Re: [RFC] Handle unary pass-through jump functions for ipa-vrp

["Bin.Cheng" <amker.cheng@gmail.com>]

On Tue, Nov 8, 2016 at 10:13 AM, kugan
<kugan.vivekanandarajah@linaro.org> wrote:
> Hi,
>
>
>
> On 04/11/16 04:36, Martin Jambor wrote:
>>
>> Hi,
>>
>> On Fri, Oct 28, 2016 at 02:03:47PM +1100, kugan wrote:
>>>
>>>
>>> ...snip...
>>>
>>> I have also separated the constant parameter conversion out and posted as
>>> https://gcc.gnu.org/ml/gcc-patches/2016-10/msg02309.html. This is now
>>> handling just unary pass-through jump functions.
>>>
>>> Bootstrapped and regression tested on x86_64-linux-gnu with no new
>>> regressions.
>>>
>>> Is this OK for trunk?
>>>
>>> Thanks,
>>> Kugan
>>>
>>> gcc/testsuite/ChangeLog:
>>>
>>> 2016-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>>
>>>         * gcc.dg/ipa/vrp7.c: New test.
>>>
>>>
>>> gcc/ChangeLog:
>>>
>>> 2016-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
>>>
>>>         * ipa-cp.c (ipa_get_jf_pass_through_result): Handle unary
>>> expressions.
>>>         (propagate_vr_accross_jump_function): Likewise.
>>>         * ipa-prop.c (ipa_set_jf_unary_pass_through): New.
>>>         (load_from_param_1): New.
>>>         (load_from_unmodified_param): Factor common part into
>>> load_from_param_1.
>>>         (load_from_param): New.
>>>         (compute_complex_assign_jump_func): Handle unary expressions.
>>>         (ipa_write_jump_function): Likewise.
>>>         (ipa_read_jump_function): Likewise.
>>>
>>>
>>>> Patch is OK with changes Martin suggested.
>>>>
>>>> Honza
>>>>
>>
>>> From b7d9b413951ba20d156a7801640cc7d7bc57c062 Mon Sep 17 00:00:00 2001
>>> From: Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
>>> Date: Fri, 28 Oct 2016 10:16:38 +1100
>>> Subject: [PATCH 2/2] add unary jump function
>>>
>>> ---
>>>  gcc/ipa-cp.c                    | 39 +++++++++++++++---
>>>  gcc/ipa-prop.c                  | 89
>>> +++++++++++++++++++++++++++++++++++------
>>>  gcc/testsuite/gcc.dg/ipa/vrp7.c | 32 +++++++++++++++
>>>  3 files changed, 142 insertions(+), 18 deletions(-)
>>>  create mode 100644 gcc/testsuite/gcc.dg/ipa/vrp7.c
>>>
>>> diff --git a/gcc/ipa-cp.c b/gcc/ipa-cp.c
>>> index 9f28557..8fc95dd 100644
>>> --- a/gcc/ipa-cp.c
>>> +++ b/gcc/ipa-cp.c
>>> @@ -1225,13 +1225,21 @@ ipa_get_jf_pass_through_result (struct
>>> ipa_jump_func *jfunc, tree input)
>>>      return NULL_TREE;
>>>
>>>    if (TREE_CODE_CLASS (ipa_get_jf_pass_through_operation (jfunc))
>>> -      == tcc_comparison)
>>> -    restype = boolean_type_node;
>>> +      == tcc_unary)
>>> +    {
>>> +      res = fold_unary (ipa_get_jf_pass_through_operation (jfunc),
>>> +                       TREE_TYPE (input), input);
>>> +    }
>>
>>
>> Please do not put curly braces around a single statement.  Apart from
>> that, no objection from me.
>
>
> Thanks Martin, I will fix this.
>
> Honza, is this OK for you with the above fix?
>
> Thanks,
> Kugan
>>
>>
>> Thanks,
>>
>> Martin
>>
>
Hi,
This new test fails on AArch64.  I created PR78316 for tracking.

Thanks,
bin