C: working with char array as 8 byte blocks

Question

I have an array of chars and a pointer to it. I need to add the first 8Bytes (binary value of the 8 Bytes) to the second 8 bytes modulo 2exp(64). How could I do that?

I found a solution. But it is definitely not good to do such things (see the code). Nevertheless it would be nice to have the result in chars array.

void init(const unsigned char *k) {
    uint64_t *kp = k;
    uint64_t sum = (*kp + *(kp+1)) % pow(2,64)
}

Thanks for help

Answer 1

You don't need the % pow(2,64), and in fact it introduces errors, since it will convert the result to and from double. Arithmetic in the type uint64_t is already modulo 2^64.

The issue of reading 8 bytes as a uint64_t is basically about what C implementations you want it to work on, and the format of the 8 bytes in your char array.

If you know that the 8 bytes in the char array have the same endian-ness as your system (that is, most significant byte first if your system is big-endian, most significant byte last if your system is little-endian), then what you're doing is probably OK. Whoever supplies the buffer will have to make sure that it's correctly aligned, though, if there are alignment restrictions for uint64_t on your system.

If the data came from some file or over the internet, then the endian-ness might not match the system, and it probably won't be aligned correctly.

If the 8 bytes are little-endian, this code would fail on a big-endian system, so if you wanted your code to be portable you'd have to do something else. There are systems where a unit64_t doesn't occupy 8 bytes, and even memcpy doesn't work. The specification of your function might mean that it's completely inapplicable to such systems, but in some circumstances it does make sense to convert an array of 8 char to a uint64_t, irrespective of whether actually CHAR_BIT is bigger than 8.

For a reliable basic solution: read one byte at a time and add the parts together (left-shifting by 0, then 8, and so on). If you don't like the look of that, make some assumptions about what platforms you need to support, and optimize.

However you read the data, if you want the result back in a char array then reverse the process - in the case of your code, *((uint64_t*)output_buffer) = kp[0] + kp[1];.

Answer 2

I'm pretty sure that 2pow64 modulo naturally comes out of the precision of your data (64 bits in this case)

so create 2 variables integers, 64 bits precision, unsigned. called a, and b.

and move you first 4 bytes into a, 4 next bytes into b, add a to b.

In fact you just don't need to do modulo.