Re: [PHP4BETA] RFC: Request startup optimization
| From: | thies at digicol dot de | Date: | Fri, 11 Feb 2000 18:07:46 +0000 |
| Subject: | Re: [PHP4BETA] RFC: Request startup optimization | ||
| References: | 1 2 | Groups: | php.version4 |
| Request: | Send a blank email to php-version4+get-10130@lists.php.net to get a copy of this message | ||
On Fri, Feb 11, 2000 at 07:59:50PM +0200, Zeev Suraski wrote:
> Generally I'm -1 for the change if there's no significant noticeable
> change. There's a considerable penalty for the shutdown process involved
> with this patch, and if the gain is unnoticeable, I don't see the point
> (that's what we arrived at in the PDM, as far as I recall).
zeev, let's wait for some "real-life" benchmarks (if there is
such thing). in general i would say that anything that
improves performance (and this patch is designed to do so) is
a GoodThing(tm).
again, i hope to do some more testing later.
BTW: i don't know what part of the patch will cause the
"considerable penalty for the shutdown"?
tc
>
> Zeev
>
> On Fri, 11 Feb 2000 thies@digicol.de wrote:
>
> > On Fri, Feb 11, 2000 at 04:50:26PM +0100, Sascha Schumann wrote:
> > > What?
> > >
> > > Defer the request startup until one of the module specific
> > > functions is used. Request shutdowns are avoided completely,
> > > if a module has a startup handler, but was not started up.
> > >
> > > Why?
> > >
> > > Module request startups (and respective shutdowns) can add
> > > (noticeable?) overhead to requests.
> > >
> > > How is it implemented?
> > >
> > > We use a stub function in the function_table which
> > >
> > > (a) calls the request_startup_func of the module
> > > (b) adds the module to the list of lazy_started_modules
> > > (c) sets all per-module function_table entries to their real
> > > handler
> > >
> > > Step (c) is reversed during module_registry_cleanup(), if the
> > > specific module was found in lazy_started_modules.
> > >
> > > Your comments and benchmark results are welcome.
> > >
> > > --
> > >
> > > Regards,
> > >
> > > Sascha Schumann
> > > Consultant
> >
> > my initial "hello world" benchmark showed no measurable jump
> > at all (could be me beeing stupid - will do some quantifying
> > early next week). im +1 for the change - what's your milage
> > sascha?
> >
>
> --
> Zeev Suraski <zeev@zend.com>
> http://www.zend.com/
>
--
Thies C. Arntzen "One Big-Mac, Small Fries and a Coke!"
Digital Collections Phone +49 40 235350 Fax +49 40 23535180
Hammerbrookstr. 93 20097 Hamburg / Germany