sizeof vs count

Involved Joined: Apr 06, 2008 Posts: 484

I was mucking around with a new idea I had for a module when I realised that somewhere my script had started to take a long time.

So after much debugging I finally worked it out.I then decided to waste half a day mucking around benchmarking it for absolutely no other reason than pure stupidity. Sad

anyway this is what I found.......

I found the following example seems to be a lot faster;

Code:




$l = sizeof($arr);


for ( $i = 0; $i < $l; $i++ ) {


      $bf .= $arr[$i] . "\n";


  }

compared to this;

Code:




for ( $i = 0; $i < sizeof($arr); $i++ ) {


      $bf .= $arr[$i] . "\n";


  }

I found that you can gain upto 96% Smile

I then decided to do a number of comparisons between sizeof and count whilst running it through a 100000 loop.

Here is the script:

Code:




<?php





  function microtime_float()


  {


      list( $usec, $sec ) = explode( " ", microtime() );


      return ( (float)$usec + (float)$sec );


  }





  for ( $i = 0; $i < 100000; $i++ ) {


      $arr[] = $i;


  }





  $timeA1 = microtime_float();


  $bf = "";


  for ( $i = 0; $i < sizeof($arr); $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeA2 = microtime_float();


  $answerA = $timeA2 - $timeA1;





  print "<br />A: (sizeof) " . $answerA . "<br />\n";





  $timeB1 = microtime_float();


  $l = sizeof( $arr );


  for ( $i = 0; $i < $l; $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeB2 = microtime_float();


  $answerB = $timeB2 - $timeB1;





  print "<br />B: (sizeof) " . $answerB . "<br />\n";





  $timeC1 = microtime_float();


  $bf = "";


  for ( $i = 0; $i < count($arr); $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeC2 = microtime_float();


  $answerC = $timeC2 - $timeC1;





  print "<br />C: (count) " . $answerC . "<br />\n";





  $timeD1 = microtime_float();


  $bf = "";


  $l = count( $arr );


  for ( $i = 0; $i < $l; $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeD2 = microtime_float();


  $answerD = $timeD2 - $timeD1;





  print "<br />D: (count) " . $answerD . "<br />\n";





  echo '<br /><br /><br /><br />';


  $percent = $answerA - $answerB;


  $percent1 = $percent / $answerA;


  $answerAB = $percent1 * 100;





  echo '<strong>Difference Between A and B ' . $answerAB . '&nbsp;%</strong>';





  $percent = $answerC - $answerD;


  $percent1 = $percent / $answerC;


  $answerCD = $percent1 * 100;





  echo '<br /><br /><strong>Difference Between C and D ' . $answerCD . '&nbsp;%</strong>';





  $percent = $answerA - $answerC;


  $percent1 = $percent / $answerA;


  $answerAC = $percent1 * 100;





  if ( $answerAC < 0 ) {


      echo '<br /><br /><strong>Difference Between A and C <span style="color: red;">' . $answerAC . '&nbsp;%</span></strong>';


  } else {


      echo '<br /><br /><strong>Difference Between A and C ' . $answerAC . '&nbsp;%</strong>';


  }





  $percent = $answerB - $answerD;


  $percent1 = $percent / $answerB;


  $answerBD = $percent1 * 100;





  if ( $answerBD < 0 ) {


      echo '<br /><br /><strong>Difference Between B and D <span style="color: red;">' . $answerBD . '&nbsp;%</span></strong>';


  } else {


      echo '<br /><br /><strong>Difference Between B and D ' . $answerBD . '&nbsp;%</strong>';


  }





?>

My next question is........is there a gain to be had if we optimized all out for loops in ravennuke?

Or did I waste half a day as per usual Smile

Posted: Mon Feb 23, 2009 1:41 am

as a follow up there is also gains of 60% or more on loops of 10

registered · Posted: Mon Feb 23, 2009 6:32 am

Absolutely makes sense as in your "bad" example, it is having to execute the function call for "sizeof" for every pass through the loop. It is ALWAYS advisable to do what you have done.

You also should use pre-increment/decrement within your loops rather than post. It is faster.

Here is a nice blog post that I ran into years back: http://ilia.ws/archives/12-PHP-Optimization-Tricks.html

As far as RN is concerned, we have made some of these types of changes as I see them in the code, but its not like we have focused specifically on fixing loops.

Posted: Mon Feb 23, 2009 7:40 am

arrrrrrrrgh its nearly midnight.... how am I going to sleep now Smile

Posted: Mon Feb 23, 2009 5:54 pm

So, did you get any sleep? lol.

One of the biggest killers within *nuke are the SQL calls. So inefficient in how things are coded in many areas. We've tried to shore some of those up as we've run into them, but it takes a lot of work and every single call should really be analyzed - JMO.

For example, you will see allot a structure similar to this:

SQL call for a higher level category of data
Loop through result set
SQL call to go get related details for a given outer loop row
InnerLoop through result set
Endloop

What can be done many times, with just a little bit of smarter code, is to use an outer join on the two tables and just loop through the results once. I.e., let MySQL handle the "heavy lifting" of joining the data together. Can reduce the number of calls dramatically.

Another thing to consider are good, smart indexes. These haven't really been closely looked at either. So, more opportunity over and above an already well-performing RN system.

Site Admin/Owner Joined: Aug 27, 2002 Posts: 17088

There is no difference between sizeof and count, other than sizeof is an alias for count Smile

.

Also as an fyi, when benchmarking you need to make sure that all your variables are in the same state for each test. I have modified your code and added statements to make sure that the $i, $l, and $br variables are reset to their original states before the next test is run. In you particular benchmark tests the differences are more subtle but increase your loop up to 1,000,000 and they become more apparent. Also, if you feel the need for speed, try using foreach instead of the slower FOR loop Wink

Code:




<?php





  function microtime_float()


  {


      list( $usec, $sec ) = explode( " ", microtime() );


      return ( (float)$usec + (float)$sec );


  }





$arr = array(); //initialize array


  for ( $i = 0; $i < 100000; $i++ ) {


      $arr[] = $i;


  }








// Housekeeping


// $i  is used up above so it needs to be unset


// $bf is unset so it just needs to be initialized


  unset($i);


  $bf = '';





  $timeA1 = microtime_float();


  for ( $i = 0; $i < sizeof($arr); $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeA2 = microtime_float();


  $answerA = $timeA2 - $timeA1;





  print "<br />A: (sizeof) " . $answerA . "<br />\n";





// Housekeeping


// $i  is used up above so it needs to be unset


// $bf is used up above so it needs to be unset


// $bf is unset so it just needs to be initialized


// $l  is unset so it just needs to be initialized


  unset($i);


  unset($bf);


  $bf = '';


  $l  = 0;





  $timeB1 = microtime_float();


  $l = sizeof( $arr );


  for ( $i = 0; $i < $l; $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeB2 = microtime_float();


  $answerB = $timeB2 - $timeB1;





  print "<br />B: (sizeof) " . $answerB . "<br />\n";





// Housekeeping


// $i  is used up above so it needs to be unset


// $bf is used up above so it needs to be unset


// $l  is used up above so it needs to be unset


// $bf is unset so it just needs to be initialized


// $l  is unset so it just needs to be initialized


  unset($i);


  unset($bf);


  unset($l);


  $bf = '';


  $l  = 0;





  $timeC1 = microtime_float();


  for ( $i = 0; $i < count($arr); $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeC2 = microtime_float();


  $answerC = $timeC2 - $timeC1;





  print "<br />C: (count) " . $answerC . "<br />\n";





// Housekeeping


// $i  is used up above so it needs to be unset


// $bf is used up above so it needs to be unset


// $bf is unset so it just needs to be initialized


  unset($i);


  unset($bf);


  $bf = '';





  $timeD1 = microtime_float();


  $l = count( $arr );


  for ( $i = 0; $i < $l; $i++ ) {


      $bf .= $arr[$i] . "\n";


  }


  $timeD2 = microtime_float();


  $answerD = $timeD2 - $timeD1;





  print "<br />D: (count) " . $answerD . "<br />\n";





  echo '<br /><br /><br /><br />';


  $percent = $answerA - $answerB;


  $percent1 = $percent / $answerA;


  $answerAB = $percent1 * 100;





  echo '<strong>Difference Between A and B ' . $answerAB . '&nbsp;%</strong>';





  $percent = $answerC - $answerD;


  $percent1 = $percent / $answerC;


  $answerCD = $percent1 * 100;





  echo '<br /><br /><strong>Difference Between C and D ' . $answerCD . '&nbsp;%</strong>';





  $percent = $answerA - $answerC;


  $percent1 = $percent / $answerA;


  $answerAC = $percent1 * 100;





  if ( $answerAC < 0 ) {


      echo '<br /><br /><strong>Difference Between A and C <span style="color: red;">' . $answerAC . '&nbsp;%</span></strong>';


  } else {


      echo '<br /><br /><strong>Difference Between A and C ' . $answerAC . '&nbsp;%</strong>';


  }





  $percent = $answerB - $answerD;


  $percent1 = $percent / $answerB;


  $answerBD = $percent1 * 100;





  if ( $answerBD < 0 ) {


      echo '<br /><br /><strong>Difference Between B and D <span style="color: red;">' . $answerBD . '&nbsp;%</span></strong>';


  } else {


      echo '<br /><br /><strong>Difference Between B and D ' . $answerBD . '&nbsp;%</strong>';


  }





?>