Compare performance of checking for existence 10,000 times using (a) function_exists or (b) a data structure cached in DB

gistfile1.php

<?php

civicrm_initialize();

$a = microtime(TRUE);

function_exists('abc');

$b = microtime(TRUE);

for ($i = 0; $i < 10000; $i++) {
  function_exists("abc_$i");
}

$c = microtime(TRUE);

// a reasonably large serialized structure
$settingsMetadata = CRM_Core_BAO_Cache::getItem('CiviCRM setting Spec', 'All', NULL);

$d = microtime(TRUE);

for ($i = 0; $i < 10000; $i++) {
  // isset() seems slightly faster than !empty()
  if (isset($settingsMetadata["abc_$i"])) {
    print "iteration $i\n"; // never actually prints, but we do the if() check 10k times
  }
}

$e = microtime(TRUE);

printf("%-35s: %.10f\n", 'a=>b: function_exists (1)', $b-$a);
printf("%-35s: %.10f\n", 'b=>c: function_exists (10k)', $c-$b);
printf("%-35s: %.10f\n", 'b=>c: function_exists (10k/10k)', ($c-$b) / 10000);
printf("%-35s: %.10f\n", 'c=>d: cache fetch', $d-$c);
printf("%-35s: %.10f\n", 'c=>d: cache check (10k)', $e-$d);
printf("%-35s: %.10f\n", 'c=>d: cache check (10k/10k)', ($e-$d) / 10000);
printf("%-35s: %.10f\n", 'c=>d: cache fetch+check (10k)', $e-$c);
printf("%-35s: %.10f\n", 'c=>d: cache fetch+check (10k/10k)', ($e-$c) / 10000);

/* Output on mylaptop:

a=>b: function_exists (1)          : 0.0000028610
b=>c: function_exists (10k)        : 0.0116860867
b=>c: function_exists (10k/10k)    : 0.0000011686
c=>d: cache fetch                  : 0.0022399426
c=>d: cache check (10k)            : 0.0058488846
c=>d: cache check (10k/10k)        : 0.0000005849
c=>d: cache fetch+check (10k)      : 0.0080888271
c=>d: cache fetch+check (10k/10k)  : 0.0000008089

*/

I think line 46 and 47 should be e=>d. Isn't it so?