From: Paolo Bonzini <bonzini@gnu.org>
To: Bernd Schmidt <bernds@codesourcery.com>
Cc: Jeff Law <law@redhat.com>, Eric Botcazou <ebotcazou@adacore.com>,
gcc-patches@gcc.gnu.org, Steven Bosscher <stevenb.gcc@gmail.com>,
Jim Wilson <wilson@codesourcery.com>
Subject: Re: ifcvt/crossjump patch: Fix PR 42496, 21803
Date: Thu, 29 Jul 2010 17:13:00 -0000 [thread overview]
Message-ID: <4C51B605.2050206@gnu.org> (raw)
In-Reply-To: <4C51B3B4.5020602@codesourcery.com>
On 07/29/2010 07:00 PM, Bernd Schmidt wrote:
> On 07/29/2010 06:19 PM, Paolo Bonzini wrote:
>> I'd like to have a note to the reader that df_analyze is only invoked
>> when you do crossjumping. Please add an assert like
>>
>> if (block_was_dirty)
>> {
>> gcc_assert (mode & CLEANUP_CROSSJUMP);
>> df_analyze ();
>> }
>
> Can do.
>
>> We do not use dataflow otherwise, and it is not necessary to call it
>> gratuitously. Passes know that CFG cleanup destroys dataflow and call
>> it themselves if necessary.
>
> Then again, we probably won't lose much by calling df_analyze during
> cfgcleanup if the following pass needs it anyway - right?
What I meant is I want to document that it's for a special case. I
wouldn't like someone to randomly remove the if just because it happens
to fix his bug. Certainly I didn't want to imply any further change. :-)
>> Second, crossjumping is now more expensive. Does it buy much really to
>> iterate it? Something like
>>
>> mode &= ~CLEANUP_CROSSJUMP;
>>
>> just before iterating may still leave it "good enough".
>
> A quick experiment shows that this causes many missed opportunities.
> (Placed it after the run_fast_dce call).
Thanks. Wishful thinking.
Paolo
next prev parent reply other threads:[~2010-07-29 17:10 UTC|newest]
Thread overview: 95+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-03-31 22:08 Bernd Schmidt
2010-04-01 18:00 ` Steven Bosscher
2010-04-01 18:01 ` Bernd Schmidt
2010-04-02 9:45 ` Bernd Schmidt
2010-04-06 9:21 ` Bernd Schmidt
2010-04-10 10:37 ` Eric Botcazou
2010-04-12 23:34 ` Bernd Schmidt
2010-04-13 21:14 ` Eric Botcazou
2010-04-13 21:36 ` Bernd Schmidt
2010-04-13 21:51 ` Eric Botcazou
2010-04-14 20:51 ` Bernd Schmidt
2010-04-14 21:09 ` Bernd Schmidt
2010-04-19 22:05 ` Eric Botcazou
2010-04-19 22:14 ` Steven Bosscher
2010-04-19 22:18 ` Steven Bosscher
2010-04-19 22:47 ` Steven Bosscher
2010-04-20 10:34 ` Eric Botcazou
2010-04-20 11:26 ` Bernd Schmidt
2010-04-23 9:25 ` Eric Botcazou
2010-04-23 11:15 ` Steven Bosscher
2010-05-15 11:24 ` Steven Bosscher
2010-05-28 10:00 ` Eric Botcazou
2010-05-28 11:20 ` Steven Bosscher
2010-04-20 12:30 ` Bernd Schmidt
2010-07-20 20:43 ` Bernd Schmidt
2010-07-22 19:47 ` Eric Botcazou
2010-07-22 21:09 ` Bernd Schmidt
2010-07-23 22:06 ` Eric Botcazou
2010-07-23 22:13 ` Bernd Schmidt
2010-07-24 13:07 ` Eric Botcazou
2010-07-26 9:42 ` Bernd Schmidt
2010-07-26 13:40 ` Paolo Bonzini
2010-07-26 13:50 ` Paolo Bonzini
2010-07-26 13:56 ` Bernd Schmidt
2010-07-26 14:14 ` Paolo Bonzini
2010-07-27 8:31 ` Eric Botcazou
2010-07-27 9:37 ` Bernd Schmidt
2010-07-27 13:35 ` Bernd Schmidt
2010-07-27 22:38 ` Eric Botcazou
2010-07-28 16:58 ` Jeff Law
2010-07-29 8:25 ` Eric Botcazou
2010-07-27 17:39 ` Jeff Law
2010-07-27 22:05 ` Bernd Schmidt
2010-07-27 22:40 ` Eric Botcazou
2010-07-28 17:06 ` Jeff Law
2010-07-29 17:28 ` Jeff Law
2010-07-29 17:43 ` Bernd Schmidt
2010-07-27 22:23 ` Eric Botcazou
2010-07-27 23:04 ` Bernd Schmidt
2010-07-28 8:40 ` Eric Botcazou
2010-07-28 10:13 ` Bernd Schmidt
2010-07-28 19:40 ` Jeff Law
2010-07-28 20:15 ` Bernd Schmidt
2010-07-29 16:00 ` Jeff Law
2010-07-29 16:21 ` Paolo Bonzini
2010-07-29 17:09 ` Bernd Schmidt
2010-07-29 17:13 ` Paolo Bonzini [this message]
2010-07-30 0:55 ` Steven Bosscher
2010-07-28 18:31 ` Jeff Law
2010-07-28 18:36 ` Paolo Bonzini
2010-07-29 9:07 ` Eric Botcazou
2010-07-27 23:08 ` Paolo Bonzini
2010-07-28 21:44 ` Bernd Schmidt
2010-07-29 14:31 ` Jeff Law
2010-07-27 15:31 ` Jeff Law
2010-07-27 22:18 ` Eric Botcazou
2010-07-28 17:07 ` Jeff Law
2010-07-28 17:38 ` Bernd Schmidt
2010-08-02 15:57 ` Jeff Law
2010-08-02 15:59 ` Bernd Schmidt
2010-08-02 16:05 ` Jeff Law
2010-08-02 16:15 ` Bernd Schmidt
2010-08-03 14:10 ` Bernd Schmidt
2010-08-03 15:16 ` Jeff Law
2010-08-03 15:31 ` Bernd Schmidt
2010-08-03 17:13 ` Jeff Law
2010-08-04 13:36 ` Bernd Schmidt
2010-08-30 16:00 ` Bernd Schmidt
2010-09-20 10:23 ` Bernd Schmidt
2010-09-20 16:25 ` Jeff Law
2010-09-23 15:53 ` Bernd Schmidt
2010-09-23 22:00 ` Richard Guenther
2010-09-23 22:03 ` Richard Guenther
2010-09-23 22:18 ` Bernd Schmidt
2010-09-24 11:29 ` Richard Guenther
2010-09-27 15:56 ` Fix PR45792, cris-elf build breakage from PR44374-fix "ifcvt/crossjump patch: Fix PR 42496, 21803" Hans-Peter Nilsson
2010-09-27 20:34 ` Bernd Schmidt
2010-09-27 23:38 ` Hans-Peter Nilsson
2010-09-28 0:07 ` Bernd Schmidt
2010-10-02 13:07 ` ifcvt/crossjump patch: Fix PR 42496, 21803 H.J. Lu
2010-10-03 11:33 ` Bernd Schmidt
2010-10-03 11:39 ` H.J. Lu
2010-10-06 1:12 ` H.J. Lu
2010-10-06 2:46 ` H.J. Lu
2010-04-12 20:43 ` Jim Wilson
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=4C51B605.2050206@gnu.org \
--to=bonzini@gnu.org \
--cc=bernds@codesourcery.com \
--cc=ebotcazou@adacore.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=law@redhat.com \
--cc=stevenb.gcc@gmail.com \
--cc=wilson@codesourcery.com \
/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).