From: Jeff Law <law@redhat.com>
To: Bernd Schmidt <bschmidt@redhat.com>,
GCC Patches <gcc-patches@gcc.gnu.org>,
Jakub Jelinek <jakub@redhat.com>,
Sebastian Pop <sebpop@gmail.com>,
Michael Matz <matz@suse.de>
Subject: Re: Remove noce_mem_write_may_trap_or_fault_p in ifcvt
Date: Fri, 20 Nov 2015 18:57:00 -0000 [thread overview]
Message-ID: <564F6D06.1010806@redhat.com> (raw)
In-Reply-To: <564F297B.8040603@redhat.com>
On 11/20/2015 07:08 AM, Bernd Schmidt wrote:
>
> BZ27313 is marked as fixed by the introduction of the tree cselim pass,
> thus the problem won't even be seen at RTL level.
Cool.
> I'm undecided on whether cs-elim is safe wrt the store speculation vs
> locks concerns raised in the thread discussing Ian's
> noce_can_store_speculate_p, but that's not something we have to consider
> to solve the problem at hand.
I don't think cs-elim is safe WRT locks and such in multi-threaded code.
In particular it replaces this:
bb0:
if (cond) goto bb2; else goto bb1;
bb1:
*p = RHS;
bb2:
with
bb0:
if (cond) goto bb1; else goto bb2;
bb1:
condtmp' = *p;
bb2:
condtmp = PHI <RHS, condtmp'>
*p = condtmp;
If *p is a shared memory location, then there may be another writer. If
that writer happens to store something in that location after the load
of *p, but before the store to *p, then that store will get lost in the
transformed pseudo code.
That seems to introduce a data race. Presumably one would call the
original code ill-formed WRT the C11/C++11 memory model since the shared
location is not marked as such.
I'm willing to consider this an independent problem.
>
> As far as I can tell hmmer and the 27313 testcase are unaffected at -O2
> (if anything, hmmer was very slightly faster afterwards). The run wasn't
> super-scientific, but I wouldn't have expected anything else given the
> existence of cs-elim.
Cool. It doesn't have to be super-scientific. Good to see it didn't
harm anything -- thanks for going the extra mile on this one.
>
> Ok to do the above, removing all the bits made unnecessary (including
> memory_must_be_modified_in_insn_p in alias.c)?
Yup. Zap it.
jeff
next prev parent reply other threads:[~2015-11-20 18:57 UTC|newest]
Thread overview: 24+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-11-06 17:21 Bernd Schmidt
2015-11-06 18:39 ` Jeff Law
2015-11-06 18:44 ` Bernd Schmidt
2015-11-06 19:20 ` Jeff Law
2015-11-06 19:30 ` Bernd Schmidt
2015-11-06 21:09 ` Jeff Law
2015-11-18 19:16 ` Bernd Schmidt
2015-11-18 23:49 ` Jeff Law
2015-11-20 14:09 ` Bernd Schmidt
2015-11-20 18:57 ` Jeff Law [this message]
2015-11-23 16:07 ` Michael Matz
2015-11-25 10:46 ` Bernd Schmidt
2015-11-25 11:53 ` Richard Biener
2015-11-25 13:18 ` Michael Matz
2015-11-25 14:52 ` Richard Biener
2015-11-25 15:02 ` Jakub Jelinek
2015-11-25 15:18 ` Michael Matz
2015-11-25 15:49 ` Bernd Schmidt
2015-11-25 15:55 ` Michael Matz
2015-11-26 9:50 ` Richard Biener
2015-11-27 10:22 ` Bernd Schmidt
2015-11-25 13:18 ` Michael Matz
2015-11-25 10:36 ` Bernd Schmidt
2015-11-23 16:03 ` Michael Matz
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=564F6D06.1010806@redhat.com \
--to=law@redhat.com \
--cc=bschmidt@redhat.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jakub@redhat.com \
--cc=matz@suse.de \
--cc=sebpop@gmail.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).