ivopts improvement

[Tom de Vries]

On 03/14/2011 01:55 PM, Zdenek Dvorak wrote:
> Hi,
> 
>>>> it is trying to allow for
>>>>
>>>> do
>>>>   {
>>>>     *p = '\0';
>>>>     i++; /* use_uses_inced_iv == true */
>>>>     p++; /* use_after_cand_inc == true */
>>>>     if (!(i < n))
>>>>       break;
>>>>   }
>>>> while (1);
>>>>
>>>> and for
>>>>
>>>> do
>>>>   {
>>>>     *p = '\0';
>>>>     if (!(i < n))
>>>>       break;
>>>>     i++; /* use_uses_inced_iv == false */
>>>>     p++; /* use_after_cand_inc == false */
>>>>   }
>>>> while (1);
>>>>
>>>> but not for
>>>>
>>>> do
>>>>   {
>>>>     *p = '\0';
>>>>     i++; /* use_uses_inced_iv == true */
>>>>     if (!(i < n))
>>>>       break;
>>>>     p++; /* use_after_cand_inc == false */
>>>>   }
>>>> while (1);
>>>>
>>>> and not for
>>>>
>>>> do
>>>>   {
>>>>     *p = '\0';
>>>>     p++; /* use_after_cand_inc == true */
>>>>     if (!(i < n))
>>>>       break;
>>>>     i++; /* use_uses_inced_iv == false */
>>>>   }
>>>> while (1);
>>>>
>>>> In the latter 2 cases, I cannot simply replace i < n with p < base + n.
>>>
>>> Why not (other than that the value to compare with varies in these cases, but
>>> it always is "the value of p in the last iteration of the loop")?
>>
>> In the 2 latter cases, the value to compare with will be either base + n
>> + 1 or base + n - 1. The code does not handle these cases yet.
> 
> well, if not, then it certainly handles one of the first two cases incorrectly

The ivopts code for the first example is:

<bb 2>:
  p_5 = p_3(D) + i_4(D);
  D.2698_12 = p_3(D) + n_8(D);

<bb 3>:
  # p_1 = PHI <p_5(2), p_7(4)>
  *p_1 = 0;
  p_7 = p_1 + 1;
  if (p_7 >= D.2698_12)
    goto <bb 5>;
  else
    goto <bb 4>;

<bb 4>:
  goto <bb 3>;

<bb 5>:
  return;

and for the second example is:

<bb 2>:
  p_5 = p_3(D) + i_4(D);
  D.2704_13 = p_3(D) + n_6(D);

<bb 3>:
  # p_1 = PHI <p_5(2), p_8(4)>
  *p_1 = 0;
  if (p_1 >= D.2704_13)
    goto <bb 5>;
  else
    goto <bb 4>;

<bb 4>:
  p_8 = p_1 + 1;
  goto <bb 3>;

<bb 5>:
  return;

both seem correct to me.

> (since the use_uses_inced_iv test is meaningless).

To me it seems use_uses_inced_iv has meaning:
- it models something: it states whether the comparison is using
  the iv increment result or the iv phi result.
- it is used to detect cases were we would generate incorrect code.

>>> One more thing: what is fold_walk_def_plus for?
>>
>> It tries to fold a plus, and if not successful, finds ssa vars in
>> operands of the plus, substitutes the defining statements of ssa vars
>> for those ssa vars and retries folding.
> 
> Sorry for not being more clear; I meant, why is it used?
> 

In the following code

<bb 2>:
  p_5 = p_3(D) + i_4(D);

<bb 3>:
  # p_1 = PHI <p_5(2), p_7(4)>
  # i_2 = PHI <i_4(D)(2), i_6(4)>
  *p_1 = 0;
  i_6 = i_2 + 1;
  p_7 = p_1 + 1;
  if (i_6 >= n_8(D))
    goto <bb 5>;
  else
    goto <bb 4>;

<bb 4>:
  goto <bb 3>;

<bb 5>:
  return;


I'm trying to prove that p_5 == x + i_4 where x is an ssa_name. I can't
do that without going through the defining statement of p_5.

> Zdenek