[Bug tree-optimization/119155] New: Aligned vector element accesses emitted for packed struct access

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119155

            Bug ID: 119155
           Summary: Aligned vector element accesses emitted for packed
                    struct access
           Product: gcc
           Version: 15.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rguenth at gcc dot gnu.org
  Target Milestone: ---

struct s { int x; } __attribute__((packed));

void f (char *xc, char *yc, int z)
{
  for (int i = 0; i < 100; ++i)
    {
      struct s *x = (struct s *) xc;
      struct s *y = (struct s *) yc;
      x->x += y->x;
      xc += z;
      yc += z;
    }
}

shows we emit

  _52 = __MEM <int> ((int *)_50);
  _53 = _50 + _4;
  _54 = __MEM <int> ((int *)_53);
  _55 = _53 + _4;
  _56 = __MEM <int> ((int *)_55);
  _57 = _55 + _4;
  _58 = __MEM <int> ((int *)_57);
  _59 = _Literal (int [[gnu::vector_size(16)]]) {_52, _54, _56, _58};

instead of byte-aligned int loads with -O3 -fno-vect-cost-model when
doing VMAT_STRIDED_SLP or VMAT_ELEMENTWISE, likely because we
represent alignment info and not check the DRs base alignment when
interpreting "unknown misalignment".