Re: Robust detection of endianness at compile time.

[Andrew Haley <aph-gcc at littlepinkcloud dot COM>]

Lee Rhodes writes:

 >   Thinking about detection of endianness at *run-time* got me
 > thinking...would this work, at least on machines with 32 bit IEEE 754 float,
 > 32 bit unsigned long and 8 bit char?  
 > 
 > #include <iostream>
 > 
 > union endian {
 > 	float f;
 > 	unsigned long i; 
 > 	unsigned char c[4];
 > };
 > 
 > inline int testEndian();
 > 
 > int testEndian() {
 >       register endian e;
 > 	register int out = 0;
 > 	e.i = 1U << 24;
 > 	if (e.c[3] == 1) out |= 1; //Little Endian Integers
 > 	e.i = 3U << 30;
 > 	if (e.f < 0) out |= 2; //Little Endian Float
 > 	return out;
 > }
 > 
 > using namespace std;
 > int main() {
 > 	int t = testEndian();
 > 	if ((t & 1) == 1) cout << "Little Endian Integer" << endl;
 > 	else cout << "Big Endian Integer" << endl;
 > 	if ((t & 2) == 2) cout << "Little Endian Float" << endl;
 > 	else cout << "Big Endian Float" << endl;
 > }
 > 
 > This may not require a memory access if the compiler actually uses registers
 > and if the machine instruction set utilized small immediate values within
 > the instruction itself.  I don't have a big-endian machine, but if one of
 > you has one could you try it?

It would work on gcc, because gcc allows members of a union to be
written as one type and read as a different incompatible type.  Other
C compilers allow this too.  However, it's is not guaranteed by the
language standards, so it doesn't much help you with truly portable
code.

Andrew.