PHP dynamical value as time() function speed in bigdata

What is the preferred (less memory consuming and fastest) approach to use time() or any similar dynamic value in billion+ iterations?

A)

$time = time();
foreach($_billion_items_array as $_i => $_v) {
  if($_v['time_saved'] < $time + rand(1,100)) {
    // do something
  }
}

B)

$this->time = time();

foreach($this->$_billion_items_array as $_i => $_v) {
  if($_v['time_saved'] < $this->time + rand(1,100)) {
    $this->do_something(_v);
  }
}

C)

$this->time = time();

function fixTime($_correction) {
  return $this->time + $_correction;
}

foreach($this->$_billion_items_array as $_i => $_v) {
  if($_v['time_saved'] < $this->fixTime(rand(1,100)) {
    $this->do_something(_v);
  }
}

I would personally prefer C) but I don’t know how will PHP use memory, if every iteration is storing time as variable? Is it the same in A) and B)?

Answer

A is surely the fastest, because it uses the simplest way to access your $time variable inside your loop.

C is by far the slowest, because it must invoke your function on every iteration of the loop.

All your choices use roughly the same amount of RAM.

If you were doing 10^3 iterations, none of this would make much difference. But you are doing 10^9 interactions so you should simplify the code in your loop as much as you possibly can.

And, I think you want foreach() in place of for().