Add lots more TensorPrimitives operations (#97192)

* Add more TensorPrimitive operations

All are functional and tested, some are vectorized, others still need to be.

* Re-fix IndexOfXx operations

* Rename MultiplyAddEstimate to FusedMultiplyAdd

* Remove some NotSupportedException throws

* Simplify CopySignOperator

* Parameter renames

* Add scalar overloads of Atan2 and Atan2Pi

* Add CopySign scalar overload

* Add Ieee754Remainder scalar overloads

* Add Lerp scalar overloads

* Add Pow scalar overload

* Consolidate inverted operators

* Add missing Max/Min scalar overloads

* Use ElementWiseSelect

Author: stephentoub

src/libraries/System.Numerics.Tensors/ref/System.Numerics.Tensors.netcore.cs

@@ -303,7 +303,7 @@ public static void Exp(ReadOnlySpan<float> x, Span<float> destination) =>
         /// </para>
         /// </remarks>
         public static int IndexOfMax(ReadOnlySpan<float> x) =>
-            IndexOfMinMaxCore<IndexOfMaxOperator_Single>(x);
+            IndexOfMinMaxCore<float, IndexOfMaxOperator_Single>(x);
 
         /// <summary>Searches for the index of the single-precision floating-point number with the largest magnitude in the specified tensor.</summary>
         /// <param name="x">The tensor, represented as a span.</param>
@@ -320,7 +320,7 @@ public static int IndexOfMax(ReadOnlySpan<float> x) =>
         /// </para>
         /// </remarks>
         public static int IndexOfMaxMagnitude(ReadOnlySpan<float> x) =>
-            IndexOfMinMaxCore<IndexOfMaxMagnitudeOperator_Single>(x);
+            IndexOfMinMaxCore<float, IndexOfMaxMagnitudeOperator_Single>(x);
 
         /// <summary>Searches for the index of the smallest single-precision floating-point number in the specified tensor.</summary>
         /// <param name="x">The tensor, represented as a span.</param>
@@ -336,7 +336,7 @@ public static int IndexOfMaxMagnitude(ReadOnlySpan<float> x) =>
         /// </para>
         /// </remarks>
         public static int IndexOfMin(ReadOnlySpan<float> x) =>
-            IndexOfMinMaxCore<IndexOfMinOperator_Single>(x);
+            IndexOfMinMaxCore<float, IndexOfMinOperator_Single>(x);
 
         /// <summary>Searches for the index of the single-precision floating-point number with the smallest magnitude in the specified tensor.</summary>
         /// <param name="x">The tensor, represented as a span.</param>
@@ -353,7 +353,7 @@ public static int IndexOfMin(ReadOnlySpan<float> x) =>
         /// </para>
         /// </remarks>
         public static int IndexOfMinMagnitude(ReadOnlySpan<float> x) =>
-            IndexOfMinMaxCore<IndexOfMinMagnitudeOperator_Single>(x);
+            IndexOfMinMaxCore<float, IndexOfMinMagnitudeOperator_Single>(x);
 
         /// <summary>Computes the element-wise natural (base <c>e</c>) logarithm of single-precision floating-point numbers in the specified tensor.</summary>
         /// <param name="x">The tensor, represented as a span.</param>

src/libraries/System.Numerics.Tensors/src/System/Numerics/Tensors/TensorPrimitives.Single.cs

@@ -1,7 +1,7 @@
 // Licensed to the .NET Foundation under one or more agreements.
 // The .NET Foundation licenses this file to you under the MIT license.
 
-// This file exists to enable TensorPrimitives.float.cs to be compiled for both
+// This file exists to enable TensorPrimitives.Single.cs to be compiled for both
 // netstandard2.0 and net8.0+ targets. It uses the XX_Single names and the operation
 // methods tied to float, whereas the net8.0+ worker implementations use generic math.
 // This file provides float-bound types and type defs that route one to the other.
@@ -14,6 +14,10 @@
 global using DivideOperator_Single = System.Numerics.Tensors.TensorPrimitives.DivideOperator<float>;
 global using MultiplyOperator_Single = System.Numerics.Tensors.TensorPrimitives.MultiplyOperator<float>;
 global using ExpOperator_Single = System.Numerics.Tensors.TensorPrimitives.ExpOperator<float>;
+global using IndexOfMaxOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMaxOperator<float>;
+global using IndexOfMaxMagnitudeOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMaxMagnitudeOperator<float>;
+global using IndexOfMinOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMinOperator<float>;
+global using IndexOfMinMagnitudeOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMinMagnitudeOperator<float>;
 global using LogOperator_Single = System.Numerics.Tensors.TensorPrimitives.LogOperator<float>;
 global using Log2Operator_Single = System.Numerics.Tensors.TensorPrimitives.Log2Operator<float>;
 global using MaxOperator_Single = System.Numerics.Tensors.TensorPrimitives.MaxOperator<float>;
@@ -33,12 +37,6 @@
 global using SquaredOperator_Single = System.Numerics.Tensors.TensorPrimitives.SquaredOperator<float>;
 global using TanhOperator_Single = System.Numerics.Tensors.TensorPrimitives.TanhOperator<float>;
 
-// TODO: These should be made generic. Their implementations are still currently bound to float.
-global using IndexOfMaxOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMaxOperator;
-global using IndexOfMaxMagnitudeOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMaxMagnitudeOperator;
-global using IndexOfMinOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMinOperator;
-global using IndexOfMinMagnitudeOperator_Single = System.Numerics.Tensors.TensorPrimitives.IndexOfMinMagnitudeOperator;
-
 namespace System.Numerics.Tensors
 {
     public static unsafe partial class TensorPrimitives