diff options
Diffstat (limited to 'include/iprt/nocrt/math.h')
-rw-r--r-- | include/iprt/nocrt/math.h | 859 |
1 files changed, 859 insertions, 0 deletions
diff --git a/include/iprt/nocrt/math.h b/include/iprt/nocrt/math.h new file mode 100644 index 00000000..4ae2c48c --- /dev/null +++ b/include/iprt/nocrt/math.h @@ -0,0 +1,859 @@ +/** @file + * IPRT / No-CRT - math.h. + */ + +/* + * Copyright (C) 2006-2022 Oracle and/or its affiliates. + * + * This file is part of VirtualBox base platform packages, as + * available from https://www.virtualbox.org. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation, in version 3 of the + * License. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, see <https://www.gnu.org/licenses>. + * + * The contents of this file may alternatively be used under the terms + * of the Common Development and Distribution License Version 1.0 + * (CDDL), a copy of it is provided in the "COPYING.CDDL" file included + * in the VirtualBox distribution, in which case the provisions of the + * CDDL are applicable instead of those of the GPL. + * + * You may elect to license modified versions of this file under the + * terms and conditions of either the GPL or the CDDL or both. + * + * SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0 + * -------------------------------------------------------------------- + * + * This code is based on: + * + * from: @(#)fdlibm.h 5.1 93/09/24 + * $FreeBSD: src/lib/msun/src/math.h,v 1.61 2005/04/16 21:12:47 das Exp $ + * FreeBSD HEAD 2005-06-xx + * + * ==================================================== + * 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. + * ==================================================== + */ + +#ifndef IPRT_INCLUDED_nocrt_math_h +#define IPRT_INCLUDED_nocrt_math_h +#ifndef RT_WITHOUT_PRAGMA_ONCE +# pragma once +#endif + +#include <iprt/types.h> +/*#include <machine/_limits.h>*/ + +/* from sys/cdefs.h */ +#if defined(__GNUC__) && !defined(__INTEL_COMPILER) +#define __GNUC_PREREQ__(ma, mi) \ + (__GNUC__ > (ma) || __GNUC__ == (ma) && __GNUC_MINOR__ >= (mi)) +#else +#define __GNUC_PREREQ__(ma, mi) 0 +#endif +#undef __pure2 /* darwin: avoid conflict with system headers when doing syntax checking of the headers */ +#define __pure2 + + +/* + * ANSI/POSIX + */ +extern const union __infinity_un { + RTFLOAT64U __uu; + double __ud; +} RT_NOCRT(__infinity); + +extern const union __nanf_un { + RTFLOAT32U __uu; + float __uf; +} RT_NOCRT(__nanf); + +#if __GNUC_PREREQ__(3, 3) || (defined(__INTEL_COMPILER) && __INTEL_COMPILER >= 800) +#define __MATH_BUILTIN_CONSTANTS +#endif + +#if __GNUC_PREREQ__(3, 0) && !defined(__INTEL_COMPILER) +#define __MATH_BUILTIN_RELOPS +#endif + +#ifndef IPRT_NOCRT_WITHOUT_CONFLICTING_CONSTANTS + +# if defined(__MATH_BUILTIN_CONSTANTS) \ + || (RT_MSC_PREREQ(RT_MSC_VER_VC140) && defined(__cplusplus)) /** @todo when was this added exactly? 2015, 2017 & 2019 has it for C++. */ +# define HUGE_VAL __builtin_huge_val() +# else +# define HUGE_VAL (RT_NOCRT(__infinity).__ud) +# endif + +/* + * XOPEN/SVID + */ +# if 1/* __BSD_VISIBLE || __XSI_VISIBLE*/ +# define M_E 2.7182818284590452354 /* e */ +# define M_LOG2E 1.4426950408889634074 /* log 2e */ +# define M_LOG10E 0.43429448190325182765 /* log 10e */ +# define M_LN2 0.69314718055994530942 /* log e2 */ +# define M_LN10 2.30258509299404568402 /* log e10 */ +# define M_PI 3.14159265358979323846 /* pi */ +# define M_PI_2 1.57079632679489661923 /* pi/2 */ +# define M_PI_4 0.78539816339744830962 /* pi/4 */ +# define M_1_PI 0.31830988618379067154 /* 1/pi */ +# define M_2_PI 0.63661977236758134308 /* 2/pi */ +# define M_2_SQRTPI 1.12837916709551257390 /* 2/sqrt(pi) */ +# define M_SQRT2 1.41421356237309504880 /* sqrt(2) */ +# define M_SQRT1_2 0.70710678118654752440 /* 1/sqrt(2) */ + +# define MAXFLOAT ((float)3.40282346638528860e+38) +extern int RT_NOCRT(signgam); +# endif /* __BSD_VISIBLE || __XSI_VISIBLE */ + +# if 1/* __BSD_VISIBLE*/ +# if 0 +/* Old value from 4.4BSD-Lite math.h; this is probably better. */ +# define HUGE HUGE_VAL +# else +# define HUGE MAXFLOAT +# endif +# endif /* __BSD_VISIBLE */ + +#endif /* !IPRT_NOCRT_WITHOUT_MATH_CONSTANTS */ + +/* + * Most of these functions depend on the rounding mode and have the side + * effect of raising floating-point exceptions, so they are not declared + * as __pure2. In C99, FENV_ACCESS affects the purity of these functions. + */ +RT_C_DECLS_BEGIN +/* + * ANSI/POSIX + */ +int RT_NOCRT(__fpclassifyd)(double) __pure2; +int RT_NOCRT(__fpclassifyf)(float) __pure2; +int RT_NOCRT(__fpclassifyl)(long double) __pure2; +int RT_NOCRT(__isfinitef)(float) __pure2; +int RT_NOCRT(__isfinite)(double) __pure2; +int RT_NOCRT(__isfinitel)(long double) __pure2; +int RT_NOCRT(__isinff)(float) __pure2; +int RT_NOCRT(__isinfl)(long double) __pure2; +int RT_NOCRT(__isnanl)(long double) __pure2; +int RT_NOCRT(__isnormalf)(float) __pure2; +int RT_NOCRT(__isnormal)(double) __pure2; +int RT_NOCRT(__isnormall)(long double) __pure2; +int RT_NOCRT(__signbit)(double) __pure2; +int RT_NOCRT(__signbitf)(float) __pure2; +int RT_NOCRT(__signbitl)(long double) __pure2; + +double RT_NOCRT(acos)(double); +double RT_NOCRT(asin)(double); +double RT_NOCRT(atan)(double); +double RT_NOCRT(atan2)(double, double); +double RT_NOCRT(cos)(double); +double RT_NOCRT(sin)(double); +double RT_NOCRT(tan)(double); + +double RT_NOCRT(cosh)(double); +double RT_NOCRT(sinh)(double); +double RT_NOCRT(tanh)(double); + +double RT_NOCRT(exp)(double); +double RT_NOCRT(frexp)(double, int *); /* fundamentally !__pure2 */ +double RT_NOCRT(ldexp)(double, int); +double RT_NOCRT(log)(double); +double RT_NOCRT(log10)(double); +double RT_NOCRT(modf)(double, double *); /* fundamentally !__pure2 */ + +double RT_NOCRT(pow)(double, double); +double RT_NOCRT(sqrt)(double); + +double RT_NOCRT(ceil)(double); +double RT_NOCRT(fabs)(double) __pure2; +double RT_NOCRT(floor)(double); +double RT_NOCRT(fmod)(double, double); + +/* + * These functions are not in C90. + */ +#if 1 /*__BSD_VISIBLE || __ISO_C_VISIBLE >= 1999 || __XSI_VISIBLE*/ +double RT_NOCRT(acosh)(double); +double RT_NOCRT(asinh)(double); +double RT_NOCRT(atanh)(double); +double RT_NOCRT(cbrt)(double); +double RT_NOCRT(erf)(double); +double RT_NOCRT(erfc)(double); +double RT_NOCRT(exp2)(double); +double RT_NOCRT(expm1)(double); +double RT_NOCRT(fma)(double, double, double); +double RT_NOCRT(hypot)(double, double); +int RT_NOCRT(ilogb)(double) __pure2; +int RT_NOCRT(isinf)(double) __pure2; +int RT_NOCRT(isnan)(double) __pure2; +double RT_NOCRT(lgamma)(double); +long long RT_NOCRT(llrint)(double); +long long RT_NOCRT(llround)(double); +double RT_NOCRT(log1p)(double); +double RT_NOCRT(logb)(double); +long RT_NOCRT(lrint)(double); +long RT_NOCRT(lround)(double); +double RT_NOCRT(nextafter)(double, double); +double RT_NOCRT(remainder)(double, double); +double RT_NOCRT(remquo)(double, double, int *); +double RT_NOCRT(rint)(double); +#endif /* __BSD_VISIBLE || __ISO_C_VISIBLE >= 1999 || __XSI_VISIBLE */ + +#if 1/* __BSD_VISIBLE || __XSI_VISIBLE*/ +double RT_NOCRT(j0)(double); +double RT_NOCRT(j1)(double); +double RT_NOCRT(jn)(int, double); +double RT_NOCRT(scalb)(double, double); +double RT_NOCRT(y0)(double); +double RT_NOCRT(y1)(double); +double RT_NOCRT(yn)(int, double); + +#if 1/* __XSI_VISIBLE <= 500 || __BSD_VISIBLE*/ +double RT_NOCRT(gamma)(double); +#endif +#endif /* __BSD_VISIBLE || __XSI_VISIBLE */ + +#if 1/* __BSD_VISIBLE || __ISO_C_VISIBLE >= 1999*/ +double RT_NOCRT(copysign)(double, double) __pure2; +double RT_NOCRT(fdim)(double, double); +double RT_NOCRT(fmax)(double, double) __pure2; +double RT_NOCRT(fmin)(double, double) __pure2; +double RT_NOCRT(nearbyint)(double); +double RT_NOCRT(round)(double); +double RT_NOCRT(scalbln)(double, long); +double RT_NOCRT(scalbn)(double, int); +double RT_NOCRT(tgamma)(double); +double RT_NOCRT(trunc)(double); +#endif + +/* + * BSD math library entry points + */ +#if 1/* __BSD_VISIBLE*/ +double RT_NOCRT(drem)(double, double); +int RT_NOCRT(finite)(double) __pure2; +int RT_NOCRT(isnanf)(float) __pure2; + +/* + * Reentrant version of gamma & lgamma; passes signgam back by reference + * as the second argument; user must allocate space for signgam. + */ +double RT_NOCRT(gamma_r)(double, int *); +double RT_NOCRT(lgamma_r)(double, int *); + +/* + * IEEE Test Vector + */ +double RT_NOCRT(significand)(double); +#endif /* __BSD_VISIBLE */ + +/* float versions of ANSI/POSIX functions */ +#if 1/* __ISO_C_VISIBLE >= 1999*/ +float RT_NOCRT(acosf)(float); +float RT_NOCRT(asinf)(float); +float RT_NOCRT(atanf)(float); +float RT_NOCRT(atan2f)(float, float); +float RT_NOCRT(cosf)(float); +float RT_NOCRT(sinf)(float); +float RT_NOCRT(tanf)(float); + +float RT_NOCRT(coshf)(float); +float RT_NOCRT(sinhf)(float); +float RT_NOCRT(tanhf)(float); + +float RT_NOCRT(exp2f)(float); +float RT_NOCRT(expf)(float); +float RT_NOCRT(expm1f)(float); +float RT_NOCRT(frexpf)(float, int *); /* fundamentally !__pure2 */ +int RT_NOCRT(ilogbf)(float) __pure2; +float RT_NOCRT(ldexpf)(float, int); +float RT_NOCRT(log10f)(float); +float RT_NOCRT(log1pf)(float); +float RT_NOCRT(logf)(float); +float RT_NOCRT(modff)(float, float *); /* fundamentally !__pure2 */ + +float RT_NOCRT(powf)(float, float); +float RT_NOCRT(sqrtf)(float); + +float RT_NOCRT(ceilf)(float); +float RT_NOCRT(fabsf)(float) __pure2; +float RT_NOCRT(floorf)(float); +float RT_NOCRT(fmodf)(float, float); +float RT_NOCRT(roundf)(float); + +float RT_NOCRT(erff)(float); +float RT_NOCRT(erfcf)(float); +float RT_NOCRT(hypotf)(float, float); +float RT_NOCRT(lgammaf)(float); + +float RT_NOCRT(acoshf)(float); +float RT_NOCRT(asinhf)(float); +float RT_NOCRT(atanhf)(float); +float RT_NOCRT(cbrtf)(float); +float RT_NOCRT(logbf)(float); +float RT_NOCRT(copysignf)(float, float) __pure2; +long long RT_NOCRT(llrintf)(float); +long long RT_NOCRT(llroundf)(float); +long RT_NOCRT(lrintf)(float); +long RT_NOCRT(lroundf)(float); +float RT_NOCRT(nearbyintf)(float); +float RT_NOCRT(nextafterf)(float, float); +float RT_NOCRT(remainderf)(float, float); +float RT_NOCRT(remquof)(float, float, int *); +float RT_NOCRT(rintf)(float); +float RT_NOCRT(scalblnf)(float, long); +float RT_NOCRT(scalbnf)(float, int); +float RT_NOCRT(truncf)(float); + +float RT_NOCRT(fdimf)(float, float); +float RT_NOCRT(fmaf)(float, float, float); +float RT_NOCRT(fmaxf)(float, float) __pure2; +float RT_NOCRT(fminf)(float, float) __pure2; +#endif + +/* + * float versions of BSD math library entry points + */ +#if 1/* __BSD_VISIBLE*/ +float RT_NOCRT(dremf)(float, float); +int RT_NOCRT(finitef)(float) __pure2; +float RT_NOCRT(gammaf)(float); +float RT_NOCRT(j0f)(float); +float RT_NOCRT(j1f)(float); +float RT_NOCRT(jnf)(int, float); +float RT_NOCRT(scalbf)(float, float); +float RT_NOCRT(y0f)(float); +float RT_NOCRT(y1f)(float); +float RT_NOCRT(ynf)(int, float); + +/* + * Float versions of reentrant version of gamma & lgamma; passes + * signgam back by reference as the second argument; user must + * allocate space for signgam. + */ +float RT_NOCRT(gammaf_r)(float, int *); +float RT_NOCRT(lgammaf_r)(float, int *); + +/* + * float version of IEEE Test Vector + */ +float RT_NOCRT(significandf)(float); +#endif /* __BSD_VISIBLE */ + +/* + * long double versions of ISO/POSIX math functions + */ +#if 1/* __ISO_C_VISIBLE >= 1999*/ +#if 1 /* bird: we've got these */ +long double RT_NOCRT(acoshl)(long double); +long double RT_NOCRT(acosl)(long double); +long double RT_NOCRT(asinhl)(long double); +long double RT_NOCRT(asinl)(long double); +long double RT_NOCRT(atan2l)(long double, long double); +long double RT_NOCRT(atanhl)(long double); +long double RT_NOCRT(atanl)(long double); +long double RT_NOCRT(cbrtl)(long double); +#endif +long double RT_NOCRT(ceill)(long double); +long double RT_NOCRT(copysignl)(long double, long double) __pure2; +#if 1 /* bird */ +long double RT_NOCRT(coshl)(long double); +long double RT_NOCRT(cosl)(long double); +long double RT_NOCRT(erfcl)(long double); +long double RT_NOCRT(erfl)(long double); +long double RT_NOCRT(exp2l)(long double); +long double RT_NOCRT(expl)(long double); +long double RT_NOCRT(expm1l)(long double); +#endif +long double RT_NOCRT(fabsl)(long double) __pure2; +long double RT_NOCRT(fdiml)(long double, long double); +long double RT_NOCRT(floorl)(long double); +long double RT_NOCRT(fmal)(long double, long double, long double); +long double RT_NOCRT(fmaxl)(long double, long double) __pure2; +long double RT_NOCRT(fminl)(long double, long double) __pure2; +#if 1 /* bird */ +long double RT_NOCRT(fmodl)(long double, long double); +#endif +long double RT_NOCRT(frexpl)(long double value, int *); /* fundamentally !__pure2 */ +#if 1 /* bird */ +long double RT_NOCRT(hypotl)(long double, long double); +#endif +int RT_NOCRT(ilogbl)(long double) __pure2; +long double RT_NOCRT(ldexpl)(long double, int); +#if 1 /* bird */ +long double RT_NOCRT(lgammal)(long double); +long long RT_NOCRT(llrintl)(long double); +#endif +long long RT_NOCRT(llroundl)(long double); +#if 1 /* bird */ +long double RT_NOCRT(log10l)(long double); +long double RT_NOCRT(log1pl)(long double); +long double RT_NOCRT(log2l)(long double); +long double RT_NOCRT(logbl)(long double); +long double RT_NOCRT(logl)(long double); +long RT_NOCRT(lrintl)(long double); +#endif +long RT_NOCRT(lroundl)(long double); +#if 1 /* bird */ +long double RT_NOCRT(modfl)(long double, long double *); /* fundamentally !__pure2 */ +long double RT_NOCRT(nanl)(const char *) __pure2; +long double RT_NOCRT(nearbyintl)(long double); +#endif +long double RT_NOCRT(nextafterl)(long double, long double); +double RT_NOCRT(nexttoward)(double, long double); +float RT_NOCRT(nexttowardf)(float, long double); +long double RT_NOCRT(nexttowardl)(long double, long double); +#if 1 /* bird */ +long double RT_NOCRT(powl)(long double, long double); +long double RT_NOCRT(remainderl)(long double, long double); +long double RT_NOCRT(remquol)(long double, long double, int *); +long double RT_NOCRT(rintl)(long double); +#endif +long double RT_NOCRT(roundl)(long double); +long double RT_NOCRT(scalblnl)(long double, long); +long double RT_NOCRT(scalbnl)(long double, int); +#if 1 /* bird: we 've got most of these. */ +long double RT_NOCRT(sinhl)(long double); +long double RT_NOCRT(sinl)(long double); +long double RT_NOCRT(sqrtl)(long double); +long double RT_NOCRT(tanhl)(long double); +long double RT_NOCRT(tanl)(long double); +long double RT_NOCRT(tgammal)(long double); +#endif +long double RT_NOCRT(truncl)(long double); + +/* bird: these were missing, gcc apparently inlines them. */ +double RT_NOCRT(nan)(const char *); +float RT_NOCRT(nanf)(const char *); + +#endif /* __ISO_C_VISIBLE >= 1999 */ + +#ifndef IPRT_NOCRT_WITHOUT_CONFLICTING_CONSTANTS /*def __USE_GNU*/ +/* + * In GLIBC there are long variants of the XOPEN/SVID constant + * block some pages ago. We need this to get the math tests going. + */ +# define M_El 2.7182818284590452353602874713526625L +# define M_LOG2El 1.4426950408889634073599246810018921L +# define M_LOG10El 0.4342944819032518276511289189166051L +# define M_LN2l 0.6931471805599453094172321214581766L +# define M_LN10l 2.3025850929940456840179914546843642L +# define M_PIl 3.1415926535897932384626433832795029L +# define M_PI_2l 1.5707963267948966192313216916397514L +# define M_PI_4l 0.7853981633974483096156608458198757L +# define M_1_PIl 0.3183098861837906715377675267450287L +# define M_2_PIl 0.6366197723675813430755350534900574L +# define M_2_SQRTPIl 1.1283791670955125738961589031215452L +# define M_SQRT2l 1.4142135623730950488016887242096981L +# define M_SQRT1_2l 0.7071067811865475244008443621048490L +#endif /* !IPRT_NOCRT_WITHOUT_MATH_CONSTANTS */ + +#if 1/*def __USE_GNU*/ + +void RT_NOCRT(sincos)(double, double *, double *); +void RT_NOCRT(sincosf)(float, float *, float *); +void RT_NOCRT(sincosl)(long double, long double *, long double *); +float RT_NOCRT(exp10f)(float); +double RT_NOCRT(exp10)(double); +long double RT_NOCRT(exp10l)(long double); +float RT_NOCRT(log2f)(float); +double RT_NOCRT(log2)(double); +long double RT_NOCRT(log2l)(long double); +float RT_NOCRT(tgammaf)(float); +long double RT_NOCRT(significandl)(long double); +long double RT_NOCRT(j0l)(long double); +long double RT_NOCRT(j1l)(long double); +long double RT_NOCRT(jnl)(int, long double); +long double RT_NOCRT(scalbl)(long double, long double); +long double RT_NOCRT(y0l)(long double); +long double RT_NOCRT(y1l)(long double); +long double RT_NOCRT(ynl)(int, long double); +long double RT_NOCRT(lgammal_r)(long double,int *); +long double RT_NOCRT(gammal)(long double); +#endif + + +RT_C_DECLS_END + + +/** @name fpclassify return values + * @{ */ +#define RT_NOCRT_FP_INFINITE 0x01 +#define RT_NOCRT_FP_NAN 0x02 +#define RT_NOCRT_FP_NORMAL 0x04 +#define RT_NOCRT_FP_SUBNORMAL 0x08 +#define RT_NOCRT_FP_ZERO 0x10 +/** @} */ + +/* bird 2022-08-03: moved this block down so we can prototype isnan & isinf without runnning into the macro forms. */ +#ifndef IPRT_NOCRT_WITHOUT_CONFLICTING_CONSTANTS /* __ISO_C_VISIBLE >= 1999*/ +# define FP_ILOGB0 (-__INT_MAX) +# define FP_ILOGBNAN __INT_MAX + +# ifdef __MATH_BUILTIN_CONSTANTS +# define HUGE_VALF __builtin_huge_valf() +# define HUGE_VALL __builtin_huge_vall() +# define INFINITY __builtin_inf() +# define NAN __builtin_nan("") +# elif RT_MSC_PREREQ(RT_MSC_VER_VC140) && defined(__cplusplus) +/** @todo When were these introduced exactly? 2015, 2017 & 2019 has them. + * However, they only work in C++ even if the c1.dll includes the strings. Oh, well. */ +# define HUGE_VALF __builtin_huge_valf() +# define HUGE_VALL __builtin_huge_val() +# define INFINITY __builtin_huge_val() +# define NAN __builtin_nan("0") /* same as we use in climits */ +# else +# define HUGE_VALF (float)HUGE_VAL +# define HUGE_VALL (long double)HUGE_VAL +# define INFINITY HUGE_VALF +# define NAN (__nanf.__uf) +# endif /* __MATH_BUILTIN_CONSTANTS */ + +# ifndef IPRT_NO_CRT +# define MATH_ERRNO 1 +# endif +# define MATH_ERREXCEPT 2 +# define math_errhandling MATH_ERREXCEPT + +/* XXX We need a <machine/math.h>. */ +# if defined(__ia64__) || defined(__sparc64__) +# define FP_FAST_FMA +# endif +# ifdef __ia64__ +# define FP_FAST_FMAL +# endif +# define FP_FAST_FMAF + +/* Symbolic constants to classify floating point numbers. */ +# define FP_INFINITE RT_NOCRT_FP_INFINITE +# define FP_NAN RT_NOCRT_FP_NAN +# define FP_NORMAL RT_NOCRT_FP_NORMAL +# define FP_SUBNORMAL RT_NOCRT_FP_SUBNORMAL +# define FP_ZERO RT_NOCRT_FP_ZERO +# define fpclassify(x) \ + ((sizeof (x) == sizeof (float)) ? RT_NOCRT(__fpclassifyf)(x) \ + : (sizeof (x) == sizeof (double)) ? RT_NOCRT(__fpclassifyd)(x) \ + : RT_NOCRT(__fpclassifyl)(x)) + +# define isfinite(x) \ + ((sizeof (x) == sizeof (float)) ? RT_NOCRT(__isfinitef)(x) \ + : (sizeof (x) == sizeof (double)) ? RT_NOCRT(__isfinite)(x) \ + : RT_NOCRT(__isfinitel)(x)) +# define isinf(x) \ + ((sizeof (x) == sizeof (float)) ? RT_NOCRT(__isinff)(x) \ + : (sizeof (x) == sizeof (double)) ? RT_NOCRT(isinf)(x) \ + : RT_NOCRT(__isinfl)(x)) +# define isnan(x) \ + ((sizeof (x) == sizeof (float)) ? RT_NOCRT(isnanf)(x) \ + : (sizeof (x) == sizeof (double)) ? RT_NOCRT(isnan)(x) \ + : RT_NOCRT(__isnanl)(x)) +# define isnormal(x) \ + ((sizeof (x) == sizeof (float)) ? RT_NOCRT(__isnormalf)(x) \ + : (sizeof (x) == sizeof (double)) ? RT_NOCRT(__isnormal)(x) \ + : RT_NOCRT(__isnormall)(x)) + +# ifdef __MATH_BUILTIN_RELOPS +# define isgreater(x, y) __builtin_isgreater((x), (y)) +# define isgreaterequal(x, y) __builtin_isgreaterequal((x), (y)) +# define isless(x, y) __builtin_isless((x), (y)) +# define islessequal(x, y) __builtin_islessequal((x), (y)) +# define islessgreater(x, y) __builtin_islessgreater((x), (y)) +# define isunordered(x, y) __builtin_isunordered((x), (y)) +# else +# define isgreater(x, y) (!isunordered((x), (y)) && (x) > (y)) +# define isgreaterequal(x, y) (!isunordered((x), (y)) && (x) >= (y)) +# define isless(x, y) (!isunordered((x), (y)) && (x) < (y)) +# define islessequal(x, y) (!isunordered((x), (y)) && (x) <= (y)) +# define islessgreater(x, y) (!isunordered((x), (y)) && \ + ((x) > (y) || (y) > (x))) +# define isunordered(x, y) (isnan(x) || isnan(y)) +# endif /* __MATH_BUILTIN_RELOPS */ + +# define signbit(x) \ + ((sizeof (x) == sizeof (float)) ? RT_NOCRT(__signbitf)(x) \ + : (sizeof (x) == sizeof (double)) ? RT_NOCRT(__signbit)(x) \ + : RT_NOCRT(__signbitl)(x)) + +typedef double double_t; +typedef float float_t; +#endif /* !IPRT_NOCRT_WITHOUT_MATH_CONSTANTS */ /* __ISO_C_VISIBLE >= 1999 */ + + +#if !defined(RT_WITHOUT_NOCRT_WRAPPERS) && !defined(RT_WITHOUT_NOCRT_WRAPPER_ALIASES) +/* sed -e "/#/d" -e "/RT_NOCRT/!d" -e "s/^.*RT_NOCRT(\([a-z0-9_]*\)).*$/# define \1 RT_NOCRT(\1)/" */ +# define __fpclassifyf RT_NOCRT(__fpclassifyf) +# define __fpclassifyd RT_NOCRT(__fpclassifyd) +# define __fpclassifyl RT_NOCRT(__fpclassifyl) +# define __isfinitef RT_NOCRT(__isfinitef) +# define __isfinite RT_NOCRT(__isfinite) +# define __isfinitel RT_NOCRT(__isfinitel) +# define __isinff RT_NOCRT(__isinff) +# define __isinfl RT_NOCRT(__isinfl) +# define __isnanl RT_NOCRT(__isnanl) +# define __isnormalf RT_NOCRT(__isnormalf) +# define __isnormal RT_NOCRT(__isnormal) +# define __isnormall RT_NOCRT(__isnormall) +# define __signbitf RT_NOCRT(__signbitf) +# define __signbit RT_NOCRT(__signbit) +# define __signbitl RT_NOCRT(__signbitl) +# define signgam RT_NOCRT(signgam) +# define __fpclassifyd RT_NOCRT(__fpclassifyd) +# define __fpclassifyf RT_NOCRT(__fpclassifyf) +# define __fpclassifyl RT_NOCRT(__fpclassifyl) +# define __isfinitef RT_NOCRT(__isfinitef) +# define __isfinite RT_NOCRT(__isfinite) +# define __isfinitel RT_NOCRT(__isfinitel) +# define __isinff RT_NOCRT(__isinff) +# define __isinfl RT_NOCRT(__isinfl) +# define __isnanl RT_NOCRT(__isnanl) +# define __isnormalf RT_NOCRT(__isnormalf) +# define __isnormal RT_NOCRT(__isnormal) +# define __isnormall RT_NOCRT(__isnormall) +# define __signbit RT_NOCRT(__signbit) +# define __signbitf RT_NOCRT(__signbitf) +# define __signbitl RT_NOCRT(__signbitl) +# define acos RT_NOCRT(acos) +# define asin RT_NOCRT(asin) +# define atan RT_NOCRT(atan) +# define atan2 RT_NOCRT(atan2) +# define cos RT_NOCRT(cos) +# define sin RT_NOCRT(sin) +# define tan RT_NOCRT(tan) +# define cosh RT_NOCRT(cosh) +# define sinh RT_NOCRT(sinh) +# define tanh RT_NOCRT(tanh) +# define exp RT_NOCRT(exp) +# define frexp RT_NOCRT(frexp) +# define ldexp RT_NOCRT(ldexp) +# define log RT_NOCRT(log) +# define log10 RT_NOCRT(log10) +# define modf RT_NOCRT(modf) +# define pow RT_NOCRT(pow) +# define sqrt RT_NOCRT(sqrt) +# define ceil RT_NOCRT(ceil) +# define fabs RT_NOCRT(fabs) +# define floor RT_NOCRT(floor) +# define fmod RT_NOCRT(fmod) +# define acosh RT_NOCRT(acosh) +# define asinh RT_NOCRT(asinh) +# define atanh RT_NOCRT(atanh) +# define cbrt RT_NOCRT(cbrt) +# define erf RT_NOCRT(erf) +# define erfc RT_NOCRT(erfc) +# define exp2 RT_NOCRT(exp2) +# define expm1 RT_NOCRT(expm1) +# define fma RT_NOCRT(fma) +# define hypot RT_NOCRT(hypot) +# define ilogb RT_NOCRT(ilogb) +# define lgamma RT_NOCRT(lgamma) +# define llrint RT_NOCRT(llrint) +# define llround RT_NOCRT(llround) +# define log1p RT_NOCRT(log1p) +# define logb RT_NOCRT(logb) +# define lrint RT_NOCRT(lrint) +# define lround RT_NOCRT(lround) +# define nextafter RT_NOCRT(nextafter) +# define remainder RT_NOCRT(remainder) +# define remquo RT_NOCRT(remquo) +# define rint RT_NOCRT(rint) +# define j0 RT_NOCRT(j0) +# define j1 RT_NOCRT(j1) +# define jn RT_NOCRT(jn) +# define scalb RT_NOCRT(scalb) +# define y0 RT_NOCRT(y0) +# define y1 RT_NOCRT(y1) +# define yn RT_NOCRT(yn) +# define gamma RT_NOCRT(gamma) +# define copysign RT_NOCRT(copysign) +# define fdim RT_NOCRT(fdim) +# define fmax RT_NOCRT(fmax) +# define fmin RT_NOCRT(fmin) +# define nearbyint RT_NOCRT(nearbyint) +# define round RT_NOCRT(round) +# define scalbln RT_NOCRT(scalbln) +# define scalbn RT_NOCRT(scalbn) +# define tgamma RT_NOCRT(tgamma) +# define trunc RT_NOCRT(trunc) +# define drem RT_NOCRT(drem) +# define finite RT_NOCRT(finite) +/*# define isinf RT_NOCRT(isinf) - already a macro */ +/*# define isnan RT_NOCRT(isnan) - already a macro */ +# define isnanf RT_NOCRT(isnanf) +# define gamma_r RT_NOCRT(gamma_r) +# define lgamma_r RT_NOCRT(lgamma_r) +# define significand RT_NOCRT(significand) +# define acosf RT_NOCRT(acosf) +# define asinf RT_NOCRT(asinf) +# define atanf RT_NOCRT(atanf) +# define atan2f RT_NOCRT(atan2f) +# define cosf RT_NOCRT(cosf) +# define sinf RT_NOCRT(sinf) +# define tanf RT_NOCRT(tanf) +# define coshf RT_NOCRT(coshf) +# define sinhf RT_NOCRT(sinhf) +# define tanhf RT_NOCRT(tanhf) +# define exp2f RT_NOCRT(exp2f) +# define expf RT_NOCRT(expf) +# define expm1f RT_NOCRT(expm1f) +# define frexpf RT_NOCRT(frexpf) +# define ilogbf RT_NOCRT(ilogbf) +# define ldexpf RT_NOCRT(ldexpf) +# define log10f RT_NOCRT(log10f) +# define log1pf RT_NOCRT(log1pf) +# define logf RT_NOCRT(logf) +# define modff RT_NOCRT(modff) +# define powf RT_NOCRT(powf) +# define sqrtf RT_NOCRT(sqrtf) +# define ceilf RT_NOCRT(ceilf) +# define fabsf RT_NOCRT(fabsf) +# define floorf RT_NOCRT(floorf) +# define fmodf RT_NOCRT(fmodf) +# define roundf RT_NOCRT(roundf) +# define erff RT_NOCRT(erff) +# define erfcf RT_NOCRT(erfcf) +# define hypotf RT_NOCRT(hypotf) +# define lgammaf RT_NOCRT(lgammaf) +# define acoshf RT_NOCRT(acoshf) +# define asinhf RT_NOCRT(asinhf) +# define atanhf RT_NOCRT(atanhf) +# define cbrtf RT_NOCRT(cbrtf) +# define logbf RT_NOCRT(logbf) +# define copysignf RT_NOCRT(copysignf) +# define llrintf RT_NOCRT(llrintf) +# define llroundf RT_NOCRT(llroundf) +# define lrintf RT_NOCRT(lrintf) +# define lroundf RT_NOCRT(lroundf) +# define nearbyintf RT_NOCRT(nearbyintf) +# define nextafterf RT_NOCRT(nextafterf) +# define remainderf RT_NOCRT(remainderf) +# define remquof RT_NOCRT(remquof) +# define rintf RT_NOCRT(rintf) +# define scalblnf RT_NOCRT(scalblnf) +# define scalbnf RT_NOCRT(scalbnf) +# define truncf RT_NOCRT(truncf) +# define fdimf RT_NOCRT(fdimf) +# define fmaf RT_NOCRT(fmaf) +# define fmaxf RT_NOCRT(fmaxf) +# define fminf RT_NOCRT(fminf) +# define dremf RT_NOCRT(dremf) +# define finitef RT_NOCRT(finitef) +# define gammaf RT_NOCRT(gammaf) +# define j0f RT_NOCRT(j0f) +# define j1f RT_NOCRT(j1f) +# define jnf RT_NOCRT(jnf) +# define scalbf RT_NOCRT(scalbf) +# define y0f RT_NOCRT(y0f) +# define y1f RT_NOCRT(y1f) +# define ynf RT_NOCRT(ynf) +# define gammaf_r RT_NOCRT(gammaf_r) +# define lgammaf_r RT_NOCRT(lgammaf_r) +# define significandf RT_NOCRT(significandf) +# define acoshl RT_NOCRT(acoshl) +# define acosl RT_NOCRT(acosl) +# define asinhl RT_NOCRT(asinhl) +# define asinl RT_NOCRT(asinl) +# define atan2l RT_NOCRT(atan2l) +# define atanhl RT_NOCRT(atanhl) +# define atanl RT_NOCRT(atanl) +# define cbrtl RT_NOCRT(cbrtl) +# define ceill RT_NOCRT(ceill) +# define copysignl RT_NOCRT(copysignl) +# define coshl RT_NOCRT(coshl) +# define cosl RT_NOCRT(cosl) +# define erfcl RT_NOCRT(erfcl) +# define erfl RT_NOCRT(erfl) +# define exp2l RT_NOCRT(exp2l) +# define expl RT_NOCRT(expl) +# define expm1l RT_NOCRT(expm1l) +# define fabsl RT_NOCRT(fabsl) +# define fdiml RT_NOCRT(fdiml) +# define floorl RT_NOCRT(floorl) +# define fmal RT_NOCRT(fmal) +# define fmaxl RT_NOCRT(fmaxl) +# define fminl RT_NOCRT(fminl) +# define fmodl RT_NOCRT(fmodl) +# define frexpl RT_NOCRT(frexpl) +# define hypotl RT_NOCRT(hypotl) +# define ilogbl RT_NOCRT(ilogbl) +# define ldexpl RT_NOCRT(ldexpl) +# define lgammal RT_NOCRT(lgammal) +# define llrintl RT_NOCRT(llrintl) +# define llroundl RT_NOCRT(llroundl) +# define log10l RT_NOCRT(log10l) +# define log1pl RT_NOCRT(log1pl) +# define log2l RT_NOCRT(log2l) +# define logbl RT_NOCRT(logbl) +# define logl RT_NOCRT(logl) +# define lrintl RT_NOCRT(lrintl) +# define lroundl RT_NOCRT(lroundl) +# define modfl RT_NOCRT(modfl) +# define nanl RT_NOCRT(nanl) +# define nearbyintl RT_NOCRT(nearbyintl) +# define nextafterl RT_NOCRT(nextafterl) +# define nexttoward RT_NOCRT(nexttoward) +# define nexttowardf RT_NOCRT(nexttowardf) +# define nexttowardl RT_NOCRT(nexttowardl) +# define powl RT_NOCRT(powl) +# define remainderl RT_NOCRT(remainderl) +# define remquol RT_NOCRT(remquol) +# define rintl RT_NOCRT(rintl) +# define roundl RT_NOCRT(roundl) +# define scalblnl RT_NOCRT(scalblnl) +# define scalbnl RT_NOCRT(scalbnl) +# define sinhl RT_NOCRT(sinhl) +# define sinl RT_NOCRT(sinl) +# define sqrtl RT_NOCRT(sqrtl) +# define tanhl RT_NOCRT(tanhl) +# define tanl RT_NOCRT(tanl) +# define tgammal RT_NOCRT(tgammal) +# define truncl RT_NOCRT(truncl) +# define nan RT_NOCRT(nan) +# define nanf RT_NOCRT(nanf) +# define sincos RT_NOCRT(sincos) +# define sincosf RT_NOCRT(sincosf) +# define sincosl RT_NOCRT(sincosl) +# define exp10f RT_NOCRT(exp10f) +# define exp10 RT_NOCRT(exp10) +# define exp10l RT_NOCRT(exp10l) +# define log2f RT_NOCRT(log2f) +# define log2 RT_NOCRT(log2) +# define log2l RT_NOCRT(log2l) +# define tgammaf RT_NOCRT(tgammaf) +# define significandl RT_NOCRT(significandl) +# define j0l RT_NOCRT(j0l) +# define j1l RT_NOCRT(j1l) +# define jnl RT_NOCRT(jnl) +# define scalbl RT_NOCRT(scalbl) +# define y0l RT_NOCRT(y0l) +# define y1l RT_NOCRT(y1l) +# define ynl RT_NOCRT(ynl) +# define lgammal_r RT_NOCRT(lgammal_r) +# define gammal RT_NOCRT(gammal) +#endif + +/* + * Include inlined implementations. + */ +#ifdef RT_ARCH_AMD64 +# include <iprt/nocrt/amd64/math.h> +#elif defined(RT_ARCH_X86) +# include <iprt/nocrt/x86/math.h> +#endif + +#endif /* !IPRT_INCLUDED_nocrt_math_h */ + |