From: Richard Biener <richard.guenther@gmail.com>
To: Jeff Law <law@redhat.com>, Jan Hubicka <hubicka@ucw.cz>
Cc: Xi Ruoyao <ryxi@stu.xidian.edu.cn>,
GCC Patches <gcc-patches@gcc.gnu.org>,
Florent Hivert <florent.hivert@lri.fr>
Subject: Re: [PATCH] Destroy arguments for _Cilk_spawn calling in the child (PR 80038)
Date: Thu, 13 Apr 2017 07:05:00 -0000 [thread overview]
Message-ID: <CAFiYyc0TPi_gEA0f8zJU5oUEBPhoTMXUwU9ysG1NeH5+qmKZGg@mail.gmail.com> (raw)
In-Reply-To: <f2f74bb5-7f36-6306-8b81-de7d87a46a14@redhat.com>
On Wed, Apr 12, 2017 at 11:16 PM, Jeff Law <law@redhat.com> wrote:
> On 04/07/2017 08:02 AM, Xi Ruoyao wrote:
>>
>> On 2017-04-06 11:12 -0600, Jeff Law wrote:
>>
>>> With the likely deprecation in mind, I've only done a cursory review of
>>> the changes -- mostly to verify that they hit Cilk+ paths only.
>>
>>
>>> What's the purpose behind changing when we set the in_lto_p flag?
>>
>>
>> Without that change, GCC with my patch ICEed with _Cilk_spawn and
>> -flto -O3 -fcilkplus since __cilkrts_stack_frame.ctx's type (array of void
>> *)
>> was not TYPE_STRUCTURAL_EQUALITY_P in lto stage.
>> If this change is not proper, I'll work on modifying my patch to work
>> without touching in_lto_p.
>
> It's certainly be preferable to not change in_lto_p-- unless Richi wants to
> chime in on the safety of setting in_lto_p earlier.
>
> I'm not familiar enough with the LTO interactions to know if movement of
> in_lto_p is safe.
It should be safe and even technically more correct given we now have
various conditionals in the type building routines in tree.c that check
in_lto_p and avoid setting TYPE_CANONICAL from them -- TYPE_CANONICAL
is re-computed later, and for the builtin types we first zero TYPE_CANONICAL
(see lto.c:read_cgraph_and_symbols).
Now... I think those checks are somewhat wrong given that the middle-end
_does_ create types later, hopefully not ones we use for alias purposes,
but I'm not 100% sure. So it somewhat feels like a hack ;)
So in theory the change is a good one. I'm still nervous at this stage.
Did you verify LTO bootstrap still works with the patch?
CCing Honza who fiddled with this last (and introduced all those
in_lto_p checks).
Thanks,
Richard.
> Jeff
next prev parent reply other threads:[~2017-04-13 7:05 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-03-31 13:58 Xi Ruoyao
2017-04-06 17:12 ` Jeff Law
2017-04-07 14:03 ` Xi Ruoyao
2017-04-12 21:17 ` Jeff Law
2017-04-13 7:05 ` Richard Biener [this message]
2017-04-14 7:00 ` Xi Ruoyao
2017-04-15 4:05 ` Xi Ruoyao
2017-04-25 15:37 ` Jeff Law
2017-04-28 14:51 ` Xi Ruoyao
2017-04-28 15:29 ` Jeff Law
2017-04-29 8:51 ` [PATCH v2] " Xi Ruoyao
2017-05-01 22:26 ` Jeff Law
2017-05-02 7:56 ` Andreas Schwab
2017-05-02 8:18 ` Xi Ruoyao
2017-05-02 15:45 ` Jeff Law
2017-08-21 17:33 ` [PATCH, gcc-7-branch] Backport PR80038 Xi Ruoyao
2017-08-21 17:46 ` Xi Ruoyao
2017-08-22 9:04 ` Richard Biener
2017-08-24 13:06 ` Xi Ruoyao
2017-09-06 4:33 ` Xi Ruoyao
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=CAFiYyc0TPi_gEA0f8zJU5oUEBPhoTMXUwU9ysG1NeH5+qmKZGg@mail.gmail.com \
--to=richard.guenther@gmail.com \
--cc=florent.hivert@lri.fr \
--cc=gcc-patches@gcc.gnu.org \
--cc=hubicka@ucw.cz \
--cc=law@redhat.com \
--cc=ryxi@stu.xidian.edu.cn \
/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).