Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:37513
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: error (pb1.pair.com: domain realplain.com from 209.151.69.1 cause and error)
Message-ID: <016801c8b051$c02db680$0201a8c0@pc1>
To: <internals@lists.php.net>,
	"Felipe Pena" <felipensp@gmail.com>
References: <013a01c8aec5$b189d650$0201a8c0@pc1> <481F8BCD.30207@lerdorf.com> <1210034790.5744.5.camel@pena> <01e801c8af82$3ed8bd20$0201a8c0@pc1> <1210103098.5344.17.camel@felipe>
Date: Wed, 7 May 2008 09:50:51 -0500
MIME-Version: 1.0
Content-Type: text/plain;
	charset="Windows-1252"
Content-Transfer-Encoding: 7bit
Subject: Re: [PHP-DEV] [PATCH] Some string changes/optimizations
From: php_lists@realplain.com ("Matt Wilmas")

Hi Felipe,

I don't know. :-/  That's a pretty big discrepancy...  (Did you do a
complete recompile before and after the patch?)  With your test, timed with
microtime() on my old Win2k system (release/non-debug), I get 1.56s before
and 1.54s after -- about what I'd expect after doing my own random testing
previously.  It's not expected to be a big optimization, just eliminate the
overhead of the old INIT_STRING only allocating an empty string initially.
(Of course, the more ADD_STRING/ADD_VAR ops for a string, the less relative
difference there is.)  And saving an opcode makes creating/destroying op
arrays just a bit faster without a cache.

BTW, after recompiling the latest 5.3 code, after some changes in the last
days (mine had been a few days old), I observed a couple different things
after my changes: the binary size was 8K smaller (no change a few days ago),
and there was a surprising improvement on a lot of the bench.php tests (I
think "ackermann" is the only one that uses interpolated strings).  Just
compiler randomness I guess, but going from 27.9s to 27.1s is more than the
fluctuation I've seen before. :-)


- Matt


----- Original Message -----
From: "Felipe Pena"
Sent: Tuesday, May 06, 2008

> Hi Matt,
> about optimization..., do you have any test for comparison? Why, in my
> crazy test (http://rafb.net/p/ZzQQQP97.html), the actual code is more
> faster than your patch.
>
> real    0m1.156s
> vs
> real    0m1.368s
>
> (using time command)
>
>
> I'm just curious, i'm not against your patch.
>
> Thanks.
> --
> Regards,
> Felipe Pena.