How do I convert between big-endian and little-end

Question

How do I convert between big-endian and little-endian values in C++?

EDIT: For clarity, I have to translate binary data (double-precision floating point values and 32-bit and 64-bit integers) from one CPU architecture to another. This doesn't involve networking, so ntoh() and similar functions won't work here.

EDIT #2: The answer I accepted applies directly to compilers I'm targetting (which is why I chose it). However, there are other very good, more portable answers here.

Answer 1

If you take the common pattern for reversing the order of bits in a word, and cull the part that reverses bits within each byte, then you're left with something which only reverses the bytes within a word. For 64-bits:

x = ((x & 0x00000000ffffffff) << 32) ^ ((x >> 32) & 0x00000000ffffffff);
x = ((x & 0x0000ffff0000ffff) << 16) ^ ((x >> 16) & 0x0000ffff0000ffff);
x = ((x & 0x00ff00ff00ff00ff) <<  8) ^ ((x >>  8) & 0x00ff00ff00ff00ff);

The compiler should clean out the superfluous bit-masking operations (I left them in to highlight the pattern), but if it doesn't you can rewrite the first line this way:

x = ( x                       << 32) ^  (x >> 32);

That should normally simplify down to a single rotate instruction on most architectures (ignoring that the whole operation is probably one instruction).

On a RISC processor the large, complicated constants may cause the compiler difficulties. You can trivially calculate each of the constants from the previous one, though. Like so:

uint64_t k = 0x00000000ffffffff; /* compiler should know a trick for this */
x = ((x & k) << 32) ^ ((x >> 32) & k);
k ^= k << 16;
x = ((x & k) << 16) ^ ((x >> 16) & k);
k ^= k << 8;
x = ((x & k) <<  8) ^ ((x >>  8) & k);

If you like, you can write that as a loop. It won't be efficient, but just for fun:

int i = sizeof(x) * CHAR_BIT / 2;
uintmax_t k = (1 << i) - 1;
while (i >= 8)
{
    x = ((x & k) << i) ^ ((x >> i) & k);
    i >>= 1;
    k ^= k << i;
}

And for completeness, here's the simplified 32-bit version of the first form:

x = ( x               << 16) ^  (x >> 16);
x = ((x & 0x00ff00ff) <<  8) ^ ((x >>  8) & 0x00ff00ff);

Answer 2

From The Byte Order Fallacy by Rob Pyke:

Let's say your data stream has a little-endian-encoded 32-bit integer. Here's how to extract it (assuming unsigned bytes):

i = (data[0]<<0) | (data[1]<<8) | (data[2]<<16) | (data[3]<<24);

If it's big-endian, here's how to extract it:

i = (data[3]<<0) | (data[2]<<8) | (data[1]<<16) | (data[0]<<24);

TL;DR: don't worry about your platform native order, all that counts is the byte order of the stream your are reading from, and you better hope it's well defined.

Note: it was remarked in the comment that absent explicit type conversion, it was important that data be an array of unsigned char or uint8_t. Using signed char or char (if signed) will result in data[x] being promoted to an integer and data[x] << 24 potentially shifting a 1 into the sign bit which is UB.

Answer 3

Simply put:

#include <climits>

template <typename T>
T swap_endian(T u)
{
    static_assert (CHAR_BIT == 8, "CHAR_BIT != 8");

    union
    {
        T u;
        unsigned char u8[sizeof(T)];
    } source, dest;

    source.u = u;

    for (size_t k = 0; k < sizeof(T); k++)
        dest.u8[k] = source.u8[sizeof(T) - k - 1];

    return dest.u;
}

usage: swap_endian<uint32_t>(42).

Answer 4

The procedure for going from big-endian to little-endian is the same as going from little-endian to big-endian.

Here's some example code:

void swapByteOrder(unsigned short& us)
{
    us = (us >> 8) |
         (us << 8);
}

void swapByteOrder(unsigned int& ui)
{
    ui = (ui >> 24) |
         ((ui<<8) & 0x00FF0000) |
         ((ui>>8) & 0x0000FF00) |
         (ui << 24);
}

void swapByteOrder(unsigned long long& ull)
{
    ull = (ull >> 56) |
          ((ull<<40) & 0x00FF000000000000) |
          ((ull<<24) & 0x0000FF0000000000) |
          ((ull<<8) & 0x000000FF00000000) |
          ((ull>>8) & 0x00000000FF000000) |
          ((ull>>24) & 0x0000000000FF0000) |
          ((ull>>40) & 0x000000000000FF00) |
          (ull << 56);
}

Answer 5

If a big-endian 32-bit unsigned integer looks like 0xAABBCCDD which is equal to 2864434397, then that same 32-bit unsigned integer looks like 0xDDCCBBAA on a little-endian processor which is also equal to 2864434397.

If a big-endian 16-bit unsigned short looks like 0xAABB which is equal to 43707, then that same 16-bit unsigned short looks like 0xBBAA on a little-endian processor which is also equal to 43707.

Here are a couple of handy #define functions to swap bytes from little-endian to big-endian and vice-versa -->

// can be used for short, unsigned short, word, unsigned word (2-byte types)
#define BYTESWAP16(n) (((n&0xFF00)>>8)|((n&0x00FF)<<8))

// can be used for int or unsigned int or float (4-byte types)
#define BYTESWAP32(n) ((BYTESWAP16((n&0xFFFF0000)>>16))|((BYTESWAP16(n&0x0000FFFF))<<16))

// can be used for unsigned long long or double (8-byte types)
#define BYTESWAP64(n) ((BYTESWAP32((n&0xFFFFFFFF00000000)>>32))|((BYTESWAP32(n&0x00000000FFFFFFFF))<<32))

Answer 6

Here's how to read a double stored in IEEE 754 64 bit format, even if your host computer uses a different system.

/*
* read a double from a stream in ieee754 format regardless of host
*  encoding.
*  fp - the stream
*  bigendian - set to if big bytes first, clear for little bytes
*              first
*
*/
double freadieee754(FILE *fp, int bigendian)
{
    unsigned char buff[8];
    int i;
    double fnorm = 0.0;
    unsigned char temp;
    int sign;
    int exponent;
    double bitval;
    int maski, mask;
    int expbits = 11;
    int significandbits = 52;
    int shift;
    double answer;

    /* read the data */
    for (i = 0; i < 8; i++)
        buff[i] = fgetc(fp);
    /* just reverse if not big-endian*/
    if (!bigendian)
    {
        for (i = 0; i < 4; i++)
        {
            temp = buff[i];
            buff[i] = buff[8 - i - 1];
            buff[8 - i - 1] = temp;
        }
    }
    sign = buff[0] & 0x80 ? -1 : 1;
    /* exponet in raw format*/
    exponent = ((buff[0] & 0x7F) << 4) | ((buff[1] & 0xF0) >> 4);

    /* read inthe mantissa. Top bit is 0.5, the successive bits half*/
    bitval = 0.5;
    maski = 1;
    mask = 0x08;
    for (i = 0; i < significandbits; i++)
    {
        if (buff[maski] & mask)
            fnorm += bitval;

        bitval /= 2.0;
        mask >>= 1;
        if (mask == 0)
        {
            mask = 0x80;
            maski++;
        }
    }
    /* handle zero specially */
    if (exponent == 0 && fnorm == 0)
        return 0.0;

    shift = exponent - ((1 << (expbits - 1)) - 1); /* exponent = shift + bias */
    /* nans have exp 1024 and non-zero mantissa */
    if (shift == 1024 && fnorm != 0)
        return sqrt(-1.0);
    /*infinity*/
    if (shift == 1024 && fnorm == 0)
    {

#ifdef INFINITY
        return sign == 1 ? INFINITY : -INFINITY;
#endif
        return  (sign * 1.0) / 0.0;
    }
    if (shift > -1023)
    {
        answer = ldexp(fnorm + 1.0, shift);
        return answer * sign;
    }
    else
    {
        /* denormalised numbers */
        if (fnorm == 0.0)
            return 0.0;
        shift = -1022;
        while (fnorm < 1.0)
        {
            fnorm *= 2;
            shift--;
        }
        answer = ldexp(fnorm, shift);
        return answer * sign;
    }
}

For the rest of the suite of functions, including the write and the integer routines see my github project

https://github.com/MalcolmMcLean/ieee754