Newsgroups: php.internals
Path: news.php.net
Xref: news.php.net php.internals:36164
Mailing-List: contact internals-help@lists.php.net; run by ezmlm
Received-SPF: pass (pb1.pair.com: domain zend.com designates 212.25.124.162 as permitted sender)
Content-class: urn:content-classes:message
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Date: Mon, 17 Mar 2008 10:07:41 +0200
Message-ID: <06B0D32C7A96544490D18AF653D6BDE502759909@il-ex1.zend.net>
Thread-Topic: [PHP-DEV] Patch for opcode caches
Thread-Index: AciF+EEug5sQh3NJT2+izBve+g/gWQCDQT7Q
To: "Marcus Boerger" <helly@php.net>,
	"Stas Malyshev" <stas@zend.com>
Cc: "Andi Gutmans" <andi@zend.com>,
	"phpxcache" <phpxcache@gmail.com>,
	<internals@lists.php.net>
Subject: RE: [PHP-DEV] Patch for opcode caches
From: dmitry@zend.com ("Dmitry Stogov")



> -----Original Message-----
> From: Marcus Boerger [mailto:helly@php.net]=20
> Sent: Friday, March 14, 2008 8:24 PM
> To: Stas Malyshev
> Cc: Dmitry Stogov; Andi Gutmans; phpxcache; internals@lists.php.net
> Subject: Re: [PHP-DEV] Patch for opcode caches
>=20
>=20
> Hello Stanislav,
>=20
> Friday, March 14, 2008, 5:51:49 PM, you wrote:
>=20
> >> But that means we are compiling inheritance when a file is loaded=20
> >> form the cache. The goal should be to compiling inheritance when=20
> >> writing to the opcode cache file. All we achieve here is a=20
> slow down.=20
> >> If there is
>=20
> > You can not do that. You do not know that until runtime.
>=20
> >> something that makes this required than at least we need=20
> to warn the=20
> >> user about the slow code. I Think what we need to do is=20
> deprecating=20
> >> class and
>=20
> > It won't be slow. It would be much faster (in some cases=20
> 10x faster).
> > This specific assembly of instructions might be slower in=20
> some cases,=20
> > but the code altogether will be faster. That's the idea of=20
> the opcode=20
> > caches.
>=20
> >> namespace use in non main blocks as well as include and require in=20
> >> non main blocks. And if there is no such case than=20
> everything can be=20
> >> bound early which is the fast state we should aim for.
>=20
> Lemme just think, doing inheritance at compile time before we=20
> send the stuff to an opcode cache can be slower then?

It's always faster, but with opcode caches we cannot always do it in
compile-time, because the parent (in another file) might be changed on
next load of children class.

> How is=20
> that possible? After all late binding means we do it at run=20
> time. And no matter how much faster we can do it. It will=20
> always be slower than doing the same thing only once.
>=20
> > We do not need to deprecate anything, and inheritance can=20
> not be bound
> > before parent class is known, which in the case of bytecode=20
> cache means=20
> > - not before the file is loaded from the cache. That's=20
> exactly what this=20
> > patch enables. It does not change performance=20
> characteristics of neither=20
> > cached not non-cached code, it just makes it much simpler=20
> and working in=20
> > all cases.
>=20
> Inheritance cannot be done before the parent class is known.=20
> We still need to do the prototype checks even if we assume=20
> the class is there and insert a virtual class as parent somehow.

Right. This is the reason why we need to delay early binding. But we
don't like to delay it to run-time, because it will breake the class
creaton order and as result may breake some scripts.

Thanks. Dmitry.