[patch] Fix PR43464: update loop closed SSA form once copy prop is done

[Richard Guenther]

On Fri, Apr 2, 2010 at 1:57 PM, Jack Howarth <howarth@bromo.med.uc.edu> wrote:
> On Fri, Apr 02, 2010 at 12:56:36PM +0200, Richard Guenther wrote:
>> On Thu, Apr 1, 2010 at 8:24 PM, Sebastian Pop <sebpop@gmail.com> wrote:
>> > On Thu, Apr 1, 2010 at 08:48, Richard Guenther <rguenther@suse.de> wrote:
>> >> Sort of.  Only verify loop closed SSA if loop_state_satisifes_p
>> >> has LCSSA.
>> >
>> > done, see 0001
>> >
>> >>  Also add a param to verify_loop_closed_ssa whether
>> >> verify_ssa should be called inside it (we do that from passes.c
>> >> already, so avoid the duplicate).
>> >
>> > done, see 0002
>> >
>> >>  For simplicity I'd just
>> >> add TODO_verify_loops to the flags when loop state satisifes
>> >> LCSSA in passes.c (which can then also avoid passing true to
>> >> verify_loop_closed_ssa if ssa verification was already done).
>> >>
>> >
>> > This confuses me: why do you want to keep the TODO_verify_loops?
>> > TODO_verify_loops won't have any effect on all the LNO passes
>> > if I'm adding this change to execute_function_todo
>> >
>> > if (current_loops && loops_state_satisfies_p (LOOP_CLOSED_SSA))
>> >  flags = flags & TODO_verify_loops;
>>
>> Indeed.  Deleting the TODO as you do is also ok.
>>
>> The patches are ok for 4.6 stage1.
>>
>
> Richard,
>   Will these be suitable for backporting into gcc 4.5 branch
> after the gcc 4.5.0 release? Reading the gcc 4.6 plans...

Why should they?  They only add additional verifiers.

Richard.

>
> http://gcc.gnu.org/ml/gcc/2009-09/msg00501.html
>
> ...it would appear that gcc 4.6 could be a rather odd
> release since it mainly focuses on compiler tuning rather
> than new branches being merged. The cut-off for what is
> backportable might be fuzzier than in the past development
> cycles.
>                  Jack
>
>> Thanks,
>> Richard.
>>
>> > Sebastian
>> >
>