Do not give realistic estimates for loop with array accesses

[Bin.Cheng]

On Wed, Mar 30, 2016 at 3:22 PM, Jan Hubicka <hubicka@ucw.cz> wrote:
>> > -  /* If access is not executed on every iteration, we must ensure that overlow may
>> > -     not make the access valid later.  */
>> > +  /* If access is not executed on every iteration, we must ensure that overlow
>> > +     may not make the access valid later.  */
>> >    if (!dominated_by_p (CDI_DOMINATORS, loop->latch, gimple_bb (data->stmt))
>> >        && scev_probably_wraps_p (initial_condition_in_loop_num (ev, loop->num),
>> >                                 step, data->stmt, loop, true))
>> > -    reliable = false;
>> > +    upper = false;
>> >
>> > -  record_nonwrapping_iv (loop, init, step, data->stmt, low, high, reliable, upper);
>> > +  record_nonwrapping_iv (loop, init, step, data->stmt, low, high, false, upper);
>> >    return true;
>> >  }
>> Hi,
>> I have a concern that GCC records bound information from basic blocks
>> even that don't dominate loop latch.  Given below example:
>>
>> extern int b[];
>> void bar (int *);
>> int foo (int x, int n)
>> {
>>   int i;
>>   int arr[128] = {0};
>>
>>   for (i = 0; i < n; i++)
>>     {
>>       if (x > i)
>>         {
>>           a[i] = i;
>>           b[i] = i;
>>         }
>>     }
>>   bar (arr);
>>   return 0;
>> }
>
> This testcase is not affected, becase scev_probably_wraps_p returns false in this case.
> In the wrapping case, we can't derive upper bound - this is indeed a correctness issue.
In the wrapping case, we still can derive upper bound if the index's
wrapping range is larger than array bound.  But I agree it looks like
very corner case and not likely be useful in practice.

Thanks,
bin
>
> I am experiemtning with enabling loop peeling by default. For that I extended the code
> to record likely upper bounds, which can be used to test if the loop most probably has
> low trip count. This is also useful to trottle other transformations.
>
> In this case we can probably assume that no sane person would count
> on wrapping and record this as likely bound.
>
> Honza