Retro68/gcc/newlib/libm/mathfp/wf_jn.c
2022-10-27 20:55:19 +02:00

98 lines
1.8 KiB
C

/* wf_jn.c -- float version of w_jn.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.
* ====================================================
*/
#include "fdlibm.h"
#include <errno.h>
#ifdef __STDC__
float jnf(int n, float x) /* wrapper jnf */
#else
float jnf(n,x) /* wrapper jnf */
float x; int n;
#endif
{
#ifdef _IEEE_LIBM
return jnf(n,x);
#else
float z;
z = jnf(n,x);
if(_LIB_VERSION == _IEEE_ || isnanf(x) ) return z;
if(fabsf(x)>(float)X_TLOSS) {
/* jnf(|x|>X_TLOSS) */
errno = ERANGE;
return 0.0f;
} else
return z;
#endif
}
#ifdef __STDC__
float ynf(int n, float x) /* wrapper ynf */
#else
float ynf(n,x) /* wrapper ynf */
float x; int n;
#endif
{
#ifdef _IEEE_LIBM
return ynf(n,x);
#else
float z;
z = ynf(n,x);
if(_LIB_VERSION == _IEEE_ || isnanf(x) ) return z;
if(x <= 0.0f){
/* ynf(n,0) = -inf or ynf(x<0) = NaN */
#ifndef HUGE_VAL
#define HUGE_VAL inf
double inf = 0.0;
SET_HIGH_WORD(inf,0x7ff00000); /* set inf to infinite */
#endif
errno = EDOM;
return (float)-HUGE_VAL;
}
if(x>(float)X_TLOSS) {
/* ynf(x>X_TLOSS) */
errno = ERANGE;
return 0.0f;
} else
return z;
#endif
}
#ifdef _DOUBLE_IS_32BITS
#ifdef __STDC__
double jn(int n, double x)
#else
double jn(n,x)
double x; int n;
#endif
{
return (double) jnf(n, (float) x);
}
#ifdef __STDC__
double yn(int n, double x)
#else
double yn(n,x)
double x; int n;
#endif
{
return (double) ynf(n, (float) x);
}
#endif /* defined(_DOUBLE_IS_32BITS) */