maths - z88dk/z88dk GitHub Wiki
| Header | {z88dk}/include/math.h, {z88dk}/include/float.h |
| ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| Source | {z88dk}/libsrc/math |
| Include | #include <math.h> |
| Linking | -lm (or target specific alternative, i.e. -lmzx) |
| Compile | |
| Comments | none |
The maths library
========= Target specific alternatives ========= On some target other it is possible to optionally link-in smaller libraries interfacing to the native ROM code. Such alternatives are normally slighlty weaker either in speed or precision but are still a good option when the program size becomes an issue.
The available options at the moment are:
| Target platform ^ Small maths LIB ^ Even smaller option |
|---|
| Sinclair ZX 81 |
| Lambda 8300 |
| Sinclair ZX Spectrum |
| Cambridge Z88 |
| Amstrad CPC |
| Amstrad CPC 464 |
| Amstrad CPC 664 |
| Amstrad CPC 6128 |
========= Floating Point Precision ========= The precision of an FP number is determined by the so called "EPSILON" value. It is the smallest value which can be added to '1' still making a difference.
The given example tests the pre-defined FLT_EPSILON constant and tries to determine EPSILON experimentally to compare the empiric results to the compiler's presets.
Compiler command line examples:
zcc +`<target>` -lndos -create-app -lm epsilon.c
zcc +zx -lndos -create-app -lmzx -DFASTMATH epsilon.c
#include `<stdio.h>`
#include `<math.h>`
#include `<float.h>`
#ifdef DOUBLE
double a,b;
#else
float a,b;
#endif
int exponent;
double getexp(double value) {
exponent=0;
while ((int)value==0.0) {
exponent--;
value *= 10.0;
}
return(value);
}
int main() {
a=1;
while ((1.0+a)!=1.0)
{
b=a;
a=a/2.0;
}
putchar(12);
getexp(b);
printf ("Computed value for Epsilon: %fe%d\n", getexp(b),exponent);
if ((1.0+b)==1.0) {
printf ("..not confirmed.\n");
}
getexp(FLT_EPSILON);
printf ("Predefined value for Epsilon: %fe%d\n", getexp(FLT_EPSILON),exponent);
a=1.0;
b=a;
while ((a+FLT_EPSILON)>a) {
b=a;
a*=10.0;
}
a=b;
if (a>1.0) {
while (((a+FLT_EPSILON)>a)&&(b!=a)) {
b=a;
/* if the loop stales, try adding a minimum value, FLT_EPSILON is fine */
/* a+=1.0+FLT_EPSILON; */
a+=1.0;
}
a=b;
}
if (a<10.0) {
printf ("FLT_EPSILON is very close to the real machine limit.\n");
} else {
if (a>2000.0)
printf ("FLT_EPSILON is far from the real machine limit.\n");
}
printf ("\nEPSILON is relevant for values <= %f\n", a);
if ((1.0+FLT_EPSILON)==1.0) {
printf ("\nError: this value for FLT_EPSILON is not correct !\n\n");
}
}