[Bug tree-optimization/127968] [16/17 Regression] wrong code at -O1, -O2 and -Os with partially overlapping memcpy in a loop since r16-5526

[pinskia at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127968

--- Comment #2 from Drea Pinski <pinskia at gcc dot gnu.org> ---
Shouldn't:

  return (operand_equal_p (load_ref->mem.base, store_ref->mem.base, 0)
          && known_eq (load_ref->mem.size, store_ref->mem.size)
          && known_eq (load_ref->mem.offset, store_ref->mem.offset));

Be a range check to check for partial overlapping first and then an exact
match?

That is:

if ([offset1, offset1+size1] overlaps [offset2, offset2+size2])

then it is not safe to hoist