Use span copy in BigInteger.TryGetBytes when applicable

[Rob-Hague]

When the value is positive (we don't need to do two's complement conversion) and on a little endian machine, we can copy the bytes out from the underlying uint[] storage.

Method	Job	BitLength	IsBigEndian	Mean	Error	StdDev	Ratio
ToByteArray	PR	1024	False	21.64 ns	0.020 ns	0.015 ns	0.41
ToByteArray	main	1024	False	53.38 ns	0.157 ns	0.140 ns	1.00
ToByteArray	PR	1024	True	23.45 ns	0.021 ns	0.018 ns	0.44
ToByteArray	main	1024	True	53.48 ns	0.150 ns	0.125 ns	1.00

public class Benchmarks
{
    [Params(1024)]
    public int BitLength;

    [Params(true, false)]
    public bool IsBigEndian;

    [Params(true)]
    public bool IsPositive;

    private BigInteger n;

    [GlobalSetup]
    public void Setup()
    {
        Random r = new(123);
        byte[] bytes = new byte[BitLength / 8];
        r.NextBytes(bytes);
        if (IsPositive)
        {
            bytes[^1] &= 0x7F;
        }
        else
        {
            bytes[^1] |= 0x80;
        }
        n = new BigInteger(bytes);
    }

    [Benchmark] public byte[] ToByteArray() => n.ToByteArray(isBigEndian: IsBigEndian);
}

[huoyaoyuan]

The whole method is not as efficient as it can. There are multiple problems, for example BitConverter.IsLittleEndian is an intrinsic so it shouldn't be passed as a parameter.

A bottom-up rewrite is better.

[dotnet-policy-service]

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

[Rob-Hague]

@huoyaoyuan isBigEndian is a parameter passed to ToByteArray

[huoyaoyuan]

@huoyaoyuan isBigEndian is a parameter passed to ToByteArray

Yes, you are right, I misread it as the same of BitConverter.IsLittleEndian.

[stephentoub]

Thanks.