CODEWITHSUNDEEP
cfloating-point
Joyce Babu
Joyce Babu

Reputation: 20674

Equality of a floating point variable initialised to zero in C

I understand that equality operation should not be performed on floating point numbers due to the way the value is stored, and instead a threshold comparison should be performed. But is it safe to use equality when the value is set to a constant value, say 0?

void test() {
   double max = 0.0;

   do {
       if (condition) {
           if (max != 0.0) { // Is this safe
               // myval is always a large positive number
               max = min(myval, max);
           } else {
               max = myval;
           }
       }
   } while (my_condition);

   return myval;
}

I can see several similar questions on StackOverflow. But my question is specific to a variable assigned the constant zero (not the result of an operation).

Upvotes: 0

Views: 89

Answers (2)

juanchopanza
juanchopanza

Reputation: 227458

0.0 has an exact representation in IEEE-754, so if two variables are set to 0.0 their comparison should always yield true. But if in your code max is really always 0.0, you might as well remove the if block because it would never execute.

Upvotes: 2

ouah
ouah

Reputation: 145879

Assuming you are using IEEE-754 binary formats, yes it is safe as 0.0 can be represented exactly in IEEE-754. There is exactly one representation for +0 in binary formats (and one for -0 and note also that+0 = -0).

Upvotes: 2

Related Questions