[PATCH] mod_include -- Only evaluate the environmentvariables wh en needed - #2

[Ian Holsman <Ia...@cnet.com>]

here is a update of the patch, (the old one had a bug in it when you used a env variable in a virtual include)

My testing on linux and solaris shows an overall improvement in times and machine load
(I've included some output from AB from the linux test)

David Reid's testing shows that it is actually slower on FreeBSD 4, but I don't have access to FreeBSD to 
find out why

> > 
> > Ian,
> > 
> > Bad news.  I've added the patch, rebuilt and don't see any 
> > significant increase
> > in speed.  In fact I'm actually seeing a very small drop.  
> > The system is
> > FreeBSD 4 STABLE running on a PIII 700 using localhost.  2 
> > test files, one with
> > only an include (test 2) and another with the same include, 
> > printenv and echo
> > (test 1) commands.  The difference goes like...
> > 
> >                 before            after
> > test 1        480-490        455-475
> > test 2        590-600        560-580
> > 
> > Not sure why there should be a drop, but it could be 
> > something to do with the
> > extra table calls that your patch adds, or seems to.
> > 
> > set_lazy_vars has a call to apr_setn and we call it a few 
> > times now that we
> > didn't before...  table operations can be expensive so this 
> > could be part of
> > the problem.  I guess it depends on your system ??  Just for 
> > my information,
> > why do you make the call to add the table variables?
> > 
> > So, I'm not going to add the patch, sorry.  I like the idea, 
> > but for some
> > reason it isn't quite working...  If you want to try again 
> > just let me know.
> > 
> >> 

Re: [PATCH] mod_include -- Only evaluate the environmentvariables when needed - #2

[David Reid <dr...@jetnet.co.uk>]

> here is a update of the patch, (the old one had a bug in it when you used
a env variable in a virtual include)
>
> My testing on linux and solaris shows an overall improvement in times and
machine load
> (I've included some output from AB from the linux test)
>
> David Reid's testing shows that it is actually slower on FreeBSD 4, but I
don't have access to FreeBSD to
> find out why

To clarify it wasn't by much (< 10%).  I've also seen some other bizarre
results when switching on mod_include which I want to check into.  Off to
deepest, darkest Africa again for a few days so should have time :)

david

>
> > >
> > > Ian,
> > >
> > > Bad news.  I've added the patch, rebuilt and don't see any
> > > significant increase
> > > in speed.  In fact I'm actually seeing a very small drop.
> > > The system is
> > > FreeBSD 4 STABLE running on a PIII 700 using localhost.  2
> > > test files, one with
> > > only an include (test 2) and another with the same include,
> > > printenv and echo
> > > (test 1) commands.  The difference goes like...
> > >
> > >                 before            after
> > > test 1        480-490        455-475
> > > test 2        590-600        560-580
> > >
> > > Not sure why there should be a drop, but it could be
> > > something to do with the
> > > extra table calls that your patch adds, or seems to.
> > >
> > > set_lazy_vars has a call to apr_setn and we call it a few
> > > times now that we
> > > didn't before...  table operations can be expensive so this
> > > could be part of
> > > the problem.  I guess it depends on your system ??  Just for
> > > my information,
> > > why do you make the call to add the table variables?
> > >
> > > So, I'm not going to add the patch, sorry.  I like the idea,
> > > but for some
> > > reason it isn't quite working...  If you want to try again
> > > just let me know.
> > >
> > >>
>
>