-1
$\begingroup$

This question already has an answer here:

I am wondering what the time complexity of a for loop that increments the control variable, but also multiplies it inside the loop. For example

for (int k = 0; k < n; k++)
{
    stuff

    k *= 2;
}

I am thinking it must be something like O(log n), but I know that is not it. Any help would be appreciated.

$\endgroup$

marked as duplicate by Raphael Jan 31 at 21:49

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

1
$\begingroup$

It is $O(\log n)$. Simply take a look at the number of iterations your loop produces when $n$ goes up. I know this is not a proof, btw. You can observe that if $n = 2^p$, for some $p \geqslant 0$, the number of iterations inside the loop increases by one. Clearly, this happens only $O(\log n)$ times, and therefore your loop has $O(\log n)$ iterations.

Now a lightly more formal attempt, but still with a few holes here and there.

We observe that at the end of the loop $$k = 2^{\lfloor \log n \rfloor + 1} - 1.$$

Technically, we should prove this using a loop invariant or induction, but I'm a little too lazy for that now. If we let $n = 2^p$, for some integer $p \geqslant 1$, then we get that $\lfloor \log n \rfloor = \lfloor \log(n - 1) \rfloor + 1$. Let $k_n$ denote the value of $k$ after $n$ iterations. Then

$$k_n = 2^{p + 1} - 1 = 2^{\lfloor \log(n - 1) \rfloor + 2} - 1 > 2^{\lfloor \log(n - 1) \rfloor + 1} - 1 = k_{n - 1}.$$

So we conclude that $k_n$ has grown over $k_{n-1}$, but only if $n = 2^p$. Therefore, the number of iterations is $O(\log n)$.

Edit

A far more intuitive approach is to consider the code

int k = 0;
for (int i = 0; k < n; i++) {
  if (i = 1) {
    k = 1;  
  }
  k *= 2;
}

Clearly, your code causes $k$ to converge more quickly to $n$ than this example, and this loop finishes after $O(\log n)$ iterations. Clearly, your code will also at most use $O(\log n)$ iterations.

$\endgroup$
  • $\begingroup$ Nice answer. It seems the argument could be simpler. At the end of $i$-th iteration, $k=2^i-2$. It follows there are $\lceil\log_2(n+1)\rceil$ iterations. $\endgroup$ – Apass.Jack Jan 31 at 22:30

Not the answer you're looking for? Browse other questions tagged or ask your own question.