JIT: Fix profiled memcmp specialization for inline candidates

[EgorBo]

Since #134160 led to 69 benchmarks improved, I decided to do the same for memcmp idiom.

Benchmark

using BenchmarkDotNet.Attributes;

public class Bench
{
    private byte[] _bytes1 = null!;
    private byte[] _bytes2 = null!;

    [Params(1, 2, 5, 8, 16, 20, 32, 50, 64, 128)]
    public int Length { get; set; }

    [GlobalSetup]
    public void Setup()
    {
        _bytes1 = new byte[Length];
        _bytes2 = new byte[Length];
    }

    [Benchmark]
    public bool Bytes() => _bytes1.AsSpan().SequenceEqual(_bytes2);
}

Results: EgorBot/Benchmarks#612 (arm64 only handles up to 32 bytes today).

[azure-pipelines]

Azure Pipelines:
Successfully started running 5 pipeline(s).
11 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[dotnet-policy-service]

Tagging subscribers to this area: @JulieLeeMSFT, @jakobbotsch
See info in area-owners.md if you want to be subscribed.

[Copilot]

Copilot review overview

🟡 Changes recommended

Address the fallback-threshold regression and add TieredPGO regression coverage.

Get a fresh assessment by requesting another Copilot review.

Review effort: Lite
Findings: 1

Open (1)

• Preserve non-SIMD memcmp unroll thresholds on non-AMD64/ARM64 targets · New

Resolved since last review (1)

• Add PGO regression test for profiled inline-candidate SequenceEqual

[EgorBo]

@EgorBot -macos_arm -linux_amd -linux_arm

using BenchmarkDotNet.Attributes;

public class Bench
{
    private byte[] _bytes1 = null!;
    private byte[] _bytes2 = null!;

    [Params(1, 2, 5, 8, 16, 20, 32, 50, 64, 128)]
    public int Length { get; set; }

    [GlobalSetup]
    public void Setup()
    {
        _bytes1 = new byte[Length];
        _bytes2 = new byte[Length];
    }

    [Benchmark]
    public bool Bytes() => _bytes1.AsSpan().SequenceEqual(_bytes2);
}

[EgorBo]

PTAL @AndyAyersMS follow up to #134160 with improvements for SequenceEqual this time.

[AndyAyersMS]

I wonder how often this actually kicks in for realistic code. Do we have any benchmarks that run a mixture of lengths? Have you considered multi-guess for things like this?

[EgorBo]

I wonder how often this actually kicks in for realistic code. Do we have any benchmarks that run a mixture of lengths? Have you considered multi-guess for things like this?

I expect it to be pretty common. It's basically what all string comparisons/startswith/endswith call under the hood. Because we can instrument inlined code now, this one actually manages to keep profile per call-site (acts as context-oriented).

Basically, SequenceEqual is a work horse for all comparisons of arrays/spans of anything (blittable)

[EgorBo]

/ba-g osx-arm64 stuck