On Tue, 24 Aug 2021, Xionghu Luo wrote:
>
>
> On 2021/8/19 20:11, Richard Biener wrote:
> >> - class loop *inn_loop = loop;
> >>
> >> if (ALWAYS_EXECUTED_IN (loop->header) == NULL)
> >> {
> >> @@ -3232,19 +3231,6 @@ fill_always_executed_in_1 (class loop *loop,
> >> sbitmap contain
On 2021/8/19 20:11, Richard Biener wrote:
>> - class loop *inn_loop = loop;
>>
>> if (ALWAYS_EXECUTED_IN (loop->header) == NULL)
>> {
>> @@ -3232,19 +3231,6 @@ fill_always_executed_in_1 (class loop *loop, sbitmap
>> contains_call)
>> to disprove this if possible). */
>>
CUTED bb after inner loop...
> >
> > But the code says:
> >
> > /* In a loop that is always entered we may proceed anyway.
> > But record that we entered it and stop once we leave it.
> > */
> >
> > and you d
leave it.
>> */
>>
>> and you do not remove this comment still it doesn't hold anymore
>> after your patch. I don't say the current code is optimal - I just
>> say it does exactly what is documented.
>
> Removed.
>
>>
>>
ymore
> after your patch. I don't say the current code is optimal - I just
> say it does exactly what is documented.
Removed.
>
>
>>>
>>> In fact for your testcase the x->j ref is _not_ always executed
>>> since the inner loop is conditional on n > 0.
