE;SUPPORTS_EXTENDED_INTRINSICS;SUPPORTS_SPAN_STREAM;SUPPORTS_ENCODING_STRING;SUPPORTS_RUNTIME_INTRINSICS;SUPPORTS_CODECOVERAGE;SUPPORTS_HOTPATH</DefineConstants>
   </PropertyGroup>
   <PropertyGroup Condition="'$(TargetFramework)' == 'netcoreapp2.1'">
     <DefineConstants>$(DefineConstants);SUPPORTS_MATHF;SUPPORTS_HASHCODE;SUPPORTS_EXTENDED_INTRINSICS;SUPPORTS_SPAN_STREAM;SUPPORTS_ENCODING_STRING;SUPPORTS_CODECOVERAGE</DefineConstants>

src/ImageSharp/Common/Helpers/InliningOptions.cs

@@ -1,4 +1,4 @@
-// Copyright (c) Six Labors and contributors.
+// Copyright (c) Six Labors and contributors.
 // Licensed under the Apache License, Version 2.0.
 
 // Uncomment this for verbose profiler results. DO NOT PUSH TO MAIN!
@@ -13,10 +13,16 @@ namespace SixLabors.ImageSharp
     internal static class InliningOptions
     {
 #if PROFILING
+        public const MethodImplOptions HotPath = MethodImplOptions.NoInlining;
         public const MethodImplOptions ShortMethod = MethodImplOptions.NoInlining;
 #else
+#if SUPPORTS_HOTPATH
+        public const MethodImplOptions HotPath = MethodImplOptions.AggressiveOptimization;
+#else
+        public const MethodImplOptions HotPath = MethodImplOptions.AggressiveInlining;
+#endif
         public const MethodImplOptions ShortMethod = MethodImplOptions.AggressiveInlining;
 #endif
         public const MethodImplOptions ColdPath = MethodImplOptions.NoInlining;
     }
-}
+}

src/ImageSharp/Formats/Png/PngDecoderCore.cs

@@ -380,7 +380,12 @@ namespace SixLabors.ImageSharp.Formats.Png
         private void InitializeImage<TPixel>(ImageMetadata metadata, out Image<TPixel> image)
             where TPixel : unmanaged, IPixel<TPixel>
         {
-            image = new Image<TPixel>(this.configuration, this.header.Width, this.header.Height, metadata);
+            image = Image.CreateUninitialized<TPixel>(
+                this.configuration,
+                this.header.Width,
+                this.header.Height,
+                metadata);
+
             this.bytesPerPixel = this.CalculateBytesPerPixel();
             this.bytesPerScanline = this.CalculateScanlineLength(this.header.Width) + 1;
             this.bytesPerSample = 1;

src/ImageSharp/Formats/Png/PngEncoderCore.cs

@@ -5,10 +5,8 @@ using System;
 using System.Buffers;
 using System.Buffers.Binary;
 using System.IO;
-using System.Linq;
 using System.Runtime.CompilerServices;
 using System.Runtime.InteropServices;
-
 using SixLabors.ImageSharp.Advanced;
 using SixLabors.ImageSharp.Formats.Png.Chunks;
 using SixLabors.ImageSharp.Formats.Png.Filters;
@@ -16,7 +14,6 @@ using SixLabors.ImageSharp.Formats.Png.Zlib;
 using SixLabors.ImageSharp.Memory;
 using SixLabors.ImageSharp.Metadata;
 using SixLabors.ImageSharp.PixelFormats;
-using SixLabors.ImageSharp.Processing.Processors.Quantization;
 
 namespace SixLabors.ImageSharp.Formats.Png
 {
@@ -633,10 +630,21 @@ namespace SixLabors.ImageSharp.Formats.Png
         private void WriteTextChunks(Stream stream, PngMetadata meta)
         {
             const int MaxLatinCode = 255;
-            foreach (PngTextData textData in meta.TextData)
+            for (int i = 0; i < meta.TextData.Count; i++)
             {
-                bool hasUnicodeCharacters = textData.Value.Any(c => c > MaxLatinCode);
-                if (hasUnicodeCharacters || (!string.IsNullOrWhiteSpace(textData.LanguageTag) || !string.IsNullOrWhiteSpace(textData.TranslatedKeyword)))
+                PngTextData textData = meta.TextData[i];
+                bool hasUnicodeCharacters = false;
+                foreach (var c in textData.Value)
+                {
+                    if (c > MaxLatinCode)
+                    {
+                        hasUnicodeCharacters = true;
+                        break;
+                    }
+                }
+
+                if (hasUnicodeCharacters || (!string.IsNullOrWhiteSpace(textData.LanguageTag) ||
+                                             !string.IsNullOrWhiteSpace(textData.TranslatedKeyword)))
                 {
                     // Write iTXt chunk.
                     byte[] keywordBytes = PngConstants.Encoding.GetBytes(textData.Keyword);
@@ -647,7 +655,8 @@ namespace SixLabors.ImageSharp.Formats.Png
                     byte[] translatedKeyword = PngConstants.TranslatedEncoding.GetBytes(textData.TranslatedKeyword);
                     byte[] languageTag = PngConstants.LanguageEncoding.GetBytes(textData.LanguageTag);
 
-                    Span<byte> outputBytes = new byte[keywordBytes.Length + textBytes.Length + translatedKeyword.Length + languageTag.Length + 5];
+                    Span<byte> outputBytes = new byte[keywordBytes.Length + textBytes.Length +
+                                                      translatedKeyword.Length + languageTag.Length + 5];
                     keywordBytes.CopyTo(outputBytes);
                     if (textData.Value.Length > this.options.TextCompressionThreshold)
                     {
@@ -667,7 +676,8 @@ namespace SixLabors.ImageSharp.Formats.Png
                     if (textData.Value.Length > this.options.TextCompressionThreshold)
                     {
                         // Write zTXt chunk.
-                        byte[] compressedData = this.GetCompressedTextBytes(PngConstants.Encoding.GetBytes(textData.Value));
+                        byte[] compressedData =
+                            this.GetCompressedTextBytes(PngConstants.Encoding.GetBytes(textData.Value));
                         Span<byte> outputBytes = new byte[textData.Keyword.Length + compressedData.Length + 2];
                         PngConstants.Encoding.GetBytes(textData.Keyword).CopyTo(outputBytes);
                         compressedData.CopyTo(outputBytes.Slice(textData.Keyword.Length + 2));
@@ -678,7 +688,8 @@ namespace SixLabors.ImageSharp.Formats.Png
                         // Write tEXt chunk.
                         Span<byte> outputBytes = new byte[textData.Keyword.Length + textData.Value.Length + 1];
                         PngConstants.Encoding.GetBytes(textData.Keyword).CopyTo(outputBytes);
-                        PngConstants.Encoding.GetBytes(textData.Value).CopyTo(outputBytes.Slice(textData.Keyword.Length + 1));
+                        PngConstants.Encoding.GetBytes(textData.Value)
+                            .CopyTo(outputBytes.Slice(textData.Keyword.Length + 1));
                         this.WriteChunk(stream, PngChunkType.Text, outputBytes.ToArray());
                     }
                 }

src/ImageSharp/Formats/Png/Zlib/Deflater.cs

@@ -288,8 +288,6 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
                 this.engine = null;
                 this.isDisposed = true;
             }
-
-            GC.SuppressFinalize(this);
         }
     }
 }

src/ImageSharp/Formats/Png/Zlib/DeflaterEngine.cs

@@ -362,7 +362,7 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
                     more = this.inputEnd - this.inputOff;
                 }
 
-                Array.Copy(this.inputBuf, this.inputOff, this.window, this.strstart + this.lookahead, more);
+                Buffer.BlockCopy(this.inputBuf, this.inputOff, this.window, this.strstart + this.lookahead, more);
 
                 this.inputOff += more;
                 this.lookahead += more;
@@ -397,8 +397,6 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
 
                 this.isDisposed = true;
             }
-
-            GC.SuppressFinalize(this);
         }
 
         [MethodImpl(InliningOptions.ShortMethod)]
@@ -464,6 +462,7 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
         /// </summary>
         /// <param name="curMatch">The current match.</param>
         /// <returns>True if a match greater than the minimum length is found</returns>
+        [MethodImpl(InliningOptions.HotPath)]
         private bool FindLongestMatch(int curMatch)
         {
             int match;

src/ImageSharp/Formats/Png/Zlib/DeflaterHuffman.cs

@@ -36,11 +36,6 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
 
         private const int EofSymbol = 256;
 
-        private static readonly short[] StaticLCodes;
-        private static readonly byte[] StaticLLength;
-        private static readonly short[] StaticDCodes;
-        private static readonly byte[] StaticDLength;
-
         private Tree literalTree;
         private Tree distTree;
         private Tree blTree;
@@ -58,49 +53,6 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
         private int extraBits;
         private bool isDisposed;
 
-        // TODO: These should be pre-generated array/readonlyspans.
-        static DeflaterHuffman()
-        {
-            // See RFC 1951 3.2.6
-            // Literal codes
-            StaticLCodes = new short[LiteralNumber];
-            StaticLLength = new byte[LiteralNumber];
-
-            int i = 0;
-            while (i < 144)
-            {
-                StaticLCodes[i] = BitReverse((0x030 + i) << 8);
-                StaticLLength[i++] = 8;
-            }
-
-            while (i < 256)
-            {
-                StaticLCodes[i] = BitReverse((0x190 - 144 + i) << 7);
-                StaticLLength[i++] = 9;
-            }
-
-            while (i < 280)
-            {
-                StaticLCodes[i] = BitReverse((0x000 - 256 + i) << 9);
-                StaticLLength[i++] = 7;
-            }
-
-            while (i < LiteralNumber)
-            {
-                StaticLCodes[i] = BitReverse((0x0c0 - 280 + i) << 8);
-                StaticLLength[i++] = 8;
-            }
-
-            // Distance codes
-            StaticDCodes = new short[DistanceNumber];
-            StaticDLength = new byte[DistanceNumber];
-            for (i = 0; i < DistanceNumber; i++)
-            {
-                StaticDCodes[i] = BitReverse(i << 11);
-                StaticDLength[i] = 5;
-            }
-        }
-
         /// <summary>
         /// Initializes a new instance of the <see cref="DeflaterHuffman"/> class.
         /// </summary>
@@ -122,12 +74,80 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
             this.pinnedLiteralBuffer = (short*)this.literalBufferHandle.Pointer;
         }
 
+#pragma warning disable SA1201 // Elements should appear in the correct order
+
+        // See RFC 1951 3.2.6
+        // Literal codes
+        private static readonly short[] StaticLCodes = new short[]
+        {
+            12, 140, 76, 204, 44, 172, 108, 236, 28, 156, 92, 220, 60, 188, 124, 252,
+            2, 130, 66, 194, 34, 162, 98, 226, 18, 146, 82, 210, 50, 178, 114, 242,
+            10, 138, 74, 202, 42, 170, 106, 234, 26, 154, 90, 218, 58, 186, 122, 250,
+            6, 134, 70, 198, 38, 166, 102, 230, 22, 150, 86, 214, 54, 182, 118, 246,
+            14, 142, 78, 206, 46, 174, 110, 238, 30, 158, 94, 222, 62, 190, 126, 254,
+            1, 129, 65, 193, 33, 161, 97, 225, 17, 145, 81, 209, 49, 177, 113, 241, 9,
+            137, 73, 201, 41, 169, 105, 233, 25, 153, 89, 217, 57, 185, 121, 249, 5,
+            133, 69, 197, 37, 165, 101, 229, 21, 149, 85, 213, 53, 181, 117, 245, 13,
+            141, 77, 205, 45, 173, 109, 237, 29, 157, 93, 221, 61, 189, 125, 253, 19,
+            275, 147, 403, 83, 339, 211, 467, 51, 307, 179, 435, 115, 371, 243, 499,
+            11, 267, 139, 395, 75, 331, 203, 459, 43, 299, 171, 427, 107, 363, 235, 491,
+            27, 283, 155, 411, 91, 347, 219, 475, 59, 315, 187, 443, 123, 379, 251, 507,
+            7, 263, 135, 391, 71, 327, 199, 455, 39, 295, 167, 423, 103, 359, 231, 487,
+            23, 279, 151, 407, 87, 343, 215, 471, 55, 311, 183, 439, 119, 375, 247, 503,
+            15, 271, 143, 399, 79, 335, 207, 463, 47, 303, 175, 431, 111, 367, 239, 495,
+            31, 287, 159, 415, 95, 351, 223, 479, 63, 319, 191, 447, 127, 383, 255, 511,
+            0, 64, 32, 96, 16, 80, 48, 112, 8, 72, 40, 104, 24, 88, 56, 120, 4, 68, 36,
+            100, 20, 84, 52, 116, 3, 131, 67, 195, 35, 163
+        };
+
+        private static ReadOnlySpan<byte> StaticLLength => new byte[]
+        {
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
+            9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+            9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+            9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+            9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+            9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+            9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+            9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9, 9,
+            7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7, 7,
+            7, 7, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8
+        };
+
+        // Distance codes and lengths.
+        private static readonly short[] StaticDCodes = new short[]
+        {
+            0, 16, 8, 24, 4, 20, 12, 28, 2, 18, 10, 26, 6, 22, 14,
+            30, 1, 17, 9, 25, 5, 21, 13, 29, 3, 19, 11, 27, 7, 23
+        };
+
+        private static ReadOnlySpan<byte> StaticDLength => new byte[]
+        {
+            5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5,
+            5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5, 5
+        };
+#pragma warning restore SA1201 // Elements should appear in the correct order
+
         /// <summary>
         /// Gets the lengths of the bit length codes are sent in order of decreasing probability, to avoid transmitting the lengths for unused bit length codes.
         /// </summary>
-        private static ReadOnlySpan<byte> BitLengthOrder => new byte[] { 16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15 };
+        private static ReadOnlySpan<byte> BitLengthOrder => new byte[]
+        {
+            16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15
+        };
 
-        private static ReadOnlySpan<byte> Bit4Reverse => new byte[] { 0, 8, 4, 12, 2, 10, 6, 14, 1, 9, 5, 13, 3, 11, 7, 15 };
+        private static ReadOnlySpan<byte> Bit4Reverse => new byte[]
+        {
+            0, 8, 4, 12, 2, 10, 6, 14, 1, 9, 5, 13, 3, 11, 7, 15
+        };
 
         /// <summary>
         /// Gets the pending buffer to use.
@@ -413,8 +433,6 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
                 this.distTree = null;
                 this.isDisposed = true;
             }
-
-            GC.SuppressFinalize(this);
         }
 
         [MethodImpl(InliningOptions.ShortMethod)]
@@ -553,6 +571,7 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
                 }
             }
 
+            [MethodImpl(InliningOptions.HotPath)]
             public void BuildTree()
             {
                 int numSymbols = this.elementCount;
@@ -964,8 +983,6 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
 
                     this.isDisposed = true;
                 }
-
-                GC.SuppressFinalize(this);
             }
         }
     }

src/ImageSharp/Formats/Png/Zlib/DeflaterPendingBuffer.cs

@@ -172,8 +172,6 @@ namespace SixLabors.ImageSharp.Formats.Png.Zlib
                 this.bufferMemoryOwner = null;
                 this.isDisposed = true;
             }
-
-            GC.SuppressFinalize(this);
         }
     }
 }