CareerCup

without parallel processing naive method is :-
simply % 2 to increase the count of 1 and then divide by 2(basically checking the last bit of the no till Most significant bit comes to last bit)

while(no!=0)
{
if(no%2 == 1)count++;
no = no/2;
}

With parallel processing ideal logic must be divide total no bits that your no can have into different set and parallel process each one of them

for eg if you have a 64 bit no divide 64 bit into 8 set of 8 bits

no1 = no&(2^8-1 <<0)
no2 = no&((2^8-1) << 8)
no3 = no&((2^801) << 16).
.
.
so on till no8

now for each no calculate separately in each processor the no of 1's and then add them.
..

count(int num)
{ int counter=0;

while(num>0)
{
num=num & num-1;
counter++;
}

}

Wht about hadoop ? even GB's of data

#include<iostream>

using namespace std;

int main()
{
int n,count=0;
cin>>n;
while(n!=0)
{
count++;
n=n&(n-1);
// cout<<n<<endl;
}

cout<<count;
return(0);
}

then how to make parallel call?

count the length create two or more start points like start comparing from 0th index to forwrd and nth index backward stop whre indexes r equal..(this way many start points can be created and can be executed simultaneously in threads)
run using thread to speed up