Remove unsafe code from IPAddress

[EgorBo]

Contributes to #94941

The safer patterns seem to show better codegen in fact + removed an allocation from MapToIPv6

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/ncl
See info in area-owners.md if you want to be subscribed.

[EgorBo]

@EgorBot -amd -arm

using System.Net;
using BenchmarkDotNet.Attributes;

[MemoryDiagnoser]
public class IPAddressBenchmarks
{
    static IPAddress Ipv6_1 = IPAddress.Parse("3cca:65e1:957d:e712:1c8c:ec01:6d44:febc");
    static IPAddress Ipv6_2 = IPAddress.Parse("3cca:65e1:957d:e712:1c8c:ec01:6d44:febc");
    static IPAddress Ipv6_3 = IPAddress.Parse("17e1:891a:1a00:d475:92cd:d4c7:5b30:775a");
    static IPAddress Ipv6_4 = IPAddress.Parse("33.55.77.99").MapToIPv6();
    static IPAddress Ipv4_1 = IPAddress.Parse("33.55.77.99");
    static byte[] _buffer = new byte[16];

    [Benchmark]
    public bool CompareSameIPv6() => Ipv6_1.Equals(Ipv6_2);

    [Benchmark]
    public bool CompareDifferentIPv6() => Ipv6_1.Equals(Ipv6_3);

    [Benchmark]
    public IPAddress MapToIPv6() => Ipv4_1.MapToIPv6();

    [Benchmark]
    public bool IsIPv4MappedToIPv6() => Ipv6_4.IsIPv4MappedToIPv6;

    [Benchmark]
    public bool TryWriteBytes_v6() => Ipv6_1.TryWriteBytes(_buffer, out _);

    [Benchmark]
    public IPAddress NewIPAddressv6() => new("1111222233334444"u8);

    [Benchmark]
    public IPAddress ParseIPAddressv6() => IPAddress.Parse("3cca:65e1:957d:e712:1c8c:ec01:6d44:febc");
}

[MihaZupan]

Thanks

[EgorBo]

/ba-g "Build browser-wasm linux Release LibraryTests stuck"

[quixoticaxis]

Does Debug.Assert have a side effect on JIT decisions? Even in Release builds? Is it mentioned somewhere or is it not guaranteed? What if the assertion no longer holds in Release?

[EgorBo]

Does Debug.Assert have a side effect on JIT decisions? Even in Release builds? Is it mentioned somewhere or is it not guaranteed? What if the assertion no longer holds in Release?

Debug.Assert doesn't exist in Release code, it's Debug only

[quixoticaxis]

Thank you for clarification. I misunderstood your code comment, and thus the question arose.

[stephentoub]

Instead of:

SequenceEqual((ReadOnlySpan<ushort>)[0, 0, 0, 0, 0, 0xFFFF])

could it be:

SequenceEqual<ushort>([0, 0, 0, 0, 0, 0xFFFF])

?

[EgorBo]

good point!