[Bug c/103647] New: constant array comparison not always folded

[Bug c/103647] New: constant array comparison not always folded

[denis.campredon at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103647

            Bug ID: 103647
           Summary: constant array comparison not always folded
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
          Assignee: unassigned at gcc dot gnu.org
          Reporter: denis.campredon at gmail dot com
  Target Milestone: ---

On trunk and gcc 11, with -O2 the following functions are not optimized to 
xor eax, eax

------------
bool f() {
    char a[] = {'a', 'c'};
    char b[] = {'a', 'b'};
    return __builtin_memcmp(a, b, 2);
}

bool g() {
    char a[] = {'a', 'b'};
    char b[] = {'a', 'b', 'c'};
    return __builtin_memcmp(a, b, 2);
}
------------

.LC1:
        .string "ab"
.LC0:
        .string "ac"
f():
        movzx   eax, WORD PTR .LC1[rip]
        cmp     WORD PTR .LC0[rip], ax
        setne   al
        ret
g():
        cmp     WORD PTR .LC1[rip], 25185
        setne   al
        ret

[Bug tree-optimization/103647] constant array comparison not always folded

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103647

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
well f should be optimized to return true while g should be optimized to return
false.

[Bug tree-optimization/103647] constant array comparison not always folded

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103647

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2021-12-10
           Severity|normal                      |enhancement
     Ever confirmed|0                           |1

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
So in fab GCC creates:
  a = "ac";
  b = "ab";
  _8 = MEM <unsigned short> [(char * {ref-all})&a];
  _9 = MEM <unsigned short> [(char * {ref-all})&b];
  _10 = _8 != _9;

And:
  a = "ab";
  b = "abc";
  _8 = MEM <unsigned short> [(char * {ref-all})&a];
  _9 = MEM <unsigned short> [(char * {ref-all})&b];
  _10 = _8 != _9;

But we don't have a FRE/DOM to optimize those MEM_REFs.

Confirmed.

[Bug tree-optimization/103647] constant array comparison not always folded

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=103647

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
fab is a bit late for this part of memcmp folding I guess, I'd have expected
that to happen in generic folding like when we inline memcpy.  It seems it's
actually the strlen pass that inlines the compare.