Re: Method call overhead

From:	Rowan Collins	Date:	Fri, 05 Jun 2015 09:08:04 +0000
Subject:	Re: Method call overhead
References:	1 2 3 4 5 6	Groups:	php.internals
Request:	Send a blank email to [email protected] to get a copy of this message

Sara Golemon wrote on 04/06/2015 22:55:
On Jun 4, 2015, at 10:00, Rowan Collins <[email protected]> wrote:
: http://3v4l.org/l75kf

HHVM's percentages are lower primarily because its for loop measure is much slower in absolute terms.

At the risk of distracting from the central topic, I'd like to point out that HHVM's times are almost certainly based on non-JITted code.  The JIT doesn't kick in until the 11th request, and 3v4l scripts are, by definition, only run once.

PHP is optimized for blind interp, so it's doing better in that mode.

That makes a lot of sense, and just goes to show how hard it is to get useful information out of a simple benchmark.



   

Thread (12 messages)

• Bishop BettiniWed, 03 Jun 2015 03:30:05 +0000
  • Michael WallnerWed, 03 Jun 2015 07:50:27 +0000
    • Sara GolemonWed, 03 Jun 2015 14:29:51 +0000
      • Julien PauliWed, 03 Jun 2015 15:13:07 +0000
  • Brian MoonWed, 03 Jun 2015 22:33:51 +0000
    • Sara GolemonWed, 03 Jun 2015 23:34:16 +0000
    • Yasuo OhgakiThu, 04 Jun 2015 04:35:53 +0000
      • Yasuo OhgakiThu, 04 Jun 2015 04:44:07 +0000
      • Marc BennewitzThu, 04 Jun 2015 09:01:19 +0000
        • Rowan CollinsThu, 04 Jun 2015 17:00:02 +0000
          • Sara GolemonThu, 04 Jun 2015 21:55:36 +0000
            • Rowan CollinsFri, 05 Jun 2015 09:08:04 +0000