From: Rong Xu <xur@google.com>
To: Andrew Pinski <pinskia@gmail.com>
Cc: Jan Hubicka <hubicka@ucw.cz>,
GCC Patches <gcc-patches@gcc.gnu.org>,
David Li <davidxl@google.com>,
reply@codereview.appspotmail.com
Subject: Re: [google 4.7] atomic update of profile counters (issue6965050)
Date: Thu, 20 Dec 2012 19:35:00 -0000 [thread overview]
Message-ID: <CAF1bQ=QwfgE3sRi_vWntNfNk_5VQpZCOx3vwMy+tiAPum+EgyQ@mail.gmail.com> (raw)
In-Reply-To: <CA+=Sn1nN04977i9RJd4SokcZFw-pMsTWHo35+tZszOsgNSph9A@mail.gmail.com>
we have this patch primarily for getting valid profile counts. we
observe that for some high-threaded programs, we are getting poor
counter due to data racing of counter update (like counter value is
only 15% of what it supposed to be for a 10-thread program).
In general, enabling atomic updates slows down programs. (for my some
of my toy programs, it has 3x slow down.) And that the reason I use
options to control value and edge profile count.
-Rong
On Thu, Dec 20, 2012 at 8:57 AM, Andrew Pinski <pinskia@gmail.com> wrote:
> On Thu, Dec 20, 2012 at 8:20 AM, Jan Hubicka <hubicka@ucw.cz> wrote:
>>> On Wed, Dec 19, 2012 at 4:29 PM, Andrew Pinski <pinskia@gmail.com> wrote:
>>> >
>>> > On Wed, Dec 19, 2012 at 12:08 PM, Rong Xu <xur@google.com> wrote:
>>> > > Hi,
>>> > >
>>> > > This patch adds the supprot of atomic update the profile counters.
>>> > > Tested with google internal benchmarks and fdo kernel build.
>>> >
>>> > I think you should use the __atomic_ functions instead of __sync_
>>> > functions as they allow better performance for simple counters as you
>>> > can use __ATOMIC_RELAXED.
>>>
>>> You are right. I think __ATOMIC_RELAXED should be OK here.
>>> Thanks for the suggestion.
>>>
>>> >
>>> > And this would be useful for the trunk also. I was going to implement
>>> > this exact thing this week but some other important stuff came up.
>>>
>>> I'll post trunk patch later.
>>
>> Yes, I like that patch, too. Even if the costs are quite high (and this is why
>> atomic updates was sort of voted down in the past) the alternative of using TLS
>> has problems with too-much per-thread memory.
>
> Actually sometimes (on some processors) atomic increments are cheaper
> than doing a regular incremental. Mainly because there is an
> instruction which can handle it in the L2 cache rather than populating
> the L1. Octeon is one such processor where this is true.
>
> Thanks,
> Andrew Pinski
>
>>
>> While there are even more alternatives, like recording the changes and
>> commmiting them in blocks (say at function return), I guess some solution is
>> better than no solution.
>>
>> Thanks,
>> Honza
next prev parent reply other threads:[~2012-12-20 19:35 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-12-19 20:08 Rong Xu
2012-12-20 0:25 ` Xinliang David Li
2012-12-20 0:29 ` Andrew Pinski
2012-12-20 0:56 ` Rong Xu
2012-12-20 16:21 ` Jan Hubicka
2012-12-20 16:57 ` Andrew Pinski
2012-12-20 19:35 ` Rong Xu [this message]
2012-12-20 19:42 ` Andrew Pinski
2012-12-21 9:13 ` Jan Hubicka
2012-12-21 9:55 ` Richard Biener
2012-12-21 10:36 ` Jan Hubicka
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='CAF1bQ=QwfgE3sRi_vWntNfNk_5VQpZCOx3vwMy+tiAPum+EgyQ@mail.gmail.com' \
--to=xur@google.com \
--cc=davidxl@google.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=hubicka@ucw.cz \
--cc=pinskia@gmail.com \
--cc=reply@codereview.appspotmail.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).