Extra push/pop ebp in code gen

[iSazonov]

See SharpLab

public class C {
    private static bool _isC0Code(char wch)
        => wch < 0x20 && wch is not (char)0x18 and not (char)0x1a and not (char)0x1b;
}

C._isC0Code(Char)
    L0000: push ebp
    L0001: mov ebp, esp
    L0003: movzx eax, cx
    L0006: cmp eax, 0x20
    L0009: jge short L0020
    L000b: cmp eax, 0x18
    L000e: je short L0020
    L0010: cmp eax, 0x1a
    L0013: je short L0020
    L0015: cmp eax, 0x1b
    L0018: setne al
    L001b: movzx eax, al
    L001e: pop ebp
    L001f: ret
    L0020: xor eax, eax
    L0022: pop ebp
    L0023: ret

[EgorBo]

@iSazonov this is 32-bit codegen in your link

[iSazonov]

@EgorBo Yes, thanks for the clarification. Nevertheless, it's still superfluous.

[EgorBo]

@EgorBo Yes, thanks for the clarification. Nevertheless, it's still superfluous.

I don't see a problem with the frame being setup, in this case it's triggered by an internal heuristic driven by number of blocks (5 blocks in 32-bit codegen, 4 blocks in 64-bit codegen). Maybe we could check why that number is different between the architectures, but I don't think 32-bit codegen is worth the effort.

; Decided to create an EBP based frame for ETW stackwalking (BasicBlock Count)