Boost C++ Libraries

...one of the most highly regarded and expertly designed C++ library projects in the world. Herb Sutter and Andrei Alexandrescu, C++ Coding Standards

This is the documentation for an old version of Boost. Click here to view this page for the latest version.

boost/atomic/detail/float_sizes.hpp

/*
 * Distributed under the Boost Software License, Version 1.0.
 * (See accompanying file LICENSE_1_0.txt or copy at
 * http://www.boost.org/LICENSE_1_0.txt)
 *
 * Copyright (c) 2018 Andrey Semashev
 */
/*!
 * \file   atomic/detail/float_sizes.hpp
 *
 * This header defines macros for testing buitin floating point type sizes
 */

#ifndef BOOST_ATOMIC_DETAIL_FLOAT_SIZES_HPP_INCLUDED_
#define BOOST_ATOMIC_DETAIL_FLOAT_SIZES_HPP_INCLUDED_

#include <float.h>
#include <boost/atomic/detail/config.hpp>

#ifdef BOOST_HAS_PRAGMA_ONCE
#pragma once
#endif

// Detect value sizes of the different floating point types. The value sizes may be less than the corresponding type sizes
// if the type contains padding bits. This is typical e.g. with 80-bit extended float types, which are often represented as 128-bit types.
// See: https://en.wikipedia.org/wiki/IEEE_754
// For Intel x87 extended double see: https://en.wikipedia.org/wiki/Extended_precision#x86_Architecture_Extended_Precision_Format
// For IBM extended double (a.k.a. double-double) see: https://en.wikipedia.org/wiki/Long_double#Implementations, https://gcc.gnu.org/wiki/Ieee128PowerPC
#if (FLT_RADIX+0) == 2

#if ((FLT_MANT_DIG+0) == 11) && ((FLT_MAX_EXP+0) == 16) // IEEE 754 binary16
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 2
#elif ((FLT_MANT_DIG+0) == 24) && ((FLT_MAX_EXP+0) == 128) // IEEE 754 binary32
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 4
#elif ((FLT_MANT_DIG+0) == 53) && ((FLT_MAX_EXP+0) == 1024) // IEEE 754 binary64
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 8
#elif ((FLT_MANT_DIG+0) == 64) && ((FLT_MAX_EXP+0) == 16384) // x87 extended double
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 10
#elif ((FLT_MANT_DIG+0) == 106) && ((FLT_MAX_EXP+0) == 1024) // IBM extended double
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 16
#elif ((FLT_MANT_DIG+0) == 113) && ((FLT_MAX_EXP+0) == 16384) // IEEE 754 binary128
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 16
#elif ((FLT_MANT_DIG+0) == 237) && ((FLT_MAX_EXP+0) == 262144) // IEEE 754 binary256
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 32
#endif

#if ((DBL_MANT_DIG+0) == 11) && ((DBL_MAX_EXP+0) == 16) // IEEE 754 binary16
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 2
#elif ((DBL_MANT_DIG+0) == 24) && ((DBL_MAX_EXP+0) == 128) // IEEE 754 binary32
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 4
#elif ((DBL_MANT_DIG+0) == 53) && ((DBL_MAX_EXP+0) == 1024) // IEEE 754 binary64
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 8
#elif ((DBL_MANT_DIG+0) == 64) && ((DBL_MAX_EXP+0) == 16384) // x87 extended double
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 10
#elif ((DBL_MANT_DIG+0) == 106) && ((DBL_MAX_EXP+0) == 1024) // IBM extended double
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 16
#elif ((DBL_MANT_DIG+0) == 113) && ((DBL_MAX_EXP+0) == 16384) // IEEE 754 binary128
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 16
#elif ((DBL_MANT_DIG+0) == 237) && ((DBL_MAX_EXP+0) == 262144) // IEEE 754 binary256
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 32
#endif

#if ((LDBL_MANT_DIG+0) == 11) && ((LDBL_MAX_EXP+0) == 16) // IEEE 754 binary16
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 2
#elif ((LDBL_MANT_DIG+0) == 24) && ((LDBL_MAX_EXP+0) == 128) // IEEE 754 binary32
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 4
#elif ((LDBL_MANT_DIG+0) == 53) && ((LDBL_MAX_EXP+0) == 1024) // IEEE 754 binary64
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 8
#elif ((LDBL_MANT_DIG+0) == 64) && ((LDBL_MAX_EXP+0) == 16384) // x87 extended double
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 10
#elif ((LDBL_MANT_DIG+0) == 106) && ((LDBL_MAX_EXP+0) == 1024) // IBM extended double
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 16
#elif ((LDBL_MANT_DIG+0) == 113) && ((LDBL_MAX_EXP+0) == 16384) // IEEE 754 binary128
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 16
#elif ((LDBL_MANT_DIG+0) == 237) && ((LDBL_MAX_EXP+0) == 262144) // IEEE 754 binary256
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 32
#endif

#elif (FLT_RADIX+0) == 10

#if ((FLT_MANT_DIG+0) == 7) && ((FLT_MAX_EXP+0) == 97) // IEEE 754 decimal32
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 4
#elif ((FLT_MANT_DIG+0) == 16) && ((FLT_MAX_EXP+0) == 385) // IEEE 754 decimal64
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 8
#elif ((FLT_MANT_DIG+0) == 34) && ((FLT_MAX_EXP+0) == 6145) // IEEE 754 decimal128
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE 16
#endif

#if ((DBL_MANT_DIG+0) == 7) && ((DBL_MAX_EXP+0) == 97) // IEEE 754 decimal32
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 4
#elif ((DBL_MANT_DIG+0) == 16) && ((DBL_MAX_EXP+0) == 385) // IEEE 754 decimal64
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 8
#elif ((DBL_MANT_DIG+0) == 34) && ((DBL_MAX_EXP+0) == 6145) // IEEE 754 decimal128
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE 16
#endif

#if ((LDBL_MANT_DIG+0) == 7) && ((LDBL_MAX_EXP+0) == 97) // IEEE 754 decimal32
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 4
#elif ((LDBL_MANT_DIG+0) == 16) && ((LDBL_MAX_EXP+0) == 385) // IEEE 754 decimal64
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 8
#elif ((LDBL_MANT_DIG+0) == 34) && ((LDBL_MAX_EXP+0) == 6145) // IEEE 754 decimal128
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE 16
#endif

#endif

// GCC and compatible compilers define internal macros with builtin type traits
#if defined(__SIZEOF_FLOAT__)
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT __SIZEOF_FLOAT__
#endif
#if defined(__SIZEOF_DOUBLE__)
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE __SIZEOF_DOUBLE__
#endif
#if defined(__SIZEOF_LONG_DOUBLE__)
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE __SIZEOF_LONG_DOUBLE__
#endif

#if !defined(BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT) || !defined(BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE) || !defined(BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE)

#define BOOST_ATOMIC_DETAIL_ALIGN_SIZE_TO_POWER_OF_2(x)\
    ((x) == 1u ? 1u : ((x) == 2u ? 2u : ((x) <= 4u ? 4u : ((x) <= 8u ? 8u : ((x) <= 16u ? 16u : ((x) <= 32u ? 32u : (x)))))))

// Make our best guess. These sizes may not be accurate, but they are good enough to estimate the size of the storage required to hold these types.
#if !defined(BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT) && defined(BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE)
#define BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT BOOST_ATOMIC_DETAIL_ALIGN_SIZE_TO_POWER_OF_2(BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE)
#endif
#if !defined(BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE) && defined(BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE)
#define BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE BOOST_ATOMIC_DETAIL_ALIGN_SIZE_TO_POWER_OF_2(BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE)
#endif
#if !defined(BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE) && defined(BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE)
#define BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE BOOST_ATOMIC_DETAIL_ALIGN_SIZE_TO_POWER_OF_2(BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE)
#endif

#endif // !defined(BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT) || !defined(BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE) || !defined(BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE)

#if !defined(BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT_VALUE) || !defined(BOOST_ATOMIC_DETAIL_SIZEOF_FLOAT) ||\
    !defined(BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE_VALUE) || !defined(BOOST_ATOMIC_DETAIL_SIZEOF_DOUBLE) ||\
    !defined(BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE_VALUE) || !defined(BOOST_ATOMIC_DETAIL_SIZEOF_LONG_DOUBLE)
#error Boost.Atomic: Failed to determine builtin floating point type sizes, the target platform is not supported. Please, report to the developers (patches are welcome).
#endif

#endif // BOOST_ATOMIC_DETAIL_FLOAT_SIZES_HPP_INCLUDED_