Fixed Point Types

A library created to provide fixed point numeric types for use cases where floating point numbers just won't cut it.

Jump straight to usage

When would you want to use fixed point numbers over floating point numbers?

• Are fixed point numbers faster than floating point numbers?

  Not really, dedicated FPUs in most modern processors actually make floating point operations faster than fixed point operations.

• Are fixed point numbers more precise than floating point numbers?

  No, sometimes fixed point numbers are less precise than floating point numbers. But also this can vary based on the scale the developer chooses.

• Do fixed point numbers have a larger range than floating point numbers?

  Nope, not when using the same amount of bits when storing the fixed point number anyway. This is actually a huge advantage of floating point numbers.

• So why would you want to use fixed point numbers?

  Fixed point numbers are deterministic. This is really useful for things like simulations where results need to be perfectly replicable. I originally created this library to create a game using a deterministic lockstep networking model. But There are many different applications for fixed point numbers.

Usage

Install the nuget package

OR Include the Fixed32.cs(Or the entire FixedTypes.csproj) file in your solution.

To initialize a new fixed point number:

var someDeterministicNumber = new Fixed32(16); // 0.0
var someDeterministicNumber = new Fixed32(16, 5); // 5.0

The 16 passed to constructor is the scale of the number. It basically specifies how many bits should be used to store the characteristic(bits preceding the point) and how many bits should be used to store the mantissa(bits succeeding the point).

:information_source: Tips

• A smaller scale will decrease precision and increase range.
• A larger scale will increase precision and decrease range.
• Scale should be between(and including) 1 and 31
• Operations +, -, *, / etc. should only be done on fixed point numbers of the same scale.

There is no way to initialize a fixed point number as a fraction(only as a whole number). This is because all numeric literals in C# are floating point numbers, with the exception of integers. If you wanted to initialize a fraction I would suggest dividing the numerator with denominator of the fraction. For example:

var twoAndAHalf = new Fixed32(16, 5) / new Fixed32(16, 2); // 5/2 = 2.5

Contributing

This project is covered by many tests, if you find a bug please create an issue. If you know how to solve the issue please create a pull request with the solution and a test for the scenario described in the original issue.

This library is far from complete and still requires many useful features. So feel free to create a pull request.:v: