Skip to content
public domain fast scalar math routines
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.


This is a set of branch-free, table-free implementations of the math functions in <math.h>.

All of the code is "public domain"1 - if you like, you can copy out the sin implementation if that's all you need.

These functions differ slightly from the C standard library by design - to stay fast, they don't do range checking by default, or follow C's precision requirements. Think of these routines more as "decent approximations", for when you know the particular input domain and you have leeway with the output. Video games are a good use case.

C compatibility

These functions stick to the C API, but the library doesn't implement all of math.h - notable exclusions are machine-related functions (e.g. fma, nextafter), precision-related functions (e.g. expm1), complex numbers, and tgamma/lgamma. The standard elementary functions that you would expect are included, as well as hyperbolic trig.


The library uses a select few functions, fully implemented from scratch, in order to derive the implementation of others via identities. For example, cos(x) is implemented as sin(x + π/2), and tan(x) is implemented as sin(x)/cos(x). This simplifies many of the routines, since the potential precision loss from this approach is negligible within the given bounds of the library.

Exceptions to branch-free rule

  • atan2 is highly discontinuous, so a truly branch-free implementation of this routine is likely impossible.

License & Credits

I want this library to be easy to use, so no attribution is necessary.

1Actual public domain dedications are of unclear legality worldwide, so an equivalent permissive license (no-clause BSD) is used, available at the bottom of the file.

Algorithm credit:

You can’t perform that action at this time.