Monday, January 06, 2014

C++ unsigned arithmetic gotcha

Recently I came across a piece of code that looked similar to the following:


int main() {
  unsigned int a, b;

  a = 20;
  b = 30;

  double c = double(a - b);

  printf("%lf\n", c);


One may, at first glance assume that this would spit out the required -10.0, however this is not the case. It will print out a huge number something like 4294967286.0 !

The reason for this is that the datatype of a and b are unsigned types. And at least with the GCC compiler (or for that matter, I assume any compiler), the intermediate variable that is generated would be an unsigned type, resulting in "mis-interpretation" of the sign bit when the casting operation is performed. 

In other words, when a subtraction is performed on unsigned quantities, you should better be careful to see what actually is happening! 

1 comment:

Unknown said...

Very Good Thought!!!
Keep it up