Benchmark cleanup and prepping for more refactoring

fennECS.sln

@@ -38,8 +38,6 @@ Global
 		{C75030DB-1B0D-44F7-B01F-018EA46178A4}.Debug|Any CPU.Build.0 = Debug|Any CPU
 		{C75030DB-1B0D-44F7-B01F-018EA46178A4}.Release|Any CPU.ActiveCfg = Release|Any CPU
 		{C75030DB-1B0D-44F7-B01F-018EA46178A4}.Release|Any CPU.Build.0 = Release|Any CPU
-		{08768DCD-54CB-4DAE-AD04-E618C1522753}.Debug|Any CPU.ActiveCfg = Debug|Any CPU
-		{08768DCD-54CB-4DAE-AD04-E618C1522753}.Debug|Any CPU.Build.0 = Debug|Any CPU
 		{08768DCD-54CB-4DAE-AD04-E618C1522753}.Release|Any CPU.ActiveCfg = Release|Any CPU
 		{08768DCD-54CB-4DAE-AD04-E618C1522753}.Release|Any CPU.Build.0 = Release|Any CPU
 		{67189493-DD27-484F-8761-135501F29E1E}.Debug|Any CPU.ActiveCfg = Debug|Any CPU

fennecs.benchmarks/ECS/CategoryExclusion.cs → fennecs.benchmarks/CategoryExclusion.cs

@@ -1,8 +1,8 @@
-namespace Benchmark.ECS;
-
-using BenchmarkDotNet.Filters;
+using BenchmarkDotNet.Filters;
 using BenchmarkDotNet.Running;
 
+namespace Benchmark;
+
 /// <summary>
 /// Excludes a given category from benchmarks
 /// </summary>

fennecs.benchmarks/Conceptual/RWBenchmarks.cs

@@ -1,18 +1,12 @@
-using System.Numerics;
-using System.Runtime.CompilerServices;
+using System.Runtime.CompilerServices;
 using System.Runtime.InteropServices;
 using BenchmarkDotNet.Attributes;
 using BenchmarkDotNet.Jobs;
 using fennecs;
-using fennecs.events;
-using fennecs.pools;
 using fennecs.storage;
 
 namespace Benchmark.Conceptual;
 
-public interface ITest
-{ }
-
 #if !NET9_0_OR_GREATER
     [AttributeUsage(AttributeTargets.Method | AttributeTargets.Constructor | AttributeTargets.Property, AllowMultiple = false, Inherited = false)]
     public sealed class OverloadResolutionPriorityAttribute : Attribute
@@ -34,40 +28,6 @@ public OverloadResolutionPriorityAttribute(int priority)
     }
 #endif
 
-record struct MyInt(in int value) : Fox<int>;
-
-record struct MyVector(in Vector3 value) : Fox<Vector3>
-{
-    private Vector3 _value = value;
-
-    public float X
-    {
-        get => _value.X;
-        set => _value.X = value;
-    }
-
-    public float Y
-    {
-        get => _value.Y;
-        set => _value.Y = value;
-    }
-
-    public float Z
-    {
-        get => _value.Z;
-        set => _value.Z = value;
-    }
-
-    public Vector3 value
-    {
-        readonly get => _value;
-        set => _value = value;
-    }
-}
-
-//[ShortRunJob(RuntimeMoniker.Mono)]
-[ShortRunJob(RuntimeMoniker.Net90)]
-[ShortRunJob(RuntimeMoniker.NativeAot90)]
 public class RWBenchmarks
 {
     [Params(1_000)]
@@ -79,15 +39,14 @@ public void Setup()
     {
         _stream = new(count);
     }
-    
+
     [Benchmark]
     public void OldNOP()
     {
         _stream.Old((in Entity entity, [In] ref int a, [In] ref int b) =>
-        {
-        });
+            { });
     }
-    
+
     [Benchmark(Baseline = true)]
     public void OldWW()
     {
@@ -97,7 +56,7 @@ public void OldWW()
             b++;
         });
     }
-    
+
     [Benchmark]
     public void OldWRo()
     {
@@ -106,7 +65,7 @@ public void OldWRo()
             a += b;
         });
     }
-    
+
     [Benchmark]
     public void OldEWR()
     {
@@ -151,7 +110,7 @@ public void NewWR()
             a.write++;
         });
     }
-    
+
     [Benchmark]
     public void NewRW()
     {
@@ -162,35 +121,37 @@ public void NewRW()
     }
 }
 
-internal interface Fox<T>where T : notnull
+internal interface Fox<T> where T : notnull
 {
-   T value { get; set; }
+    T value { get; set; }
 }
 
-internal readonly record struct BenchStream2<C1, C2>(int Count) 
+internal readonly record struct BenchStream2<C1, C2>(int Count)
     where C1 : notnull
-    where C2 :  notnull 
+    where C2 : notnull
 {
     public readonly Entity[] entities = new Entity[Count];
     public readonly C1[] Data1 = new C1[Count];
     public readonly C2[] Data2 = new C2[Count];
-    
-    [MethodImpl(MethodImplOptions.AggressiveInlining)]
+
+
     public void New(ComponentActionEWR<C1, C2> action)
     {
         var match = default(TypeExpression);
         for (var i = 0; i < Count; i++) action(new(ref entities[i]), new(ref Data1[i], ref entities[i], ref match), new(ref Data2[i]));
     }
+
     public void Old(EntityComponentAction<C1, C2> action)
     {
         for (var i = 0; i < Count; i++) action(entities[i], ref Data1[i], ref Data2[i]);
     }
 
+
     public void Old2(ComponentActionRead<C1, C2> action)
     {
         for (var i = 0; i < Count; i++) action(in Data1[i], in Data2[i]);
     }
-    
+
     [OverloadResolutionPriority(0)]
     public void New(ComponentActionWW<C1, C2> action)
     {
@@ -223,7 +184,7 @@ public void New(ComponentActionWR<C1, C2> action)
             action(ref1, ref2);
         }
     }
-    
+
     [OverloadResolutionPriority(2)]
     public void NewRo(ComponentActionWR<C1, C2> action)
     {
@@ -232,10 +193,10 @@ public void NewRo(ComponentActionWR<C1, C2> action)
         {
             var ref1 = new RW<C1>(ref Data1[i], ref entities[i], ref match);
             var ref2 = new R<C2>(ref Data2[i]);
-            action(ref1,  ref2);
+            action(ref1, ref2);
         }
     }
-    
+
     [OverloadResolutionPriority(3)]
     public void New(ComponentActionRR<C1, C2> action)
     {
@@ -248,6 +209,7 @@ public void New(ComponentActionRR<C1, C2> action)
         }
     }
 
+
     public void Old(ComponentActionWRo<C1, C2> action)
     {
         for (var i = 0; i < Count; i++) action(ref Data1[i], ref Data2[i]);
@@ -256,10 +218,9 @@ public void Old(ComponentActionWRo<C1, C2> action)
 
 internal delegate void ComponentActionRead<C0, C1>(in C0 comp0, in C1 comp1);
 internal delegate void ComponentActionWRo<C0, C1>(ref C0 comp0, ref readonly C1 comp1);
-
 internal delegate void ComponentActionEWR<C0, C1>([In] EntityRef entity, RW<C0> comp0, R<C1> comp1) where C0 : notnull where C1 : notnull;
 internal delegate void ComponentActionWR<C0, C1>(RW<C0> comp0, R<C1> comp1) where C0 : notnull where C1 : notnull;
 internal delegate void ComponentActionRW<C0, C1>(R<C0> comp0, RW<C1> comp1) where C0 : notnull where C1 : notnull;
 internal delegate void ComponentActionWWref<C0, C1>(ref readonly RW<C0> comp0, ref readonly RW<C1> comp1) where C0 : notnull where C1 : notnull;
 internal delegate void ComponentActionWW<C0, C1>(RW<C0> comp0, RW<C1> comp1) where C0 : notnull where C1 : notnull;
-internal delegate void ComponentActionRR<C0, C1>(R<C0> comp0, R<C1> comp1) where C0 : notnull where C1 : notnull;
+internal delegate void ComponentActionRR<C0, C1>(R<C0> comp0, R<C1> comp1) where C0 : notnull where C1 : notnull;

fennecs.benchmarks/Conceptual/TensorPrimitivesBench.cs

@@ -0,0 +1,56 @@
+// Add a <PackageReference Include="System.Numerics.Tensors" Version="9.0.0" /> to the csproj.
+// dotnet run -c Release -f net9.0 --filter "*"
+
+using System.Numerics.Tensors;
+using BenchmarkDotNet.Attributes;
+
+namespace Benchmark.Conceptual;
+
+public class TensorPrimitivesBench
+{
+    private float[] _source, _destination;
+
+    [GlobalSetup]
+    public void Setup()
+    {
+        var r = new Random(42);
+        _source = Enumerable.Range(0, 1024).Select(_ => (float)r.NextSingle()).ToArray();
+        _destination = new float[1024];
+    }
+
+    [Benchmark(Baseline = true)]
+    public void ManualLoop()
+    {
+        ReadOnlySpan<float> source = _source;
+        Span<float> destination = _destination;
+        for (int i = 0; i < source.Length; i++)
+        {
+            destination[i] = MathF.PI + source[i];
+        }
+    }
+
+    [Benchmark]
+    public void UnrolledLoop()
+    {
+        Span<float> source = _source.AsSpan();
+        Span<float> destination = _destination.AsSpan();
+        for (int i = 0; i < source.Length; i+=8)
+        {
+            destination[i] = MathF.PI + source[i];
+            destination[i+1] = MathF.PI + source[i+1];
+            destination[i+2] = MathF.PI + source[i+2];
+            destination[i+3] = MathF.PI + source[i+3];
+            destination[i+4] = MathF.PI + source[i+4];
+            destination[i+5] = MathF.PI + source[i+5];
+            destination[i+6] = MathF.PI + source[i+6];
+            destination[i+7] = MathF.PI + source[i+7];
+        }
+    }
+
+    [Benchmark]
+    public void BuiltIn()
+    {
+        //TensorPrimitives.Cosh(_source, _destination);
+        TensorPrimitives.Multiply(_source, MathF.PI, _destination);
+    }
+}

fennecs.benchmarks/Program.cs

@@ -1,12 +1,19 @@
 using System.Runtime.Intrinsics.Arm;
 using System.Runtime.Intrinsics.X86;
+using Benchmark;
 using Benchmark.Conceptual;
-using Benchmark.ECS;
 using BenchmarkDotNet.Configs;
+using BenchmarkDotNet.Environments;
+using BenchmarkDotNet.Jobs;
 using BenchmarkDotNet.Running;
 
 
-var config = ManualConfig.Create(DefaultConfig.Instance).WithOptions(ConfigOptions.JoinSummary);
+var config = ManualConfig
+    .Create(DefaultConfig.Instance)
+    .WithOptions(ConfigOptions.JoinSummary)
+    .AddJob(Job.ShortRun.WithRuntime(CoreRuntime.Core90))
+    .AddJob(Job.ShortRun.WithRuntime(NativeAotRuntime.Net90))
+    .HideColumns("Job", "Error", "Median", "RatioSD");
 
 // Most relevant vectorization instruction sets, add other intrinsics as needed.
 // These are exclusions you can use to TURN OFF specific benchmarks based on the
@@ -17,5 +24,6 @@
 if (!Sse2.IsSupported) config.AddFilter(new CategoryExclusion(nameof(Sse2)));
 if (!AdvSimd.IsSupported) config.AddFilter(new CategoryExclusion(nameof(AdvSimd)));
 
+
 //BenchmarkRunner.Run<DorakuBenchmarks>(config);
 BenchmarkRunner.Run<RWBenchmarks>(config);

fennecs.benchmarks/fennecs.benchmarks.csproj

@@ -6,14 +6,15 @@
         <ImplicitUsings>enable</ImplicitUsings>
         <Nullable>enable</Nullable>
         <RootNamespace>Benchmark</RootNamespace>
-        <Configurations>Release;Debug</Configurations>
+        <Configurations>Release</Configurations>
         <Platforms>AnyCPU</Platforms>
         <IsPackable>false</IsPackable>
         <PackageId>fennecs.benchmarks</PackageId>
         <AllowUnsafeBlocks>true</AllowUnsafeBlocks>
         <CheckForOverflowUnderflow>false</CheckForOverflowUnderflow>
         <LangVersion>preview</LangVersion>
         <PublishAot>true</PublishAot>
+        <OptimizationPreference>Speed</OptimizationPreference>
     </PropertyGroup>
 
     <PropertyGroup Condition=" '$(Configuration)' == 'Debug' ">
@@ -31,5 +32,4 @@
     <ItemGroup>
         <PackageReference Include="BenchmarkDotNet" Version="0.14.1-nightly.20241107.194" />
     </ItemGroup>
-
 </Project>

fennecs.tests/Conceptual/BloomTests.cs

@@ -47,8 +47,6 @@ public void GetInterfaces()
     {
 
         output.WriteLine($"Hardware: {Vector256.IsHardwareAccelerated}");
-
-        Vector256<uint> bloom = default;
 
         var hash = Vector256<long>.Zero;