mirror of https://github.com/tongzx/nt5src
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
101 lines
3.1 KiB
101 lines
3.1 KiB
/* xmath.h internal header for Microsoft C */
|
|
#ifndef _XMATH
|
|
#define _XMATH
|
|
#include <wctype.h>
|
|
#include <errno.h>
|
|
#include <math.h>
|
|
#include <stddef.h>
|
|
#ifndef _YMATH
|
|
#include <ymath.h>
|
|
#endif
|
|
_STD_BEGIN
|
|
/* FLOAT PROPERTIES */
|
|
#define _DBIAS 0x3fe
|
|
#define _DOFF 4
|
|
#define _FBIAS 0x7e
|
|
#define _FOFF 7
|
|
#define _FRND 1
|
|
#define _D0 3 /* little-endian, small long doubles */
|
|
#define _D1 2
|
|
#define _D2 1
|
|
#define _D3 0
|
|
#define _DLONG 0
|
|
#define _LBIAS 0x3fe
|
|
#define _LOFF 4
|
|
/* IEEE 754 double properties */
|
|
#define _DFRAC ((1 << _DOFF) - 1)
|
|
#define _DMASK (0x7fff & ~_DFRAC)
|
|
#define _DMAX ((1 << (15 - _DOFF)) - 1)
|
|
#define _DSIGN 0x8000
|
|
#define DSIGN(x) (((unsigned short *)&(x))[_D0] & _DSIGN)
|
|
#define HUGE_EXP (int)(_DMAX * 900L / 1000)
|
|
#define HUGE_RAD 3.37e9 /* ~pi * 2^30 */
|
|
#define SAFE_EXP (_DMAX >> 1)
|
|
/* IEEE 754 float properties */
|
|
#define _FFRAC ((1 << _FOFF) - 1)
|
|
#define _FMASK (0x7fff & ~_FFRAC)
|
|
#define _FMAX ((1 << (15 - _FOFF)) - 1)
|
|
#define _FSIGN 0x8000
|
|
#define FSIGN(x) (((unsigned short *)&(x))[_F0] & _FSIGN)
|
|
#define FHUGE_EXP (int)(_FMAX * 900L / 1000)
|
|
#define FHUGE_RAD 3.37e9 /* ~pi * 2^30 */
|
|
#define FSAFE_EXP (_FMAX >> 1)
|
|
#define _F0 1 /* little-endian order */
|
|
#define _F1 0
|
|
/* IEEE 754 long double properties */
|
|
#define _LFRAC (-1)
|
|
#define _LMASK 0x7fff
|
|
#define _LMAX 0x7fff
|
|
#define _LSIGN 0x8000
|
|
#define LSIGN(x) (((unsigned short *)&(x))[_L0] & _LSIGN)
|
|
#define LHUGE_EXP (int)(_LMAX * 900L / 1000)
|
|
#define LHUGE_RAD 3.37e9 /* ~pi * 2^30 */
|
|
#define LSAFE_EXP (_LMAX >> 1)
|
|
#define _L0 4 /* little-endian order */
|
|
#define _L1 3
|
|
#define _L2 2
|
|
#define _L3 1
|
|
#define _L4 0
|
|
/* return values for testing functions */
|
|
#define FINITE _FINITE
|
|
#define INF _INFCODE
|
|
#define NAN _NANCODE
|
|
_C_LIB_DECL
|
|
/* double declarations */
|
|
_CRTIMP double _Atan(double, unsigned short);
|
|
_CRTIMP short _Dint(double *, short);
|
|
_CRTIMP short _Dnorm(unsigned short *);
|
|
_CRTIMP short _Dscale(double *, long);
|
|
_CRTIMP double _Dtento(double, long);
|
|
_CRTIMP short _Dunscale(short *, double *);
|
|
_CRTIMP double _Poly(double, const double *, int);
|
|
extern _CRTIMP const _Dconst _Eps, _Rteps;
|
|
extern _CRTIMP const double _Xbig;
|
|
/* float declarations */
|
|
_CRTIMP float _FAtan(float, unsigned short);
|
|
_CRTIMP short _FDint(float *, short);
|
|
_CRTIMP short _FDnorm(unsigned short *);
|
|
_CRTIMP short _FDscale(float *, long);
|
|
_CRTIMP float _FDtento(float, long);
|
|
_CRTIMP short _FDunscale(short *, float *);
|
|
_CRTIMP float _FPoly(float, const float *, int);
|
|
extern _CRTIMP const _Dconst _FEps, _FRteps;
|
|
extern _CRTIMP const float _FXbig;
|
|
/* long double functions */
|
|
_CRTIMP long double _LAtan(long double, unsigned short);
|
|
_CRTIMP short _LDint(long double *, short);
|
|
_CRTIMP short _LDnorm(unsigned short *);
|
|
_CRTIMP short _LDscale(long double *, long);
|
|
_CRTIMP long double _LDtento(long double, long);
|
|
_CRTIMP short _LDunscale(short *, long double *);
|
|
_CRTIMP long double _LPoly(long double, const long double *, int);
|
|
extern _CRTIMP const _Dconst _LEps, _LRteps;
|
|
extern _CRTIMP const long double _LXbig;
|
|
_END_C_LIB_DECL
|
|
_STD_END
|
|
#endif /* _XMATH */
|
|
|
|
/*
|
|
* Copyright (c) 1994 by P.J. Plauger. ALL RIGHTS RESERVED.
|
|
* Consult your license regarding permissions and restrictions.
|
|
*/
|