i used this:

unsigned char *ptr;

ptr = &YOUR FLOAT;
ee_write(*ptr);
ee_write(*(ptr+1));
ee_write(*(ptr+2));
ee_write(*(ptr+3));

where ee_write is a routine that writes one byte to your memory device. This works in my application (using Keil compiler). Keil stores floats in IEEE format (4 bytes). So I need not know anything except to copy the 4 bytes from memory. I later replaced the 4 values to reload from memory.

Steve