From: Qing Zhao <qing.zhao@oracle.com>
To: Richard Biener <rguenther@suse.de>
Cc: "jason@redhat.com" <jason@redhat.com>,
gcc-patches Nick Alcock via <gcc-patches@gcc.gnu.org>
Subject: Re: [patch][middle-end/PR102359]Not add initialization for DECL_VALUE_EXPR variables with -ftrivial-auto-var-init
Date: Tue, 5 Oct 2021 14:32:54 +0000 [thread overview]
Message-ID: <C3D9506E-E85C-4058-BCCF-189EDB6D7143@oracle.com> (raw)
In-Reply-To: <2r53651-n1q1-2rp7-p316-o330p2sp161o@fhfr.qr>
> On Oct 5, 2021, at 3:19 AM, Richard Biener <rguenther@suse.de> wrote:
>
> On Tue, 5 Oct 2021, Qing Zhao wrote:
>
>> Hi,
>>
>> This is the patch to fix this issue based on our discussion.
>>
>> I have tested it on aarch64 with bootstrap and regtests. X86 bootstrap was done, regtests is ongoing.
>>
>> Okay for trunk?
>>
>> Thanks.
>>
>> Qing
>>
>> ======================
>> From d349ef0145512efe7f9af2c6bbd01f636475bce3 Mon Sep 17 00:00:00 2001
>> From: qing zhao <qing.zhao@oracle.com>
>> Date: Mon, 4 Oct 2021 15:26:03 -0700
>> Subject: [PATCH] middle-end/102359 Not add initialization for variables that
>> have been initialized by FEs.
>>
>> C++ FE creates proxy variables, which have associated DECL_VALUE_EXPR
>> and have been initialized by FE. For such auto variable, we should not
>> add initialization when -ftrivial-auto-var-init presents.
>>
>> gcc/ChangeLog:
>>
>> 2021-10-04 qing zhao <qing.zhao@oracle.com>
>>
>> * gimplify.c (is_decl_init_by_fe): New function.
>> (gimplify_decl_expr): Not add initialization for an auto variable
>> when it has been initialized by frontend.
>>
>> gcc/testsuite/ChangeLog:
>>
>> 2021-10-04 qing zhao <qing.zhao@oracle.com>
>>
>> * g++.dg/pr102359_1.C: New test.
>> * g++.dg/pr102359_2.C: New test.
>> ---
>> gcc/gimplify.c | 21 ++++++++++++++++++++-
>> gcc/testsuite/g++.dg/pr102359_1.C | 13 +++++++++++++
>> gcc/testsuite/g++.dg/pr102359_2.C | 13 +++++++++++++
>> 3 files changed, 46 insertions(+), 1 deletion(-)
>> create mode 100644 gcc/testsuite/g++.dg/pr102359_1.C
>> create mode 100644 gcc/testsuite/g++.dg/pr102359_2.C
>>
>> diff --git a/gcc/gimplify.c b/gcc/gimplify.c
>> index b27776a..d6865ad 100644
>> --- a/gcc/gimplify.c
>> +++ b/gcc/gimplify.c
>> @@ -1819,6 +1819,19 @@ gimple_add_padding_init_for_auto_var (tree decl, bool is_vla,
>> gimplify_seq_add_stmt (seq_p, call);
>> }
>>
>> +/* Return true if the DECL is initialized by FE.
>> + If the VAR_DECL has DECL_VALUE_EXPR that was created by FE (usually C++FE),
>> + it's a proxy varaible, and FE already initializd the DECL_VALUE_EXPR of it.
>> +*/
>> +static bool
>> +is_decl_init_by_fe (tree decl, bool is_created_by_fe)
>> +{
>> + if (DECL_HAS_VALUE_EXPR_P (decl)
>> + && is_created_by_fe)
>> + return true;
>> + return false;
>> +}
>> +
>> /* Return true if the DECL need to be automaticly initialized by the
>> compiler. */
>> static bool
>> @@ -1871,8 +1884,13 @@ gimplify_decl_expr (tree *stmt_p, gimple_seq *seq_p)
>> if (VAR_P (decl) && !DECL_EXTERNAL (decl))
>> {
>> tree init = DECL_INITIAL (decl);
>> + bool is_value_expr_created_by_fe = false;
>
> no need for the = false, it's always initialized below.
>
>> bool is_vla = false;
>>
>> + /* Check whether a decl has FE created VALUE_EXPR here BEFORE
>> + gimplify_vla_decl creates VALUE_EXPR for vla decl. */
>> + is_value_expr_created_by_fe = DECL_HAS_VALUE_EXPR_P (decl);
>
> That looks a bit weird when looking at ...
>
>> +
>> poly_uint64 size;
>> if (!poly_int_tree_p (DECL_SIZE_UNIT (decl), &size)
>> || (!TREE_STATIC (decl)
>> @@ -1934,7 +1952,8 @@ gimplify_decl_expr (tree *stmt_p, gimple_seq *seq_p)
>> /* When there is no explicit initializer, if the user requested,
>> We should insert an artifical initializer for this automatic
>> variable. */
>> - else if (is_var_need_auto_init (decl))
>> + else if (is_var_need_auto_init (decl)
>> + && !is_decl_init_by_fe (decl, is_value_expr_created_by_fe))
>
> ... which just expands to
>
> if (DECL_HAS_VALUE_EXPR_P (decl) && DECL_HAS_VALUE_EXPR_P (decl))
>
> can you please name 'is_value_expr_created_by_fe' as
> 'decl_had_value_expr_p' and check && !decl_had_value_expr_p here?
> So sth like
I can do this -:) I agree that the change will make the code simpler.
However, my major concern with this change is: later when people look at this change, they might ask:
Why we should not initialize a variable with VALUE_EXPR? And whether the variable whose VALUE_EXPR
was created by “gimplify_vla_decl” should be excluded?
My new function and comments were all for this purpose.
If I go with this change, at least we should add some comments to explain this as following, what do you think?
>
> diff --git a/gcc/gimplify.c b/gcc/gimplify.c
> index b27776af7c8..9013f385f13 100644
> --- a/gcc/gimplify.c
> +++ b/gcc/gimplify.c
> @@ -1872,6 +1872,7 @@ gimplify_decl_expr (tree *stmt_p, gimple_seq *seq_p)
> {
> tree init = DECL_INITIAL (decl);
> bool is_vla = false;
+ /* Check whether a decl has FE created VALUE_EXPR here BEFORE
+ gimplify_vla_decl creates VALUE_EXPR for vla decl. */
> + bool decl_had_value_expr_p = DECL_HAS_VALUE_EXPR_P (decl);
>
> poly_uint64 size;
> if (!poly_int_tree_p (DECL_SIZE_UNIT (decl), &size)
> @@ -1934,7 +1935,8 @@ gimplify_decl_expr (tree *stmt_p, gimple_seq *seq_p)
> /* When there is no explicit initializer, if the user requested,
> We should insert an artifical initializer for this automatic
> variable. */
> - else if (is_var_need_auto_init (decl))
> + else if (is_var_need_auto_init (decl)
+ /* If the VAR_DECL has DECL_VALUE_EXPR that was created by FE (usually C++FE),
+ it's a proxy varaible, and FE already initializd the DECL_VALUE_EXPR of it. */
> + && !decl_had_value_expr_p)
> {
> gimple_add_init_for_auto_var (decl,
> flag_auto_var_init,
>
> OK with that change.
thanks.
Qing
>
> Thanks,
> Richard.
>
>
>> {
>> gimple_add_init_for_auto_var (decl,
>> flag_auto_var_init,
>> diff --git a/gcc/testsuite/g++.dg/pr102359_1.C b/gcc/testsuite/g++.dg/pr102359_1.C
>> new file mode 100644
>> index 0000000..da643cd
>> --- /dev/null
>> +++ b/gcc/testsuite/g++.dg/pr102359_1.C
>> @@ -0,0 +1,13 @@
>> +/* PR middle-end/102359 ICE gimplification failed since
>> + r12-3433-ga25e0b5e6ac8a77a. */
>> +/* { dg-do compile } */
>> +/* { dg-options "-ftrivial-auto-var-init=zero" } */
>> +/* { dg-require-effective-target c++17 } */
>> +
>> +struct A {
>> + double a = 111;
>> + auto foo() {
>> + return [*this] { return a; };
>> + }
>> +};
>> +int X = A{}.foo()();
>> diff --git a/gcc/testsuite/g++.dg/pr102359_2.C b/gcc/testsuite/g++.dg/pr102359_2.C
>> new file mode 100644
>> index 0000000..d026d72
>> --- /dev/null
>> +++ b/gcc/testsuite/g++.dg/pr102359_2.C
>> @@ -0,0 +1,13 @@
>> +/* PR middle-end/102359 ICE gimplification failed since
>> + r12-3433-ga25e0b5e6ac8a77a. */
>> +/* { dg-do run} */
>> +/* { dg-options "-ftrivial-auto-var-init=zero" } */
>> +/* { dg-require-effective-target c++17 } */
>> +
>> +int main()
>> +{
>> + int i = 42;
>> + auto l = [=]() mutable { return i; };
>> + if (l() != i)
>> + __builtin_abort ();
>> +}
>>
>
> --
> Richard Biener <rguenther@suse.de>
> SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg,
> Germany; GF: Felix Imendörffer; HRB 36809 (AG Nuernberg)
next prev parent reply other threads:[~2021-10-05 14:33 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-05 4:52 Qing Zhao
2021-10-05 8:19 ` Richard Biener
2021-10-05 14:32 ` Qing Zhao [this message]
2021-10-05 15:36 ` Qing Zhao
2021-10-05 18:30 ` Jason Merrill
2021-10-05 22:39 ` Qing Zhao
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=C3D9506E-E85C-4058-BCCF-189EDB6D7143@oracle.com \
--to=qing.zhao@oracle.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jason@redhat.com \
--cc=rguenther@suse.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).