[Bug tree-optimization/96603] New: Failure to optimize memcmp out for more than one byte

[gabravier at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=96603

            Bug ID: 96603
           Summary: Failure to optimize memcmp out for more than one byte
           Product: gcc
           Version: 11.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: gabravier at gmail dot com
  Target Milestone: ---

int f(const void *s1, const void *s2)
{
    return memcmp(s1, s2, 2);
}

This can be optimized to `loadbe16(s1) - loadbe16(s2)` (where loadbe16 is a
function that read a 16-bit big endian number). This transformation is done by
LLVM, but not by GCC.