PHP XOR performance

Question

I am aware that XOR is the same as not equal, but I want to ask for the reason for the performance difference here. I know it is negligible, but I am curious to know why that happens.

Analysis

My own tests:

$ php -r '$a=true; $b=true;
$start = microtime(true);
for($i = 0; $i < 0xFFFFFF; $i++) $a xor $b;
$end = microtime(true);
echo ($end - $start) / 0xFFFFFF;
2.8898769545694E-8
$ php -r '$a=true; $b=true;
$start = microtime(true);
for($i = 0; $i < 0xFFFFFF; $i++) $a !== $b;
$end = microtime(true);
echo ($end - $start) / 0xFFFFFF;
'
2.735811385077E-8
$ php -r '$a=true; $b=true;
$start = microtime(true);
for($i = 0; $i < 0xFFFFFF; $i++) $a != $b;
$end = microtime(true);
echo ($end - $start) / 0xFFFFFF;
'
3.2480544635878E-8
$ php -r '$a=true; $b=true;
$start = microtime(true);
for($i = 0; $i < 0xFFFFFF; $i++) $a xor $b;
$end = microtime(true);
echo ($end - $start) / 0xFFFFFF;
'
2.9041645487517E-8
$ php -r '$a=true; $b=true;
$start = microtime(true);
for($i = 0; $i < 0xFFFFFF; $i++) $a !== $b;
$end = microtime(true);
echo ($end - $start) / 0xFFFFFF;
'
2.7436373032351E-8
$ php -r '$a=true; $b=true;
$start = microtime(true);
for($i = 0; $i < 0xFFFFFF; $i++) $a != $b;
$end = microtime(true);
echo ($end - $start) / 0xFFFFFF;
'
3.2506237862734E-8

Average:

• xor: 28.9 ns
• !==: 27.4 ns
• !=: 32.5 ns

I would like to ask:

Why do these operators have such significant performance?

I have actually done these tests a few more times, and !== is always faster than xor, and both are significantly (more than 10%) faster than !=.

Answer 1

VLD output from PHP 5.6.0 for a simple test of the three options:

$a=true; $b=true;
$x = $a xor $b;

echo '-';

$a=true; $b=true;
$x = $a !== $b;

echo '-';

$a=true; $b=true;
$x = $a != $b;

gives

Finding entry points
Branch analysis from position: 0
Jump found. Position 1 = -2
filename:       /in/OLfo9
function name:  
number of ops:  16
compiled vars:  !0 = $a, !1 = $b, !2 = $x
line     #* E I O op                           fetch          ext  return  operands
-------------------------------------------------------------------------------------
   3     0  E >   ASSIGN                                                   !96, 
         1        ASSIGN                                                   !112, 
   4     2        ASSIGN                                           $5      !128, !96
         3        BOOL_XOR                                         ~6      $5, !112
         4        FREE                                                     ~6
   6     5        ECHO                                                     '-'
   8     6        ASSIGN                                                   !96, 
         7        ASSIGN                                                   !112, 
   9     8        IS_NOT_IDENTICAL                                 ~9      !96, !112
         9        ASSIGN                                                   !128, ~9
  11    10        ECHO                                                     '-'
  13    11        ASSIGN                                                   !96, 
        12        ASSIGN                                                   !112, 
  14    13        IS_NOT_EQUAL                                     ~13     !96, !112
        14        ASSIGN                                                   !128, ~13
        15      > RETURN                                                   1

So xor is indeed creating a temporary result that is subsequently freed