include: Add float functions to msvcrt.

This commit is contained in:
Maarten Lankhorst 2010-05-01 16:32:45 +02:00 committed by Alexandre Julliard
parent dc04cee787
commit f4fcd81167
2 changed files with 59 additions and 0 deletions

View File

@ -135,6 +135,10 @@ int __cdecl _finite(double);
int __cdecl _isnan(double);
int __cdecl _fpclass(double);
#ifdef __x86_64__
float __cdecl _scalbf(float, __msvcrt_long);
#endif
#ifdef __cplusplus
}
#endif

View File

@ -76,6 +76,61 @@ double __cdecl y0(double);
double __cdecl y1(double);
double __cdecl yn(int, double);
#ifdef __x86_64__
float __cdecl sinf(float);
float __cdecl cosf(float);
float __cdecl tanf(float);
float __cdecl sinhf(float);
float __cdecl coshf(float);
float __cdecl tanhf(float);
float __cdecl asinf(float);
float __cdecl acosf(float);
float __cdecl atanf(float);
float __cdecl atan2f(float, float);
float __cdecl expf(float);
float __cdecl logf(float);
float __cdecl log10f(float);
float __cdecl powf(float, float);
float __cdecl sqrtf(float);
float __cdecl ceilf(float);
float __cdecl floorf(float);
float __cdecl fabsf(float);
float __cdecl ldexpf(float, int);
float __cdecl frexpf(float, int*);
float __cdecl modff(float, float*);
float __cdecl fmodf(float, float);
#else
#define sinf(x) ((float)sin((double)(x)))
#define cosf(x) ((float)cos((double)(x)))
#define tanf(x) ((float)tan((double)(x)))
#define sinhf(x) ((float)sinh((double)(x)))
#define coshf(x) ((float)cosh((double)(x)))
#define tanhf(x) ((float)tanh((double)(x)))
#define asinf(x) ((float)asin((double)(x)))
#define acosf(x) ((float)acos((double)(x)))
#define atanf(x) ((float)atan((double)(x)))
#define atan2f(x,y) ((float)atan2((double)(x), (double)(y)))
#define expf(x) ((float)exp((double)(x)))
#define logf(x) ((float)log((double)(x)))
#define log10f(x) ((float)log10((double)(x)))
#define powf(x,y) ((float)pow((double)(x), (double)(y)))
#define sqrtf(x) ((float)sqrt((double)(x)))
#define ceilf(x) ((float)ceil((double)(x)))
#define floorf(x) ((float)floor((double)(x)))
#define frexpf(x) ((float)frexp((double)(x)))
#define modff(x,y) ((float)modf((double)(x), (double*)(y)))
#define fmodf(x,y) ((float)fmod((double)(x), (double)(y)))
#endif
#define fabsf(x) ((float)fabs((double)(x)))
#define ldexpf(x) ((float)ldexp((double)(x)))
float __cdecl hypotf(float, float);
int __cdecl _matherr(struct _exception*);
double __cdecl _cabs(struct _complex);