Fastest implementation of the frac function in C#

Question

I would like to implement a frac function in C# (just like the one in hsl here http://msdn.microsoft.com/en-us/library/bb509603%28VS.85%29.aspx) but since it is for a very processor intensive application i would like the best version possible. I was using something like

public float Frac(float value)
{
   return value - (float)Math.Truncate(value);
}

but I'm having precision problems, for example for 2.6f it's returning in the unit test

Expected: 0.600000024f But was: 0.599999905f

I know that I can convert to decimal the value and then at the end convert to float to obtain the correct result something like this:

public float Frac(float value)
{
   return (float)((decimal)value - Decimal.Truncate((decimal)value));
}

But I wonder if there is a better way without resorting to decimals...

Answer 1

float (an alias for System.Single) is only precise to 7 digits. You shouldn't really compare floating point values using equality anyway. You should instead check that they are within a tolerable range - for example that the difference between the expected and actual values is not more than 0.000001.

Assert.AreEqual(expected, actual, delta);

If you use double instead, you will be precise to 15-16 digits.

Answer 2

You are going to need to use decimals as in:

public decimal Frac(decimal value) { return value - Math.Truncate(value); }

So, when you pass 2.6m you will always get .6