[Bug tree-optimization/124667] Optimize load/store IOR|AND branch mispredict (former bitmap_set_bit from GCC benchmark)

[daniel.barboza at oss dot qualcomm.com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124667

--- Comment #3 from Daniel Henrique Barboza <daniel.barboza at oss dot qualcomm.com> ---
(In reply to Andrew Pinski from comment #1)
>       bool res = (ptr->bits[word_num] & bit_val) == 0;
>       if (res)
>         ptr->bits[word_num] |= bit_val;
> 
> Converting this into:
> ptr->bits[word_num] |= bit_val;
> 
> is not valid with the C11/C++11 threading model so it needs a
> `!flag_store_data_races` check. And The only other problem is if this is
> read only memory then doing the store always could cause issues too.

Right.  I see these constraints are already implemented in
cond_store_replacement() (ref_can_have_store_data_races() checks for
flag_store_data_races):



  /* Prove that we can move the store down.  We could also check
     TREE_THIS_NOTRAP here, but in that case we also could move stores,
     whose value is not available readily, which we want to avoid.  */
  if (!nontrap->contains (lhs))
    {
      /* If LHS is an access to a local variable without address-taken
         (or when we allow data races) and known not to trap, we could
         always safely move down the store.  */
      tree base;
      if (ref_can_have_store_data_races (lhs)
          || tree_could_trap_p (lhs)
          /* tree_could_trap_p is a predicate for rvalues, so check
             for readonly memory explicitly.  */
          || ((base = get_base_address (lhs))
              && ((DECL_P (base)
                   && TREE_READONLY (base))
                  || TREE_CODE (base) == STRING_CST)))
        return false;
    }


I'll add these in this transformation too (maybe excluding the trap check).