Replace .Clone() with explicit array allocation and copy in Arrays

[paulomorgado]

Replace .Clone() with explicit array allocation and copy in Arrays

Summary

Refactors all array cloning methods in crypto/src/util/Arrays.cs to replace the (T[])array.Clone() pattern with explicit allocation and copy. This gives the runtime more precise information about the operation, avoids unnecessary zero-initialization overhead on .NET 5.0+, and uses the most efficient copy path for each element type.

Describe your changes

• Clone(T[] data) overloads (bool, byte, short, ushort, int, uint, long, ulong): replaced (T[])data.Clone() with GC.AllocateUninitializedArray<T> (on .NET 5.0+, falling back to new T[] on older targets) followed by Buffer.BlockCopy, which operates on raw bytes and is highly optimized for blittable/primitive types.
• InternalCopyBuffer<T>: replaced (T[])buf.Clone() with GC.AllocateUninitializedArray<T> + Array.Copy, since the generic constraint doesn't guarantee a blittable type suited for Buffer.BlockCopy.

Benchmarks

[MemoryDiagnoser]
[SimpleJob(RuntimeMoniker.Net10_0)]
[SimpleJob(RuntimeMoniker.Net481)]
[HideColumns(Column.Error, Column.StdDev, Column.Median, Column.Job)]
public class CloneImplementationComparison
{
    [Params(16, 128, 1024)]
    public int ArraySize;

    private byte[] sourceArray;

    [GlobalSetup]
    public void Setup()
    {
        sourceArray = new byte[ArraySize];
        for (int i = 0; i < ArraySize; i++)
            sourceArray[i] = (byte)(i & 0xFF);
    }

    [Benchmark(Baseline = true)]
    public byte[] ArrayClone()
    {
        return (byte[])sourceArray.Clone();
    }

    [Benchmark]
    public byte[] NewArrayBlockCopy()
    {
#if NET5_0_OR_GREATER
        byte[] temp = GC.AllocateUninitializedArray<byte>(sourceArray.Length);
#else
        byte[] temp = new byte[sourceArray.Length];
#endif
        Buffer.BlockCopy(sourceArray, 0, temp, 0, sourceArray.Length);
        return temp;
    }
}

Benchmark results

Benchmarked on: Windows 11, 13th Gen Intel Core i9-13900K, .NET 10.0.3 / .NET Framework 4.8.1.

BenchmarkDotNet v0.15.8, Windows 11 (10.0.26200.7918/25H2/2025Update/HudsonValley2)
13th Gen Intel Core i9-13900K 3.00GHz, 1 CPU, 32 logical and 24 physical cores
.NET SDK 10.0.200-preview.0.26103.119
  [Host]               : .NET 10.0.3 (10.0.3, 10.0.326.7603), X64 RyuJIT x86-64-v3
  .NET 10.0            : .NET 10.0.3 (10.0.3, 10.0.326.7603), X64 RyuJIT x86-64-v3
  .NET Framework 4.8.1 : .NET Framework 4.8.1 (4.8.9325.0), X64 RyuJIT VectorSize=256

Method	Runtime	ArraySize	Mean	Ratio	Allocated
ArrayClone	.NET 10.0	16	33.286 ns	1.00	40 B
NewArrayBlockCopy	.NET 10.0	16	4.787 ns	0.14	40 B
ArrayClone	.NET Framework 4.8.1	16	18.655 ns	1.00	40 B
NewArrayBlockCopy	.NET Framework 4.8.1	16	7.217 ns	0.39	40 B
ArrayClone	128	33.653 ns	1.00	152 B
NewArrayBlockCopy	.NET 10.0	128	7.731 ns	0.23	152 B
ArrayClone	.NET Framework 4.8.1	128	22.591 ns	1.00	152 B
NewArrayBlockCopy	.NET Framework 4.8.1	128	11.276 ns	0.50	152 B
ArrayClone	.NET 10.0	1024	50.312 ns	1.00	1048 B
NewArrayBlockCopy	.NET 10.0	1024	33.917 ns	0.67	1048 B
ArrayClone	.NET Framework 4.8.1	1024	44.428 ns	1.00	1051 B
NewArrayBlockCopy	.NET Framework 4.8.1	1024	34.993 ns	0.79	1051 B

The new approach is 7× faster for small arrays (16 elements) on .NET 10.0, and consistently 20–50% faster across all sizes on .NET Framework 4.8.1. Memory allocation is identical. No behavioral change — null inputs continue to return null.

How has this been tested?

Using the existing test.

Checklist before requesting a review

• I have performed a self-review of my code
• I have kept the patch limited to only change the parts related to the patch
• This change requires a documentation update

See also Contributing Guidelines.