Porting code to a different system, endianness, bytes swapping

[ppanta]

Hi,
I'm trying to read data that generated in a big-endian system (SPARC) in the little-endian cpu (intel). The following is the code I used.

Please note that the following codes are fully working in SUN but not in the LINUX system I'm currently using.
===========================================
void Initialize(.... char *FreezeTableDSN)

{

int j, handle;

float h_clim[12][NO_OF_ROWS];
....
if( (handle=open( FreezeTableDSN, O_RDONLY ) ) == -1 ) {

perror("Open error for FreezeTable");

exit(1);

}

if( read( handle, h_clim, sizeof(h_clim) ) != sizeof(h_clim) ) {

perror("Read error for FreezeTable");

exit(1);

}
for(j=0; j<NO_OF_ROWS; j++ )
FreezeLevel[j] = h_clim[month-1][j];
...
============================================
(Freezelevel is declared as 'float' and month is declared as 'int')

The problem is that the data doesn't get read properly as a result of cpu system. Thus, I'm trying to swap the bytes for readability. However, I'm confuse as to how and what should I swap?

First, should I swap the whole stream of data before it gets translate into the the h_clim ? or should I translate cell-by-cell in the h_clim[month-1][j]?

Second, assuming that I need to translate cell by cell, I proceeded to define a macro
===============================
#define SWAP32(x) \

x = ((((x) & 0x00ff0000) >> 8) | \

(((x) & 0x0000ff00) << 8) | \

(((x) & 0xff000000) >> 24) | \

(((x) & 0x000000ff) << 24) )
===============================
However, the compiler says something like "Invalid operand to binary &" when I try to use the macro as for the h_clim or FreezeLevel.
E.G. SWAP32(FreezeLevel(j))

Any help would be greatly appreciate.

Thank you.

[ppanta]

Sorry, forgot to note the detail of the machine and the compiler I'm using.

OS: Fedora 7
Compiler: native C compiler (I think it's GNU GCC)
Language: C

[JosAH]

Have a look at the htonl(), htons(), ntohl() and ntohs() macros.
The acronyms stand for: HostToNetwork and NetworkToHost; the 'l' and 's'
stand for Long and Short respectively. These macros convert an integer from
your system's endianness to network order and vice versa. The network order
is big endian per definition. (a la SPARC processors etc.)

kind regards,

Jos

[ppanta]

Thx, it solves the problem.