[Bug tree-optimization/101641] Bogus redundant store removal

["rguenth at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101641

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Wow, and this time it's even combine coming into play!

(insn 10 9 11 2 (set (reg/v:DI 82 [ xy ])
        (mem/j:DI (reg/v/f:DI 86 [ pu ]) [2 pu_6(D)->y+0 S8 A64])) "t.i":12:8
76 {*movdi_internal}
     (nil))
(insn 11 10 12 2 (set (mem/j:DI (reg/v/f:DI 86 [ pu ]) [2 pu_6(D)->x+0 S8 A64])
        (reg/v:DI 82 [ xy ])) "t.i":13:9 76 {*movdi_internal}
     (expr_list:REG_DEAD (reg/v/f:DI 86 [ pu ])
        (expr_list:REG_DEAD (reg/v:DI 82 [ xy ])
            (nil))))

Trying 10 -> 11:
   10: r82:DI=[r86:DI]
   11: [r86:DI]=r82:DI
      REG_DEAD r86:DI
      REG_DEAD r82:DI
Failed to match this instruction:
(set (mem/j:DI (reg/v/f:DI 86 [ pu ]) [2 pu_6(D)->x+0 S8 A64])
    (mem/j:DI (reg/v/f:DI 86 [ pu ]) [2 pu_6(D)->y+0 S8 A64]))
allowing combination of insns 10 and 11
original costs 4 + 4 = 8
replacement cost 4
deferring deletion of insn with uid = 10.
modifying insn i3    11: [r86:DI]=[r86:DI]
      REG_DEAD r86:DI
deferring rescan insn with uid = 11.

Trying 4 -> 11:
    4: r86:DI=r91:DI
      REG_DEAD r91:DI
   11: [r86:DI]=[r86:DI]
      REG_DEAD r86:DI
Failed to match this instruction:
(set (mem/j:DI (reg:DI 91) [2 pu_6(D)->x+0 S8 A64])
    (mem/j:DI (reg:DI 91) [2 pu_6(D)->y+0 S8 A64]))
allowing combination of insns 4 and 11
original costs 4 + 4 = 8
replacement cost 4
deferring deletion of insn with uid = 4.
modifying insn i3    11: [r91:DI]=[r91:DI]
      REG_DEAD r91:DI
deferring rescan insn with uid = 11.
deleting noop move 11


somewhere inside combine we'd have to realize that this isn't a noop move
and then maybe not allow the combination in the first place since it
isn't recognizable?  That is, somehow we must anticipate the removal,
I suppose it is via

  /* Recognize all noop sets, these will be killed by followup pass.  */
  if (insn_code_number < 0 && GET_CODE (pat) == SET && set_noop_p (pat))
    insn_code_number = NOOP_MOVE_INSN_CODE, num_clobbers_to_add = 0;

where set_noop_p for two MEMs simply dispatches to
rtx_equal_p && !side_effects_p.

Note on RTL we see that we cannot rely on MEM_ALIAS_SET but have to
use MEM_EXPR to conservatively assess that the access is _not_ through
a union ... (or as said we could annotate the alias set entry as to
belonging to a union).


In the end how we handle TBAA and unions might not be the very best way
(but I can't offer something better yet).