Improve perfomance for XXH128 & XXH3

Author: uranium62

CHANGELOG.md

@@ -1,3 +1,6 @@
+# 2022-07-17
+- Improve perfomance for XXH32 & XXH64
+- Improve perfomance for XXH128 & XXH3
 # 2022-06-13
 - Added xxHash3
 # 2022-06-05

README.md

@@ -55,10 +55,10 @@ Runtime=.NET 6.0
 | Hash32             |      x64 |       C  |  140.2 ns |  129.6 us |  150.3 ms | 6.65 GB/s  |
 | Hash64             |      x64 |       C# |  73.9 ns  |   64.6 us |  81.4 ms  | 12.28 GB/s |
 | Hash64             |      x64 |       C  |  75.5 ns  |   65.2 us |  84.5 ms  | 11.83 GB/s |
-| Hash128 (SSE2/AVX2)|      x64 |       C# |  151.6 ns |   64.5 us |  80.5 ms  | 12.04 GB/s |
-| Hash128 (SSE2/AVX2)|      x64 |       C  |  84.4 ns  |   38.3 us |  57.4 ms  | 17.42 GB/s |
-| Hash3   (SSE2/AVX2)|      x64 |       C# |  77.6 ns  |   62.1 us |  78.5 ms  | 12.08 GB/s |
-| Hash3   (SSE2/AVX2)|      x64 |       C  |  73.7 ns  |   42.2 us |  59.8 ms  | 16.72 GB/s |
+| Hash128 (SSE2/AVX2)|      x64 |       C# |  84.95 ns |   56.9 us |  73.2 ms  | 13.66 GB/s |
+| Hash128 (SSE2/AVX2)|      x64 |       C  |  84.35 ns |   38.1 us |  57.2 ms  | 17.48 GB/s |
+| Hash3   (SSE2/AVX2)|      x64 |       C# |  75.8 ns  |   56.6 us |  74.6 ms  | 13.40 GB/s |
+| Hash3   (SSE2/AVX2)|      x64 |       C  |  74.1 ns  |   42.1 us |  59.5 ms  | 16.80 GB/s |
 
 
 ## Api

nuget.props

@@ -3,7 +3,7 @@
   <PropertyGroup>
     <TargetFramework>net6.0</TargetFramework>
     <PackageId>Standart.Hash.xxHash</PackageId>
-    <VersionPrefix>4.0.3</VersionPrefix>
+    <VersionPrefix>4.0.4</VersionPrefix>
     <AssemblyName>Standart.Hash.xxHash</AssemblyName>
     <AssemblyTitle>Standart.Hash.xxHash</AssemblyTitle>
     <Authors>Oleksandr Melnyk</Authors>

src/Standart.Hash.xxHash/__inline__xxHash128.cs

@@ -1,30 +1,40 @@
 // ReSharper disable InconsistentNaming
 
 using System.Runtime.CompilerServices;
+using System.Runtime.Intrinsics;
 using System.Runtime.Intrinsics.X86;
 
 namespace Standart.Hash.xxHash
 {
     public static partial class xxHash128
     {
-        private const ulong XXH_PRIME64_1 = 11400714785074694791UL;
-        private const ulong XXH_PRIME64_2 = 14029467366897019727UL;
-        private const ulong XXH_PRIME64_3 = 1609587929392839161UL;
-        private const ulong XXH_PRIME64_4 = 9650029242287828579UL;
-        private const ulong XXH_PRIME64_5 = 2870177450012600261UL;
-        
-        private const uint XXH_PRIME32_1 = 2654435761U;
-        private const uint XXH_PRIME32_2 = 2246822519U;
-        private const uint XXH_PRIME32_3 = 3266489917U;
-        private const uint XXH_PRIME32_4 = 668265263U;
-        private const uint XXH_PRIME32_5 = 374761393U;
+        private static readonly ulong XXH_PRIME64_1 = 11400714785074694791UL;
+        private static readonly ulong XXH_PRIME64_2 = 14029467366897019727UL;
+        private static readonly ulong XXH_PRIME64_3 = 1609587929392839161UL;
+        private static readonly ulong XXH_PRIME64_4 = 9650029242287828579UL;
+        private static readonly ulong XXH_PRIME64_5 = 2870177450012600261UL;
+
+        private static readonly uint XXH_PRIME32_1 = 2654435761U;
+        private static readonly uint XXH_PRIME32_2 = 2246822519U;
+        private static readonly uint XXH_PRIME32_3 = 3266489917U;
+        private static readonly uint XXH_PRIME32_4 = 668265263U;
+        private static readonly uint XXH_PRIME32_5 = 374761393U;
+
+        private static readonly int XXH_STRIPE_LEN = 64;
+        private static readonly int XXH_ACC_NB = 8;
+        private static readonly int XXH_SECRET_CONSUME_RATE = 8;
+        private static readonly int XXH_SECRET_MERGEACCS_START = 11;
+        private static readonly int XXH_SECRET_DEFAULT_SIZE = 192;
+        private static readonly int XXH_SECRET_LASTACC_START = 7;
+
+        private static readonly byte MM_SHUFFLE_0_3_0_1 = 0b0011_0001;
+        private static readonly byte MM_SHUFFLE_1_0_3_2 = 0b0100_1110;
+
+        [FixedAddressValueType]
+        private static readonly Vector256<uint> M256i_XXH_PRIME32_1 = Vector256.Create(XXH_PRIME32_1);
+        [FixedAddressValueType]
+        private static readonly Vector128<uint> M128i_XXH_PRIME32_1 = Vector128.Create(XXH_PRIME32_1);
 
-        private const int XXH_STRIPE_LEN = 64;
-        private const int XXH_ACC_NB = XXH_STRIPE_LEN / 8;
-        private const int XXH_SECRET_CONSUME_RATE = 8;
-        private const int XXH_SECRET_MERGEACCS_START = 11;
-        private const int XXH_SECRET_DEFAULT_SIZE = 192;
-        private const int XXH_SECRET_LASTACC_START = 7;
 
         [MethodImpl(MethodImplOptions.AggressiveInlining)]
         private static uint XXH_rotl32(uint x, int r)