Apply hardware intrinsics to BitArray.*Shift

[kzrnm]

Related #63722, #113005

benchmark

BenchmarkDotNet v0.14.0, Windows 11 (10.0.26100.3194)
13th Gen Intel Core i5-13500, 1 CPU, 20 logical and 14 physical cores
.NET SDK 10.0.100-preview.3.25125.5
  [Host]     : .NET 10.0.0 (10.0.25.12411), X64 RyuJIT AVX2
  Job-UQJRRD : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX2
  Job-IYICYK : .NET 10.0.0 (42.42.42.42424), X64 RyuJIT AVX2

Method	Toolchain	N	Mean	Ratio	Allocated	Alloc Ratio
LeftShift	\main\corerun.exe	256	45.553 ns	1.00	-	NA
LeftShift	\pr\corerun.exe	256	10.037 ns	0.22	-	NA
RightShift	\main\corerun.exe	256	56.281 ns	1.00	-	NA
RightShift	\pr\corerun.exe	256	9.565 ns	0.17	-	NA
LeftShift	\main\corerun.exe	65536	10,493.384 ns	1.00	-	NA
LeftShift	\pr\corerun.exe	65536	1,699.862 ns	0.16	-	NA
RightShift	\main\corerun.exe	65536	11,729.467 ns	1.00	-	NA
RightShift	\pr\corerun.exe	65536	1,860.553 ns	0.16	-	NA
LeftShift	\main\corerun.exe	16777216	3,054,162.161 ns	1.00	-	NA
LeftShift	\pr\corerun.exe	16777216	1,051,349.065 ns	0.34	-	NA
RightShift	\main\corerun.exe	16777216	3,341,553.776 ns	1.00	-	NA
RightShift	\pr\corerun.exe	16777216	639,720.483 ns	0.19	-	NA

Benchmark code

using BenchmarkDotNet.Attributes;
using BenchmarkDotNet.Configs;
using System.Collections;

[MemoryDiagnoser(false)]
[HideColumns("Job", "Error", "StdDev", "Median", "RatioSD")]
public class ShiftTest
{
    [Params([
        1 << 8,
        1 << 16,
        1 << 24,
    ])]
    public int N;
    const int shiftSize = 13;
    BitArray b;

    [GlobalSetup]
    public void Setup()
    {
        var bytes = new byte[N];
        new Random(227).NextBytes(bytes);
        b = new BitArray(bytes);
    }

    [Benchmark] public BitArray LeftShift() => b.LeftShift(shiftSize);
    [Benchmark] public BitArray RightShift() => b.RightShift(shiftSize);
}

[eiriktsarpalis]

Hi @kzrnm, apologies for the delay in reviewing this. Is there a chance you could rebase your changes on top of the latest main? I don't have write permissions in your PR branch to do it myself.

[kzrnm]

I updated to match the latest implementation.
In commit e96a2b8, the implementation for shifting as a 32-bit integer has been updated to apply a mask within the byte range, aligning it with the behavior of the Apply method.

[Copilot]

Pull Request Overview

This PR optimizes the BitArray.LeftShift and BitArray.RightShift methods by leveraging hardware intrinsics (Vector512, Vector256, Vector128) to achieve significant performance improvements. The benchmark results show 3-6x speedup across different array sizes.

Key Changes:

• Replaced int-based shifting with byte-based operations using SIMD vectors
• Introduced vectorized processing for bulk shift operations with fallback to scalar code
• Added comprehensive test coverage for various bit array sizes and shift amounts

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
BitArray.cs	Refactored RightShift and LeftShift to use byte-level operations with SIMD intrinsics (Vector512/256/128), replacing the previous int-based implementation
BitArray_OperatorsTests.cs	Enhanced test coverage with additional test cases covering various sizes (1023-1025) and shift amounts, plus validation of high-order bit clearing