_src_math_funcs.hh File Reference

Math functions for computer graphics. More...

#include "math/funcs_debug.hh"

Go to the source code of this file.

Namespaces
namespace	math
	math functions
namespace	math::defs
Defines
#define	IfNan(N)   (UX(!std::isfinite(N)))
Functions
template<typename FP >
bool	math::FP_EQ (FP a, FP b)
template<typename FP >
bool	math::FP_LE (FP a, FP b)
template<typename FP >
bool	math::FP_GE (FP a, FP b)
template<typename T1 , typename T2 >
bool	math::IfSameSign (const T1 a, const T2 b)
template<typename FP >
FP	math::SquareRoot (FP f)
template<typename T >
T	math::Square (T x)
template<typename T >
T	math::Average (T x, T y)
template<typename T >
T	math::Range (T val, T lo, T hi)
template<typename T >
T	math::IfInRange (T a0, T a1, T b0, T b1)
INLINE fp	math::FractionInRange (fp n, fp low, fp high)
INLINE fp	math::Remap (fp value, fp rangeOld, fp rangeNew)
INLINE fp	math::Truncate (fp val, fp interval)
template<typename T >
T	math::ModuloAbs (const T n, const T limit)
Variables
const fpx	math::defs::FP_DELTA_TOLERANCE = 0.001
	used for comparing fp values

---

Detailed Description

Math functions for computer graphics.

LastChangedDate:

2010-10-23 18:22:39 -0400 (Sat, 23 Oct 2010)

Author:

Jim E. Brooks http://www.palomino3d.org

Remarks:

- A few functions are named specially to avoid ambiguity with standard functions.

• To speed or improve FP execution:
  • use multiplication instead of division
  • power of 2 operations involve no loss of accuracy
  • math::PI is preferred over M_PI etc from math.h. math.h has confusing defines such as M_2_PI which means 2/PI (not 2*PI).
• NaN are caused by division and some standard C math functions.
• All modern FPUs are based on IEEE 754. A single-precision 32-bit FP is limited to 24 bits (8 decimal digits) of precision.
• C++ <cmath> is better than C <math.h> because C++ functions are overloaded. REMINDERS: - Consider using MATH_EXCEPTION() instead of ASSERT().

---

Define Documentation

#define IfNan

(

N

)

(UX(!std::isfinite(N)))

IfNan() returns true if a Not-A-Number or Infinity. C99 isfinite() is true if neither NaN nor infinity. Technically, divide-by-zero results in Infinity, not NaN, but IfNan() is used to check divide errors. A NaN never equals any number -- not even itself. !(n==n) is unreliable because smart compilers will optimize as always false (gcc will).