sine and cosine function implementation in C using Taylor series

math_operation.c

#include <stdlib.h>
#include <stdio.h>
#include <math.h>

int compare_float(double f1, double f2)
{
 double precision = 0.00000000000000000001;
 if ((f1 - precision) < f2)
  {
 return -1;
  }
 else if ((f1 + precision) > f2)
 {
  return 1;
 }
 else
  {
 return 0;
  }
}

double cos(double x){
 if( x < 0.0f ) 
  x = -x;

  if (0 <= compare_float(x,M_PI_M_2)) 
 {
 do {
  x -= M_PI_M_2;
  }while(0 <= compare_float(x,M_PI_M_2));

  }

  if ((0 <= compare_float(x, M_PI)) && (-1 == compare_float(x, M_PI_M_2)))
  {
   x -= M_PI;
   return ((-1)*(1.0f - (x*x/2.0f)*( 1.0f - (x*x/12.0f) * ( 1.0f - (x*x/30.0f) * (1.0f - (x*x/56.0f )*(1.0f - (x*x/90.0f)*(1.0f - (x*x/132.0f)*(1.0f - (x*x/182.0f)))))))));
  } 
 return 1.0f - (x*x/2.0f)*( 1.0f - (x*x/12.0f) * ( 1.0f - (x*x/30.0f) * (1.0f - (x*x/56.0f )*(1.0f - (x*x/90.0f)*(1.0f - (x*x/132.0f)*(1.0f - (x*x/182.0f)))))));
}

double sin(double x){return cos(x-M_PI_2);}

Thanks very nice. Was this taken from a specific libraries headers?

@Lewiscowles1986 No i wrote it myself based on Taylor's series.

@giangnguyen2412 nice to see you finally. Apologies, the question was assumed as directed at @robertdetian unless these are both your github usernames, or you have an answer with regards the defines in https://gist.github.com/giangnguyen2412/bcab883b5a53b437b980d7be9745beaf?permalink_comment_id=4286644#gistcomment-4286644

@Lewiscowles1986 Sorry for my (super) late reply. No we are not :D But I think that definition works unless you want the results more accurate (i.e. extending the numbers after the floating point).

Thanks @giangnguyen2412 for this example.

I managed to implement the float port cosine, and rewrote above code as following, which can make Taylor's series more scalable, although 7 times works well for float.

float c_cos(float x)
{
const float pi = 3.1415927;

x = std::copysign(x, 1.0f);

auto pi_count = (uint32_t)(x / pi);
x -= (float)pi_count * pi;

float taylor_series = 1.0;
for (int i = 14; i > 0; i-=2)
{
    taylor_series = 1.0f - (x*x)/(float)(i*(i-1)) * taylor_series;
}

uint32_t n_one_f = 0xBF800000;
return (pi_count & 1) ? taylor_series * *(float*)&n_one_f : taylor_series;

}

Ref: https://en.wikipedia.org/wiki/Sine_and_cosine