In response to Xor and Slava:
I recommend you read up a bit more on the way PHP handles memory management. Take the following code for example:
<?php
$data = $_POST['lotsofdata'];
$data2 = $data;
$data3 = $data;
$data4 = $data;
$data5 = $data;
?>
Assuming we post 10MB of data to this PHP file, what will PHP do with the memory?
PHP uses a table of sorts that maps variable names to the data that variable refers to in memory. The $_POST superglobal will actually be the first instance of that data in the execution, so it will be the first variable referenced to that data in the memory. It will consume 10MB. Each $data var will simply point to the same data in memory. Until you change that data PHP will NOT duplicate it.
Passing a variable by value does just what I did with each $data var. There is no significant overhead to assigning a new name to the same data. It is only when you modify the data passed to the function that it must allocate memory for the data. Passing a variable by reference will do essentially the same thing when you pass the data to the function, only modifying it will modify the data that is in the memory already versus copying it to a new location in memory.
If for learning purposes you choose to disregard the obvious pointlessness in benchmarking the difference between these two methods of passing arguments, you will need to modify the data when it is passed to the function in order to obtain more accurate results.
