newlib-cygwin/newlib/libm/math/wf_sinh.c

79 lines
1.7 KiB
C

/* wf_sinh.c -- float version of w_sinh.c.
* Conversion to float by Ian Lance Taylor, Cygnus Support, ian@cygnus.com.
*/
/*
* ====================================================
* Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
*
* Developed at SunPro, a Sun Microsystems, Inc. business.
* Permission to use, copy, modify, and distribute this
* software is freely granted, provided that this notice
* is preserved.
* ====================================================
*/
/*
* wrapper sinhf(x)
*/
#include "fdlibm.h"
#include <errno.h>
#ifdef __STDC__
float sinhf(float x) /* wrapper sinhf */
#else
float sinhf(x) /* wrapper sinhf */
float x;
#endif
{
#ifdef _IEEE_LIBM
return __ieee754_sinhf(x);
#else
float z;
struct exception exc;
z = __ieee754_sinhf(x);
if(_LIB_VERSION == _IEEE_) return z;
if(!finitef(z)&&finitef(x)) {
/* sinhf(finite) overflow */
#ifndef HUGE_VAL
#define HUGE_VAL inf
double inf = 0.0;
SET_HIGH_WORD(inf,0x7ff00000); /* set inf to infinite */
#endif
exc.type = OVERFLOW;
exc.name = "sinhf";
exc.err = 0;
exc.arg1 = exc.arg2 = (double)x;
if (_LIB_VERSION == _SVID_)
exc.retval = ( (x>0.0) ? HUGE : -HUGE);
else
exc.retval = ( (x>0.0) ? HUGE_VAL : -HUGE_VAL);
if (_LIB_VERSION == _POSIX_)
errno = ERANGE;
else if (!matherr(&exc)) {
errno = ERANGE;
}
if (exc.err != 0)
errno = exc.err;
return (float)exc.retval;
} else
return z;
#endif
}
#ifdef _DOUBLE_IS_32BITS
#ifdef __STDC__
double sinh(double x)
#else
double sinh(x)
double x;
#endif
{
return (double) sinhf((float) x);
}
#endif /* defined(_DOUBLE_IS_32BITS) */