| View previous topic :: View next topic |
| Author |
Message |
stanleyxu2005
Smarty Rookie
Joined: 31 Dec 2006
Posts: 5
|
 Posted: Sun Dec 31, 2006 2:44 pm Post subject: Can this make smarty faster? |
 |
|
I have made a comparation about loading templates.
blank template ~> 0.0065 second
template with lot of content ~> 0.0135 second
It seems that to load "Smarty.class.php" takes much time.
The file smarty.class.php and smarty_compiler is roughly 140kb.
Does it make sense to remove all comments from smarty library, in order to make smarty library more lightweight?
I have just ported my website to smarty-template based. Each page requires at least 0.0135 second (previously was 0.002 second) This is a big difference. And the cpu usage is also much more. (I am afraid, when many users are viewing website at a same time, the server will be overload.)
You can see the difference at
http://gosurfbrowser.com/?&ln=en (pages are generated dynamically with simply php code)
http://gosurfbrowser.com/v2/?&ln=en (with smarty-template)
The configuration code is:
$tpl->compile_check=true;
$tpl->debugging=false;
other settings are used default values. |
|
| Back to top |
|
Pap
Smarty Regular
Joined: 21 Jun 2006
Posts: 69
Location: Denver, CO
|
| Posted: Sun Dec 31, 2006 8:20 pm Post subject: |
|
|
I suggest you use this:
http://php.net/manual/en/ref.apc.php
It will keep your most commonly requested PHP files cached in memory in bytecode format. Compiling to bytecode automatically removes comments, whitespace, etc, and leaves only pure computer instructions.
_________________
Don't be stupid, be a Smarty™.
Come and join the P-H-Party. |
|
| Back to top |
|
stanleyxu2005
Smarty Rookie
Joined: 31 Dec 2006
Posts: 5
|
| Posted: Sun Dec 31, 2006 8:59 pm Post subject: |
|
|
Thanks.
I have found a comment remover, but the difference is not obvious.
http://www.devpro.it/remove_phpcomments/
APC is a good solution, but it is required to be installed on server. This is not always possible. The better way is to find the bottleneck in Smarty. |
|
| Back to top |
|
Hielke Hoeve
Smarty Elite
Joined: 06 Jan 2006
Posts: 406
Location: Netherlands
|
| Posted: Mon Jan 01, 2007 7:36 pm Post subject: |
|
|
Comments in files are not a significant factor in the loading of php files.
If you use alot of filters and such the page loading time will indeed be somewhat higher, have you tried using caching?
_________________
Debug XHTML Compliance
SmartyPaginate
Smarty License Questions
---
(About Unix) The learning curve is full of aha! moments, such as that glorious day that the full beauty of grep and, later, find is revealed in all its majesty. --- Robert Uhl <ruhl@4dv.net> |
|
| Back to top |
|
stanleyxu2005
Smarty Rookie
Joined: 31 Dec 2006
Posts: 5
|
| Posted: Wed Jan 03, 2007 5:19 pm Post subject: |
|
|
I have not used any plugins at all.
About caching: the difference is not that much. I have mentioned, if I just create an instance without loading any template. It takes already 0.0065s. (If I generate a Web page with PHP code manually, it takes only 0.002s) This is a big difference.
My machine:
+ Intel Core Duo E6400
+ 2GB DDR 667
+ 320GB SATA 2.0
+ Windows XP SP2 with all patches.
+ PHP 4.4.2 (with Zend optimizier)
I think my machine is fast enough, but it still takes roughly at least 0.0135s to create a simple Web page. Will Smarty bring any performance issue, when many users are navigating web site at the same time? |
|
| Back to top |
|
stanleyxu2005
Smarty Rookie
Joined: 31 Dec 2006
Posts: 5
|
| Posted: Wed Jan 03, 2007 5:23 pm Post subject: |
|
|
The code of my benchmark timer
| Code: |
class TmxTimer {
var $startTime;
var $totalTime;
//
// Constructor
//
function TmxTimer() {
$this->totalTime= 0;
$this->start();
}
/**
* Gets the current timestamp
* @return float Timestamp
*/
function _getMicrotime() {
list ($usec, $sec)= explode(' ', microtime());
return ((float) $usec + (float) $sec);
}
/**
* Starts timer
*/
function start() {
$this->startTime= $this->_getMicrotime();
}
/**
* Stops timer
* @return string Total duration
*/
function stop() {
$duration= $this->_getMicrotime() - $this->startTime;
$this->totalTime += $duration;
return sprintf('%.4f', $duration);
}
}
|
My index.php
| Code: |
include_once ('timer.php');
$timer= new TmxTimer();
$timer->start();
include_once ('smarty/Smarty.class.php');
$tpl= new Smarty;
/* display time and close */
echo $timer->stop();
|
The web page displays a time between 0.0044s and 0.0060s
To note that: If I create a Web page (like http://gosurfbrowser.com/), with
+ multilanguage support, with
+ dynamic content (load my own designed templates, data files, etc)
It takes only 0.002s
 |
|
| Back to top |
|
messju
Administrator
Joined: 16 Apr 2003
Posts: 3336
Location: Oldenburg, Germany
|
| Posted: Thu Jan 04, 2007 8:13 am Post subject: |
|
|
| stanleyxu2005 wrote: | | Will Smarty bring any performance issue, when many users are navigating web site at the same time? |
Probably. Test it. |
|
| Back to top |
|
Hielke Hoeve
Smarty Elite
Joined: 06 Jan 2006
Posts: 406
Location: Netherlands
|
| Posted: Thu Jan 04, 2007 2:10 pm Post subject: |
|
|
With page loads in the 100s and 1000s of a second you will probably need alot of users and/or webpages that do lots of things. especially with a dual core processor.
Personally you should only be worried when loading times start to rise upto (half) a second when alot of users are using your site (100+?). Research shows users are often content when pages get loaded within a few second, this is ofcourse including client side rendering. Ofcourse it is good practice to prevent something rather than fixing it afterwards.
_________________
Debug XHTML Compliance
SmartyPaginate
Smarty License Questions
---
(About Unix) The learning curve is full of aha! moments, such as that glorious day that the full beauty of grep and, later, find is revealed in all its majesty. --- Robert Uhl <ruhl@4dv.net> |
|
| Back to top |
|
stanleyxu2005
Smarty Rookie
Joined: 31 Dec 2006
Posts: 5
|
| Posted: Thu Jan 04, 2007 2:30 pm Post subject: |
|
|
| Hielke Hoeve wrote: | | With page loads in the 100s and 1000s of a second you will probably need alot of users and/or webpages that do lots of things. especially with a dual core processor. |
dual core processor is my test machine. I have rent a virtual host as web server. It is shared with many other customers. 0.0135s is tested with my machine. On the real web server, it performs much slower. Usallay 0.02-0.05s.
I just wondered, even when I do nothing with smarty, just including it. It takes much time. It does much in background. Are they really required?
I have seen some other template engines: smarty light, template lite. They say they are faster than smarty. My question is that I just use the very basic features of smarty, those are supported by other template engines as well. Why smarty is slower? |
|
| Back to top |
|
messju
Administrator
Joined: 16 Apr 2003
Posts: 3336
Location: Oldenburg, Germany
|
| Posted: Thu Jan 04, 2007 2:35 pm Post subject: |
|
|
| stanleyxu2005 wrote: | I just wondered, even when I do nothing with smarty, just including it. It takes much time. It does much in background.
|
look into the source it what it does:
| Code: |
/**
* DIR_SEP isn't used anymore, but third party apps might
*/
if(!defined('DIR_SEP')) {
define('DIR_SEP', DIRECTORY_SEPARATOR);
}
/**
* set SMARTY_DIR to absolute path to Smarty library files.
* if not defined, include_path will be used. Sets SMARTY_DIR only if user
* application has not already defined it.
*/
if (!defined('SMARTY_DIR')) {
### if you think the dirname system call is too expensive then define SMARTY_DIR before
define('SMARTY_DIR', dirname(__FILE__) . DIRECTORY_SEPARATOR);
}
if (!defined('SMARTY_CORE_DIR')) {
define('SMARTY_CORE_DIR', SMARTY_DIR . 'internals' . DIRECTORY_SEPARATOR);
}
define('SMARTY_PHP_PASSTHRU', 0);
define('SMARTY_PHP_QUOTE', 1);
define('SMARTY_PHP_REMOVE', 2);
define('SMARTY_PHP_ALLOW', 3);
/**
* @package Smarty
*/
class Smarty
{
... here come only var and method definitions but *no* execution of anything
}
|
| Quote: | | Are they really required? |
yes. |
|
| Back to top |
|
|
