RE: [PHP-DEV] Re: Benchmark Results for PHP Master 2015-11-12
| From: | Andone, Bogdan | Date: | Mon, 16 Nov 2015 08:19:10 +0000 |
| Subject: | RE: [PHP-DEV] Re: Benchmark Results for PHP Master 2015-11-12 | ||
| References: | 1 2 3 | Groups: | php.internals |
| Request: | Send a blank email to internals+get-89222@lists.php.net to get a copy of this message | ||
> -----Original Message-----
> From: Matt Wilmas [mailto:php_lists@realplain.com]
> Sent: Friday, November 13, 2015 1:49 PM
> To: internals@lists.php.net; Andrea Faulds
> Subject: Re: [PHP-DEV] Re: Benchmark Results for PHP Master 2015-11-12
>
> Hi,
>
> ----- Original Message -----
> From: "Andrea Faulds"
> Sent: Friday, November 13, 2015
>
> > Hi,
> >
> > lp_benchmark_robot@intel.com wrote:
> >>
> >> ---------------------------------------------------------------------
> ----------------------
> >> benchmark relative change since change
> >> since current rev run
> >> std_dev* last run
> >> baseline with PGO
> >> ---------------------------------------------------------------------
> ----------------------
> >> :-( Wordpress 4.2.2 cgi -T10000
> >> 9% -3.67% -1.28% 6.51%
> >> :-( Drupal 7.36 cgi -T10000
> >> 8% -2.03% -0.67% 2.63%
> >> :-( MediaWiki 1.23.9 cgi -T5000
> >> 1% -3.19% -0.97% 3.84%
> >> :-( bench.php cgi -T1
> >> 08% -16.51% -12.70% 9.68%
> >> :-( micro_bench.php cgi -T1 0.03% -1.50%
> >> 0.10% 5.41%
> >> :-( mandelbrot.php cgi -T1
> >> 31% -25.68% -18.78% -0.74%
> >> ---------------------------------------------------------------------
> ----------------------
> >
> >
> > Huh? Sad faces across the board, as bad as -25% for one test?
> >
> > Did we have to change something and it inadvertently affected
> performance?
> > This looks pretty bad.
>
> Yup, I noticed right away yesterday and wondered the same, but decided
> to
> hold off on commenting until the next report(s) come(s). :-)
>
> I figured it might have happened "between some changes that weren't
> optimal," but after the commits were finalized, etc. it would be sorted
> out
> again. But this does seem pretty extreme, and I don't recall seeing
> radical
> changes since the last one... *shrug*
Indeed, it seemed that the sources "sampling" occurred on an unstable commit. We
didn't escalate explicitly the issue as we saw it solved during the next day with the
performance back on track.
Not sure if this was the cause, but having a short look at the commits applied just before and after
the performance drop we observed several iterations and reverts on #70895 and #70898.
>
>
> - Matt
>
>
> --
> PHP Internals - PHP Runtime Development Mailing List
> To unsubscribe, visit: http://www.php.net/unsub.php
Kind Regards,
Bogdan