# Dividing 2 integers with some constraints

This a problem i came across while practicing binary search. Here is the problem:

Given two integers dividend and divisor, divide two integers without using multiplication, division and mod operator.

Return the quotient after dividing dividend by divisor. The integer division should truncate toward zero.

Note:

1. Both dividend and divisor will be 32-bit signed integers.
2. The divisor will never be 0.
3. Assume we are dealing with an environment which could only store integers within the 32-bit signed integer range: [−2^31, 2^31 − 1]. For the purpose of this problem, assume that your function returns 2^31 − 1 when the division result overflows.

A Brute force Solution is that subtract the dividend with the divisor till it is greater and the number of subtractions is the result. But it is giving Time Limit Exceeding error.

How to solve the problem efficiently or using Binary Search ??

Also provide the time complexity as well.

• All time complexities of reasonable algorithms will be $O(1)$ since you are only dealing with finitely many inputs. Commented Nov 2, 2019 at 10:49
• – D.W.
Commented Nov 3, 2019 at 6:07

Here is a strategy (I will only consider positive numbers): Let $$d$$ be the dividend and $$x$$ be the divisor. Generate all values $$x_i = 2^i x$$, up to some some $$x_k$$ such that $$x_{k+1}$$ exceeds the dividend. This can be done with only one addition per value since $$x_{0} = x$$ and, for $$i \ge 1$$, $$x_i = x_{i-1} + x_{i-1}$$. Similarly, generate all values $$b_i = 2^i$$ for $$i=0, \dots, k$$.

Let $$r$$ be a variable that will hold the result. Initially $$r=0$$. For $$i=k$$ down to $$0$$ do the following:

• Check whether $$x_i$$ is bigger than $$d$$;
• If that is the case, then you know that, by subtracting $$x_i$$ from $$d$$, you are effectively subtracting $$x$$ from $$d$$ a total of $$b_i = 2^i$$ times. Update $$d = d- x_i$$, and $$r = r + b_i$$.

Finally, return $$r$$.

This strategy requires only a logarithmic number of operations w.r.t. $$d/x$$ (up to multiplicative and additive constants). Since this is at most $$2^{31}$$, the time needed is always upper bounded by a constant.

As an example, let's divide $$62$$ by $$3$$. The sequences of values $$x_i$$ will be: $$x_0 = 3, x_1 = 6, x_2 = 12, x_3 = 24, x_4 = x_k = 48$$, since $$x_5 = 96 > 62$$. The corresponding values $$b_i$$ are: $$b_0 = 1, b_1=2, b_2=4, b_3=8, b_4=16$$.

• Initially $$d=62$$, $$x=3$$, $$r=0$$.

• In the fist iteration ($$i=k=4$$) we have $$62 = d \ge 48 = x_4$$, and we update: $$d = 62 - x_4 = 62-48 = 14$$, and $$r = 0 + b_4 = 0 + 16 = 16$$.

• In the second iteration ($$i=3$$) we do nothing since $$d = 14 \not\ge 24 = x_3$$.

• In the third iteration ($$i=2$$) we have $$d = 14 \ge 12 = x_2$$, and we update $$d = 14 - x_2 = 14 - 12 = 2$$, and $$r = 16 + b_2 = 16 + 4 = 20$$.

• In the forth iteration ($$i=1$$) we do nothing since $$d = 2 \not\ge 6 = x_1$$.

• In the fifth and final iteration ($$i=0$$) we do nothing since $$d = 2 \not\ge 3 = x_0$$.

In the end we have $$r=20$$ and $$d=2$$. Indeed: $$62 = 3 \cdot 20 + 2$$.

• ,Appreciate your answer. if you give any example, it would be better.Thanks. Commented Nov 2, 2019 at 11:41
• I added an example. Commented Nov 2, 2019 at 13:25
• Understood now :). However the time complexity would be log(d/x)?? Commented Nov 2, 2019 at 13:32
• Yes, that's what I wrote. The number of iterations is proportional to the value of $k$, which is in $\Theta( \log \frac{d}{x})$ for arbitrary $x$ and $d$ with $x \le d$. However, since you restrict $d$ to be at most $2^{31}$, $k$ will always be at most $31$. Commented Nov 2, 2019 at 13:35
• Thanks. I am posting another question to determine the time-complexity.Kindly check once (Similar to your solution). Link: cs.stackexchange.com/questions/116591/… Commented Nov 2, 2019 at 13:44

With the help of Steven, i am posting the solution.

def divide(dd,dr):
'''
let dd and dr be the dividend and divisor
x be the current macium divisor, less that dividend
c be the counter
q is the qutotient
'''
x=dr
c=0
while (x<<c) <= dd:
c+=1
print(c)
q=0

#then subtract from dividend and update result as usual manner

#will run c time i.e lgx time  -----------Loop2
for j in range(c-1,-1,-1):
if x<<j <= dd :
dd-=(x<<j)
q+=1<<j

print(q,dd)

'''
total time complexity will be 2lgx i.e lgx
'''


** Just Little bit doubt in the time complexity of loop1 and loop2.**

But this solution works.

• To figure out the time complexity you just need to bound the value of variable $c$ at the end of the first loop. What is the value $x_c$ of $\texttt{x << c}$ after the $c$-th iteration of the first loop? What is the minimum value of $c$ such that the condition of the loop is not satisfied anymore, i.e., $x_c > dd$? Commented Nov 3, 2019 at 17:42

The best strategy when given a problem with stupid restrictions is to ignore the restrictions and write z = x / y. As it is, solving this problem doesn't help you understanding anything, and doesn't help you doing anything.

• Some processors don't have these instructions and that's why the question has these restrictions Commented Mar 31, 2022 at 9:00