C++ Program Giving Different Results on Different Computers

Question

My code is printing 192 for the input 100 (which is the desired result). But when I submit it on the online judge, it shows that the output of my program for the input 100 is 190. I copied and pasted the code in ideone.com and for input 100 I got result 192. I send it to my friend and on his PC the output is 190. But he also submitted the code onto ideone.com and got 192. What's the problem? Here's my code:

#include <bits/stdc++.h>
using namespace std;

typedef long long int lli;

int main(){
   lli in,ans = 0;
   cin >> in;

   if(in < 10)
       cout << in << endl;
   else{
      lli digits = 0;
      lli temp = in;
      while(temp > 0){
         digits++;
         temp /= 10;
      }
     digits--;

     while(in > 0){
         //cout << "in: " << in << endl;
         //cout << "digits: " << digits << endl;

         ans += ((in - (pow(10,digits) - 1)) * (digits + 1));
         in = in -  (in - (pow(10,digits) - 1));
         digits--;

         if(in == 9){
            ans+= 9;
            break;
         }
      } 
      cout << ans << endl;
   }
}

ideone link: http://ideone.com/zOvHzW

Why is this happening? I understand that it may be a compiler issue but what is really going on in here?

Read CodeChef guidelines on how to process input and output! :) — Ajay, Jun 09 '16 at 18:09
Side note, [don't `#include `](https://stackoverflow.com/questions/31816095/why-should-i-not-include-bits-stdc-h) — Cory Kramer, Jun 09 '16 at 18:09
I found that on first iteration (when `in` is `100`) some compilers evaluate `in = in - (in - (pow(10,digits) - 1));` as `98` and some as `99`(which is correct). I´m not sure why, I think that it will have to be something with optimization, but I'm not sure. — PcAF, Jun 09 '16 at 18:35
I wrote a custom power function, and the solution is accepted! What was wrong in pow() function? What is going on when I call the pow()? I solved the problem, but I still don't know what's wrong here. — Redwanul Sourav, Jun 09 '16 at 18:49
related: http://stackoverflow.com/questions/17063102/differences-in-rounded-result-when-calling-pow — kmdreko, Jun 09 '16 at 19:00

score 1 · Accepted Answer · answered Jun 09 '16 at 18:56

The problem is that std::pow is not exact. It is an approximating algorithm. The floating point result that it returns may be off by some, very tiny amount.

When you convert the floating point number to an integer, you round it down by removing the fractional part. But, if the correct result would have been say 100, and the error was -0.000...001 and therefore the result was 99.999...999 and you cut off the fractional part, then the integer that you get is 99.

How std::pow is implemented is... specific to the implementer. Therefore the error may be one way in one compiler / computer, but the other way in another. That is why the result may differ.

Solution: Don't use std::pow for integer calculation.

Workaround: Round the result to nearest integer first.

@RedwanulSourav no need to thank, that is what the upvote button is for. And don't forget to mark the most helpful answer as accepted :) — eerorika, Jun 09 '16 at 19:03

C++ Program Giving Different Results on Different Computers

1 Answers1