Remove span pinning associated with string.Create

[stephentoub]

This is now possible with C# 11.

[dotnet-issue-labeler]

I couldn't figure out the best area label to add to this PR. If you have write-permissions please help me learn by adding exactly one area label.

[jakobbotsch]

Keep in mind that this pattern introduces new address exposure, so it might be worth it to double check that it does not impact the preceding uses of the span in the same method significantly. If you want to be on the safe side you can create a copy of the span.

For example

[MethodImpl(MethodImplOptions.NoInlining)]
private static unsafe string Foo(ReadOnlySpan<char> chars)
{
    for (int i = 0; i < chars.Length; i++)
    {
        if (chars[i] >= 128)
            return null;
    }

    //return string.Create(chars.Length, (IntPtr)(&chars), static (s, c) =>
    //{
    //    for (int j = 0; j < s.Length; j++)
    //        s[j] = (*(ReadOnlySpan<char>*)c)[j];
    //});
    fixed (char* charsPtr = chars)
    {
        return string.Create(chars.Length, (IntPtr)charsPtr, static (s, c) =>
        {
            for (int j = 0; j < s.Length; j++)
                s[j] = ((char*)c)[j];
        });
    }
}

Codegen for the loop with the uncommented vs commented code on linux-x64:

 G_M27558_IG03:
-       mov      eax, esi
-       cmp      word  ptr [rdi+2*rax], 128
-       jae      G_M27558_IG09
-       inc      esi
-       cmp      esi, ebx
+       mov      rsi, bword ptr [rbp-28H]
+       mov      eax, edi
+       cmp      word  ptr [rsi+2*rax], 128
+       jae      G_M27558_IG07
+       inc      edi
+       cmp      edi, dword ptr [rbp-20H]
        jl       SHORT G_M27558_IG03

(Fixing this would be the "partial lifetime promotion" item in #76931)

[stephentoub]

Thanks.

Codegen for the loop with the uncommented vs commented code on linux-x64:

Ugh, it reintroduces bounds checking on the prior loop? That's not something we can fix?

SharpLab

[jakobbotsch]

Ugh, it reintroduces bounds checking on the prior loop? That's not something we can fix?

Not easily. Address exposure turns off lots of basic analyses and optimizations for the exposed local and we're very coarse grained about it today.

[stephentoub]

@jakobbotsch, does it look ok now? Thanks.