JIT: Optimize bit-wise AND with a constant mask in combination with a left shift in a compare

[varelen]

This resolves #111554.

C# code (example code from the issue):

[MethodImpl(MethodImplOptions.NoInlining)]
public static bool CheckLetterOrDigitConst(int uc)
    => (LetterOrDigitCategories & (1 << uc)) != 0;

x64 (before):

mov      eax, 1
shlx     eax, eax, ecx
test     eax, 287
setne    al
movzx    rax, al

x64 (after):

mov      eax, 287
bt       eax, edi
setb     al
movzx    rax, al

[varelen]

@dotnet-policy-service agree

[jakobbotsch]

LGTM. Thanks!

[EgorBo]

@jakobbotsch should this be merged?

[jakobbotsch]

Yes, I was waiting for CI and it fell off my radar.
@varelen For future PRs, just FYI, we do not require the PR branch to be up to date for merging, so no need to press that "Update branch" button. Things move too quickly in this repo to always have PRs up to date, and pressing it makes CI run again.

Thanks again!