/* * float.h * * Constants related to floating point arithmetic. Also included here are * some non-ANSI mechanisms for accessing the floating point controller. * * $Id: float.h,v af569eb5aad7 2017/02/27 17:46:27 keithmarshall $ * * Written by Colin Peters * Copyright (C) 1997, 1999-2005, 2007, 2010, 2011, 2017, MinGW.org Project * * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice, this permission notice, and the following * disclaimer shall be included in all copies or substantial portions of * the Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OF OR OTHER * DEALINGS IN THE SOFTWARE. * */ #ifndef _MINGW_FLOAT_H #pragma GCC system_header #define _MINGW_FLOAT_H /* * NOTE: * * GCC provides float.h, but it doesn't include the non-standard stuff for * accessing the fp controller. We parse the GCC-supplied header, for its * standard content, and then define the MS-specific extensions here. * * In a MinGW standard Win32 hosted environment, this should be the float.h * found by a system include path search, but this can't be guaranteed; for * a cross-compiler setup, the GCC-supplied header, which is guarded by the * _FLOAT_H___ macro, may be found first, thus, except when compiling any * __LIBIMPL__ stub... * */ #if !defined(_FLOAT_H___) && !defined(__FLOAT_H) /* ...when we didn't find the GCC-supplied header first, we want to pull * it in now; include_next should achieve this, (and we must rely on the * GCC header maintainers to extend us the same courtesy, to get this one * pulled in, when the GCC-supplied header is found first). */ # define _MINGW_FLOAT_H_ 1 /* GCC's needs to see this */ # include_next /* !__CRT_ALIAS __LIBIMPL__( FUNCTION = fpreset ) */ #endif /* All MinGW headers are required to include <_mingw.h> */ #include <_mingw.h> /* Functions and definitions for controlling the FPU. */ #ifndef __STRICT_ANSI__ /* TODO: These constants are only valid for x86 machines */ /* Control word masks for unMask */ #define _MCW_EM 0x0008001F /* Error masks */ #define _MCW_IC 0x00040000 /* Infinity */ #define _MCW_RC 0x00000300 /* Rounding */ #define _MCW_PC 0x00030000 /* Precision */ /* Control word values for unNew (use with related unMask above) */ #define _EM_INVALID 0x00000010 #define _EM_DENORMAL 0x00080000 #define _EM_ZERODIVIDE 0x00000008 #define _EM_OVERFLOW 0x00000004 #define _EM_UNDERFLOW 0x00000002 #define _EM_INEXACT 0x00000001 #define _IC_AFFINE 0x00040000 #define _IC_PROJECTIVE 0x00000000 #define _RC_CHOP 0x00000300 #define _RC_UP 0x00000200 #define _RC_DOWN 0x00000100 #define _RC_NEAR 0x00000000 #define _PC_24 0x00020000 #define _PC_53 0x00010000 #define _PC_64 0x00000000 #ifndef __MINGW_FPCLASS_DEFINED #define __MINGW_FPCLASS_DEFINED 1 /* These are also defined in MinGW ; these duplicates are * needed here, to work around GCC build issues. FIXME: we should * be able to factor this, at least to avoid any potential for the * introduction of inconsistent definitions. * * * Return values for fpclass. */ #define _FPCLASS_SNAN 0x0001 /* Signaling "Not a Number" */ #define _FPCLASS_QNAN 0x0002 /* Quiet "Not a Number" */ #define _FPCLASS_NINF 0x0004 /* Negative Infinity */ #define _FPCLASS_NN 0x0008 /* Negative Normal */ #define _FPCLASS_ND 0x0010 /* Negative Denormal */ #define _FPCLASS_NZ 0x0020 /* Negative Zero */ #define _FPCLASS_PZ 0x0040 /* Positive Zero */ #define _FPCLASS_PD 0x0080 /* Positive Denormal */ #define _FPCLASS_PN 0x0100 /* Positive Normal */ #define _FPCLASS_PINF 0x0200 /* Positive Infinity */ #endif /* __MINGW_FPCLASS_DEFINED */ /* Invalid subconditions (_SW_INVALID also set) */ #define _SW_UNEMULATED 0x0040 /* unemulated instruction */ #define _SW_SQRTNEG 0x0080 /* square root of a neg number */ #define _SW_STACKOVERFLOW 0x0200 /* FP stack overflow */ #define _SW_STACKUNDERFLOW 0x0400 /* FP stack underflow */ /* Floating point error signals and return codes */ #define _FPE_INVALID 0x81 #define _FPE_DENORMAL 0x82 #define _FPE_ZERODIVIDE 0x83 #define _FPE_OVERFLOW 0x84 #define _FPE_UNDERFLOW 0x85 #define _FPE_INEXACT 0x86 #define _FPE_UNEMULATED 0x87 #define _FPE_SQRTNEG 0x88 #define _FPE_STACKOVERFLOW 0x8A #define _FPE_STACKUNDERFLOW 0x8B #define _FPE_EXPLICITGEN 0x8C /* raise SIGFPE */ #ifndef RC_INVOKED _BEGIN_C_DECLS /* Set the FPU control word as cw = (cw & ~unMask) | (unNew & unMask), * i.e. change the bits in unMask to have the values they have in unNew, * leaving other bits unchanged. */ _CRTIMP __cdecl __MINGW_NOTHROW unsigned int _controlfp (unsigned int unNew, unsigned int unMask); _CRTIMP __cdecl __MINGW_NOTHROW unsigned int _control87 (unsigned int unNew, unsigned int unMask); #define _clear87 _clearfp #define _status87 _statusfp _CRTIMP __cdecl __MINGW_NOTHROW unsigned int _clearfp (void); /* Clear the FPU status word */ _CRTIMP __cdecl __MINGW_NOTHROW unsigned int _statusfp (void); /* Report the FPU status word */ /* MSVCRT.DLL provides _fpreset() to initialize the FPU; it initializes * the control register to 0x27F, (thus selecting 53-bit precision), the * status register to zero and the tag word to 0FFFFh. This differs from * Intel's default configuration, (as established by the asm instructions * "finit" and "fninit"), which set the control word to 0x37F, (yielding * 64-bit precison rather than 53-bit). * * Nominally, MinGW prefers to configure the FPU for 64-bit precision, as * that is a better fit for support of GCC's 10-byte long doubles. This * was formerly achieved by provision of an alternative implementation of * _fpreset(), intended to be statically linked into MinGW applications, * (with variants enabling 64-bit precision by default, or alternatively, * when the application is explicitly linked with CRT_fp8.o, providing * 53-bit precision instead). Unfortunately, this linking strategy may * lead to non-deterministic ambiguity over which _fpreset() variant is * called, in the event of the call originating in any user implemented, * or third party DLL. Consequently, support for this strategy has been * withdrawn. _fpreset() now ALWAYS refers to the MSVCRT.DLL function, * (and thus ALWAYS selects 53-bit precision); MinGW applications should * now use the ISO-C99 standard fesetenv() function to configure the FPU, * (as the start-up code now does, to enable 64-bit precision by default, * but continuing to support linking with CRT_fp8.o, to initialize the * configuration to 53-bit precision). */ _CRTIMP __cdecl __MINGW_NOTHROW void _fpreset (void); /* Former MinGW releases also provided fpreset(), implemented as an alias * for _fpreset(), (in whichever guise it was deployed); deprecate this, * but retain the association for backward compatibility, (noting that it * will now redirect the call to fesetenv(), as recommended above, with * the configuration selection set to FE_DFL_ENV). */ __cdecl __MINGW_NOTHROW void fpreset (void) __MINGW_ATTRIB_DEPRECATED; #ifndef __NO_INLINE__ /* Provide the implementation of fpreset() as an inline stub; for this, * we need to expose a minimal opaque subset of , and we assume * that FE_DFL_ENV is equivalent to (const fenv_t *)(0), (as it is, in * the MinGW implementation of ). */ typedef struct __fenv_t fenv_t; __cdecl __MINGW_NOTHROW int fesetenv (const fenv_t *); __CRT_ALIAS __LIBIMPL__( FUNCTION = fpreset ) void fpreset (void){ fesetenv ((const fenv_t *)(0)); } #endif /* Global 'variable' for the current floating point error code. */ #define _fpecode (*(__fpecode())) _CRTIMP __cdecl __MINGW_NOTHROW int *__fpecode (void); /* IEEE recommended functions. Microsoft puts them in , * but they really belong in math.h. */ _CRTIMP __cdecl __MINGW_NOTHROW double _chgsign (double); _CRTIMP __cdecl __MINGW_NOTHROW double _copysign (double, double); _CRTIMP __cdecl __MINGW_NOTHROW double _logb (double); _CRTIMP __cdecl __MINGW_NOTHROW double _nextafter (double, double); _CRTIMP __cdecl __MINGW_NOTHROW double _scalb (double, long); _CRTIMP __cdecl __MINGW_NOTHROW int _finite (double); _CRTIMP __cdecl __MINGW_NOTHROW int _fpclass (double); _CRTIMP __cdecl __MINGW_NOTHROW int _isnan (double); _END_C_DECLS #endif /* ! RC_INVOKED */ #endif /* !__STRICT_ANSI__ */ #endif /* _MINGW_FLOAT_H: $RCSfile: float.h,v $: end of file */