[Gnu3dkit-dev] Commit: G3DDefs.h

[Brent Gulanowski]

I have improved the precision of the following definitions to take full 
advantage of the double type:

#define DEG2RAD
#define RAD2DEG
#define TWOPI
#define PIOVER2
#define E
#define SQRT2
#define SQRT3

Also, I replaced the fractional definition of DEG2RAD and RAD2DEG with 
precalculated values.

The replacement values are accurate to 16 decimal digits. IEEE764 
double precision floation point types contain 52 bits devoted to the 
mantissa, plus one hidden bit to the left of the decimal. The 
right-most binary bit has a value 2^-52 relative to the left-most 
(hidden) bit, or 2e-16 decimal.

Values were taken from a math page on the internet, the math.h header 
(usr/include/architecture/ppc/math.h in Mac OS X), and values output by 
the sqrt() function, double- and triple-checking where possible. 
DEG2RAD and RAD2DEG were calculated using a program and printing with 
printf().

--
Brent Gulanowski                                address@hidden

"Of course the Universe hates you. You're working to reduce chaos by 
expending a lot of energy to do your job. Thus, you're contributing to 
the eventual heat death of the universe, and it's just protecting 
itself from you."               -- John Batzel