[llvm-bugs] [Bug 49594] New: weird codegen: xor eax, eax; test al, al; jne

[via llvm-bugs]

https://bugs.llvm.org/show_bug.cgi?id=49594

            Bug ID: 49594
           Summary: weird codegen: xor eax, eax; test al, al; jne
           Product: clang
           Version: 11.0
          Hardware: PC
                OS: Windows NT
            Status: NEW
          Severity: enhancement
          Priority: P
         Component: -New Bugs
          Assignee: unassignedclangbugs at nondot.org
          Reporter: redbeard0531 at gmail.com
                CC: htmldeveloper at gmail.com, llvm-bugs at lists.llvm.org,
                    neeilans at live.com, richard-llvm at metafoo.co.uk

https://godbolt.org/z/nsGj55

I noticed this while trying to minrepro another codegen issue, so the input
code is a bit odd. I can't think of any reason to ever generate this
instruction sequence, since the value of al is well known after zeroing it. It
also seems odd that it is doing anything with eax/al at all here, given that it
is clobbering it a few instructions later.

#include <cassert>
#include <cstdint>

template<int V>  int f(int = V);

int test1(const char* p) {
    switch(auto byte = uint8_t(*p)) {
        case 0: return f<0>();
        case 1: return f<1>();
        case 2: return f<2>();
        case 3 ... 10: return f<3>(byte);
        case 11 ... 100: return f<4>(byte);
        case 111 ... 120: return f<5>(byte);
        case 121 ... 130: return f<6>(byte);
        case 131 ... 140: return f<7>(byte);
        case 141 ... 150: return f<8>(byte);
        case 151 ... 160: return f<9>(byte);
        case 161 ... 170: return f<10>(byte);
        case 171 ... 180: return f<5>(byte);
        case 181 ... 190: return f<5>(byte);
        case 191 ... 200: return f<5>(byte);
        default: assert(false);
    }
}

test1(char const*):                            # @test1(char const*)
        push    rax
        movzx   edi, byte ptr [rdi]
        xor     eax, eax  # <-- zeros eax, including al
        test    al, al    # <-- why ?!?!?!?
        jne     .LBB0_6   # <-- never taken
        lea     eax, [rdi - 128]
        movzx   eax, al
        jmp     qword ptr [8*rax + .LJTI0_0]
        ...
(snipping rest since the interesting bit is above)

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20210315/8da25a98/attachment.html>