Looks like calling an empty function is faster than checking if a function exists. Could be handy designing frameworks.

results.txt

3.93 ms
1.68 ms

test.php

<?php

profiler::start('function_exists');
for ($x = 0; $x < 10000; $x++) {
  function_exists('i_dont');
}
echo text::seconds(profiler::stop('function_exists')->time->delta)."\n";

function foo() {}

profiler::start('empty');
for ($x = 0; $x < 10000; $x++) {
  foo();
}
echo text::seconds(profiler::stop('empty')->time->delta)."\n";