Thanks for clarifying my doubts - the steps below sounds right to me.
I was just considering the overall perf. of such php daemons and
whether we can get some free perf. boost
with that apc setting.
Ravi
On Dec 12, 2007 12:19 PM, Richard Lynch <ceo@xxxxxxxxx> wrote:
>
> On Wed, December 12, 2007 1:33 pm, Ravi Menon wrote:
> > We have long running daemons written in php ( cli, non-apache
> > contexts) with the typical pattern:
> >
> > while( !$shutdown )
> > {
> > $c = new SomeClass;
> >
> > $c->process();
> > }
> >
> > For performance reasons, would it help if apc.enable_cli is turned on:
> >
> > apc.enable_cli integer
> >
> > Mostly for testing and debugging. Setting this enables APC for the
> > CLI version of PHP. Normally you wouldn't want to create, populate and
> > tear down the APC cache on every CLI request, but for various test
> > scenarios it is handy to be able to enable APC for the CLI version of
> > APC easily.
> >
> >
> > I am slightly confused by the statement - 'Mostly for testing and
> > debugging.....' .
> >
> > On each loop iteration, does php recompile the code in 'SomeClass' (
> > and all its dependencies ) or it is really cached ( as it has seen the
> > class code once ).
> >
> > If there is a php internals document on such issues, do let me know.
>
> The following is almost-for-sure correct, but I wouldn't swear in
> court...
>
> Step 0. Read PHP/HTML source code.
> Step 1. PHP uses a 2-pass compiler and generates byte-code.
> Step 2. The byte-code is then feed to the executer.
> Step 3. Executer spews output (or crashes or whatever)
>
> APC and other caches add a Step 1A., which stores the byte-code in RAM
> under the filename (or full path, depending on config) as a key.
>
> Therefore, adding APC will not affect in any way the "while" loop --
> It's compiled once in Step 1, and that's it.
>
> If you re-run the same script again and again, however, APC in CLI
> might be able to keep the script around and avoid a hit to the disk to
> LOAD the script (Step 0 above) as well as avoiding the 2-pass
> compilation to byte-code (Step 1 above).
>
> NOTE:
> Step 0 is the REALLY expensive step where APC et al are REALLY
> boosting performance.
>
> APC et al *could* just insert step 0A and store the PHP source, and
> have ALMOST the same benefits.
>
> However, storing the compiled version at Step 1A gets you some free
> gravy in not re-compiling the PHP source to byte-code, so they do that
> because, well, it's essentially FREE and saves a few more cpu cycles.
>
> But the REAL boost, again, is from not hammering the hard drive (slow)
> to load PHP source into RAM, Step 0.
>
> PS
> If you are really concerned about the constructor of SomeClass being
> expensive, time it and see.
>
> YOU may be doing something incredibly expensive there like
> re-connecting to the database (slow!) each time.
>
> You may also not even NEED a whole new SomeClass every time -- Perhaps
> you could just make a singleton and then reset its values and call
> process() with the new values instead of building up a whole new
> instance each time.
>
> --
> Some people have a "gift" link here.
> Know what I want?
> I want you to buy a CD from some indie artist.
> http://cdbaby.com/from/lynch
> Yeah, I get a buck. So?
>
>
--
PHP General Mailing List (http://www.php.net/)
To unsubscribe, visit: http://www.php.net/unsub.php
