[PATCH] Check optab before transforming atomic bit test and operations

[Jeff Law]

On 11/15/2021 12:05 PM, H.J. Lu wrote:
> On Mon, Nov 15, 2021 at 10:59 AM Jeff Law <jeffreyalaw@gmail.com> wrote:
>>
>>
>> On 11/15/2021 6:39 AM, H.J. Lu via Gcc-patches wrote:
>>> Check optab before transforming equivalent, but slighly different cases
>>> of atomic bit test and operations to their canonical forms.
>>>
>>> gcc/
>>>
>>>        PR middle-end/103184
>>>        * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Check optab
>>>        before transforming equivalent, but slighly different cases to
>>>        their canonical forms.
>>>
>>> gcc/testsuite/
>>>
>>>        PR middle-end/103184
>>>        * gcc.dg/pr103184-1.c: New test.
>>>        * gcc.dg/pr103184-2.c: Likewise.
>>>        }
>>>        }
>>>
>>> -  switch (fn)
>>> -    {
>>> -    case IFN_ATOMIC_BIT_TEST_AND_SET:
>>> -      optab = atomic_bit_test_and_set_optab;
>>> -      break;
>>> -    case IFN_ATOMIC_BIT_TEST_AND_COMPLEMENT:
>>> -      optab = atomic_bit_test_and_complement_optab;
>>> -      break;
>>> -    case IFN_ATOMIC_BIT_TEST_AND_RESET:
>>> -      optab = atomic_bit_test_and_reset_optab;
>>> -      break;
>>> -    default:
>>> -      return;
>>> -    }
>>> -
>>>      if (optab_handler (optab, TYPE_MODE (TREE_TYPE (lhs))) == CODE_FOR_nothing)
>>>        return;
>> Shouldn't the test of the return value of optab_handler here just go
>> away since we're testing it earlier?  OK with that fix.
>>
> The earlier check is predicated on if (rhs_code != BIT_AND_EXPR):
>
>    if (rhs_code != BIT_AND_EXPR)
>      {
>        if (rhs_code != NOP_EXPR && rhs_code != BIT_NOT_EXPR)
>          return;
>
>        tree use_lhs = gimple_assign_lhs (use_stmt);
>        if (TREE_CODE (use_lhs) == SSA_NAME
>            && SSA_NAME_OCCURS_IN_ABNORMAL_PHI (use_lhs))
>          return;
>
>        tree use_rhs = gimple_assign_rhs1 (use_stmt);
>        if (lhs != use_rhs)
>          return;
>
>        if (optab_handler (optab, TYPE_MODE (TREE_TYPE (lhs)))
>            == CODE_FOR_nothing)
>          return;
>
> I can add an "else"
>
> else  if (optab_handler (optab, TYPE_MODE (TREE_TYPE (lhs)))
>              == CODE_FOR_nothing)
>      return;
>
> Will it be OK?
Sure.  THanks.
jeff