Skip to content
it's /me edited this page Mar 5, 2021 · 4 revisions

A stumbling block in the conversion arises with the modulo function. The _fmod() provided by DCTL unfortunately behaves a little differently than the one in WebGL. So it is sometimes absolutely necessary to use the replacement:

#define mod_f(a,b) (a-b*_floor(a/b))

OpenGL ES defines mod(x,y) to compute x modulo y as x-y*floor(x/y). In cmath.han fmod(x,y)is defined being the floating point remainder of x/y which is calculated as x-n*y with n beging x/y with it's fractional part truncated. With floor truncating the fractional part this should be floor(x/y) and thereby this interpretation should match the OpenGL ES spec. And as fmod is overloaded to work with float as well as with double (the default), one would use float fmodf(float,float) to constrain it to float and only float. For Metal (don't know if it comes with Metal or is added by DCTL) _modf comes as the type generic macro #define _modf(X, INTVAL) modf((X), (INTVAL)) - and here it even works for the vector types. Furthermore you should know that in the C standard library there also is a T modf(T,T*) function defined with T being float, double, or long double that does decompose a value in its integral and fractional part, and has its modff variant for single precision values.

Clone this wiki locally