Re: [PHP4BETA] Wacky Performance Problem w/ Windows

On Thu, 1 Jun 2000, Joel Roller wrote:

> > > Leaving aside the fact that IIS4/isapi php is so much slower for a
> > > sec..  Php4 on windows parses php (stuff within <? tags ?>) between 2
> > > and 5 times faster than html.

> > > I have tried IIS4/isapi, IIS4/cgi, Apache/cgi, all on windows, and get
> > > the same results:  File 2 is much faster than File 1.  And all are
> > > around 50 times slower than the same file on linux/apache/mod_php.

>> What happens if you set up IIS4 with neither PHP ISAPI, nor PHP cgi?  Are the
>> speeds for flat HTML files still comparable to what you see when PHP is
>> loaded?
>>

> I'm sorry, I've got a way of writing confusing things.   Of course, I meant to
> say that I'm running all these files through a .php extension bound to the php
> engine.   The raw performance of .html files is not an issue -- I haven't even
> bothered to test such things as I couldn't imagine them being affected.  I don't
> plan on using many of them in the near future, anyway, if I could only get php
> working properly.

> My problem is that the raw html in a .php page seems to be gumming up the entire
> parsing or scanning, or whatever it is that php does when its reading through a
> page.  I've got a homepage with 50k of html intermixed with a few dozen php
> commands here and there, database calls, etc.   First thing I did was take out
> the huge blocks of html, and put them in seperate files, called with
> readfile().   That more than doubled the speed of the page, then putting all
> other blocks of html within print statements doubled it again.  It seems to me,
> that none of this messyfussin around should be necessary.

Hmm. Well, that makes a little more sense, at least.  I still don't pretend to
understand why there would be such performance problems under Win32.  I do see
some Win32-specific code in main.c (UpdateIniFromRegistry)... is the registry
really that slow under Windows?  Even if it were, I don't see any reason why
plain HTML would be that much slower than PHP.  I guess this would have to be
a bug in the Zend compiler?

Steve Langasek
postmodern programmer