cprecisionmodulusdouble-precision

How to use fmod and avoid precision issues

I'm going to boil this problem down to the simplest form:

Let's iterate from [0 .. 5.0] with a step of 0.05 and print out 'X' for every 0.25 multiplier.

for(double d=0.0; d<=5.0; d+=0.05) {
  if(fmod(d,0.25) is equal 0)
    print 'X';
}

This will of course not work since d will be [0, 0.05000000001, 0.100000000002, ...] causing fmod() to fail. Extreme example is when d=1.999999999998 and fmod(d,0.25) = 1.

How to tackle this? Here is an editable online example.

Solution

  • I'd solve this by simply not using floating point variables in that way:

    for (int i = 0; i <= 500; i += 5) {
  double d = i / 100.0;  // in case you need to use it.
  if ((i % 25) == 0)
    print 'X';
}

    They're usually problematic enough that it's worth a little extra effort in avoiding them.