mirror of
https://github.com/boostorg/endian.git
synced 2025-08-01 05:24:39 +02:00
Add changes from conversion.hpp, providing synopsis with in-place and non-modifying operations, unconditional and conditional.
git-svn-id: http://svn.boost.org/svn/boost/sandbox/endian@72091 b8fc166d-592f-0410-95f2-cb63ce0dd405
This commit is contained in:
@@ -1,19 +1,60 @@
|
||||
// endian_flip.hpp -------------------------------------------------------------------//
|
||||
// boost/endian/conversion.hpp -------------------------------------------------------//
|
||||
|
||||
// Copyright Beman Dawes 2010
|
||||
// Copyright Beman Dawes 2010, 2011
|
||||
|
||||
// Distributed under the Boost Software License, Version 1.0.
|
||||
// http://www.boost.org/LICENSE_1_0.txt
|
||||
|
||||
#ifndef BOOST_ENDIAN_FLIP_HPP
|
||||
#define BOOST_ENDIAN_FLIP_HPP
|
||||
#ifndef BOOST_ENDIAN_CONVERSION_HPP
|
||||
#define BOOST_ENDIAN_CONVERSION_HPP
|
||||
|
||||
#include <boost/detail/endian.hpp>
|
||||
#include <boost/cstdint.hpp>
|
||||
|
||||
//------------------------------------- synopsis ---------------------------------------//
|
||||
|
||||
#include <boost/integer.hpp>
|
||||
namespace boost
|
||||
{
|
||||
namespace integer
|
||||
namespace endian
|
||||
{
|
||||
inline void endian_flip(int16_t& x)
|
||||
// unconditional modifying (i.e. in-place) endianness reversal
|
||||
|
||||
inline void flip(int16_t& x);
|
||||
inline void flip(int32_t& x);
|
||||
inline void flip(int64_t& x);
|
||||
inline void flip(uint16_t& x);
|
||||
inline void flip(uint32_t& x);
|
||||
inline void flip(uint64_t& x);
|
||||
|
||||
// unconditional non-modifying endianness reversing copy
|
||||
|
||||
inline void flip(int16_t source, int16_t& target);
|
||||
inline void flip(int32_t source, int32_t& target);
|
||||
inline void flip(int64_t source, int64_t& target);
|
||||
inline void flip(uint16_t source, uint16_t& target);
|
||||
inline void flip(uint32_t source, uint32_t& target);
|
||||
inline void flip(uint64_t source, uint64_t& target);
|
||||
|
||||
// conditional modifying (i.e. in-place) endianness reversal;
|
||||
// no effect if native endianness and specified endianness are the same
|
||||
|
||||
template <class T> inline void to_big(T& x); // if different, convert native to big
|
||||
template <class T> inline void to_little(T& x); // if different, convert native to little
|
||||
template <class T> inline void from_big(T& x); // if different, convert big to native
|
||||
template <class T> inline void from_little(T& x); // if different, convert little to native
|
||||
|
||||
// non-modifying copy, conditionally reversing endianness;
|
||||
// copy the first argument to the second argument, converting to or from the
|
||||
// specified endianness if different than native endianness
|
||||
|
||||
template <class T> inline void to_big(T native, T& big);
|
||||
template <class T> inline void to_little(T native, T& little);
|
||||
template <class T> inline void from_big(T big, T& native);
|
||||
template <class T> inline void from_little(T little, T& native);
|
||||
|
||||
//----------------------------------- implementation -----------------------------------//
|
||||
|
||||
inline void flip(int16_t& x)
|
||||
{
|
||||
char* rep = reinterpret_cast<char*>(&x);
|
||||
char tmp;
|
||||
@@ -22,7 +63,7 @@ namespace integer
|
||||
*(rep+1) = tmp;
|
||||
}
|
||||
|
||||
inline void endian_flip(int32_t& x)
|
||||
inline void flip(int32_t& x)
|
||||
{
|
||||
char* rep = reinterpret_cast<char*>(&x);
|
||||
char tmp;
|
||||
@@ -34,7 +75,7 @@ namespace integer
|
||||
*(rep+2) = tmp;
|
||||
}
|
||||
|
||||
inline void endian_flip(int64_t& x)
|
||||
inline void flip(int64_t& x)
|
||||
{
|
||||
char* rep = reinterpret_cast<char*>(&x);
|
||||
char tmp;
|
||||
@@ -52,7 +93,7 @@ namespace integer
|
||||
*(rep+4) = tmp;
|
||||
}
|
||||
|
||||
inline void endian_flip(uint16_t& x)
|
||||
inline void flip(uint16_t& x)
|
||||
{
|
||||
char* rep = reinterpret_cast<char*>(&x);
|
||||
char tmp;
|
||||
@@ -61,7 +102,7 @@ namespace integer
|
||||
*(rep+1) = tmp;
|
||||
}
|
||||
|
||||
inline void endian_flip(uint32_t& x)
|
||||
inline void flip(uint32_t& x)
|
||||
{
|
||||
char* rep = reinterpret_cast<char*>(&x);
|
||||
char tmp;
|
||||
@@ -73,7 +114,7 @@ namespace integer
|
||||
*(rep+2) = tmp;
|
||||
}
|
||||
|
||||
inline void endian_flip(uint64_t& x)
|
||||
inline void flip(uint64_t& x)
|
||||
{
|
||||
char* rep = reinterpret_cast<char*>(&x);
|
||||
char tmp;
|
||||
@@ -90,7 +131,72 @@ namespace integer
|
||||
*(rep+3) = *(rep+4);
|
||||
*(rep+4) = tmp;
|
||||
}
|
||||
} // namespace integer
|
||||
|
||||
inline void flip(int16_t source, int16_t& target)
|
||||
{
|
||||
const char* s (reinterpret_cast<const char*>(&source));
|
||||
char * t (reinterpret_cast<char*>(&target) + sizeof(target) - 1);
|
||||
*t = *s;
|
||||
*++t = *++s;
|
||||
}
|
||||
|
||||
inline void flip(int32_t source, int32_t& target)
|
||||
{
|
||||
const char* s (reinterpret_cast<const char*>(&source));
|
||||
char * t (reinterpret_cast<char*>(&target) + sizeof(target) - 1);
|
||||
*t = *s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
}
|
||||
|
||||
inline void flip(int64_t source, int64_t& target)
|
||||
{
|
||||
const char* s (reinterpret_cast<const char*>(&source));
|
||||
char * t (reinterpret_cast<char*>(&target) + sizeof(target) - 1);
|
||||
*t = *s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
}
|
||||
|
||||
inline void flip(uint16_t source, uint16_t& target)
|
||||
{
|
||||
const char* s (reinterpret_cast<const char*>(&source));
|
||||
char * t (reinterpret_cast<char*>(&target) + sizeof(target) - 1);
|
||||
*t = *s;
|
||||
*++t = *++s;
|
||||
}
|
||||
|
||||
inline void flip(uint32_t source, uint32_t& target)
|
||||
{
|
||||
const char* s (reinterpret_cast<const char*>(&source));
|
||||
char * t (reinterpret_cast<char*>(&target) + sizeof(target) - 1);
|
||||
*t = *s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
}
|
||||
|
||||
inline void flip(uint64_t source, uint64_t& target)
|
||||
{
|
||||
const char* s (reinterpret_cast<const char*>(&source));
|
||||
char * t (reinterpret_cast<char*>(&target) + sizeof(target) - 1);
|
||||
*t = *s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
*++t = *++s;
|
||||
}
|
||||
|
||||
} // namespace endian
|
||||
} // namespace boost
|
||||
|
||||
#endif // BOOST_ENDIAN_FLIP_HPP
|
||||
#endif // BOOST_ENDIAN_CONVERSION_HPP
|
||||
|
Reference in New Issue
Block a user