Thread: Function Compression

Hello folks,

I normally wouldn't bother you guys with silly things but I figured if y'all are
wasting time answering Ryanc's questions, maybe you're having a slow weekend
and won't mind the distraction.

I Need to shave two lines off this function, to bring it down to the required 25
line limit I have to follow. Normally I would just cut some statements and create
a helper function but I happen to be at the function limit for this translation
unit -five, in case you're wondering- and I rather not create another source file
just for this function. Besides, I need the function to be "inlineable"
and I can't include another definition in a header, so it *must* be
in this source file.

It's been a long day and I can't see any other way to shorten the code. Do you see
something obvious I'm missing?

Thanks!

PS: Can't initialize variables during declaration and can't change the coding
style. Can't use multiline macros either, and the max line limit is 80 columns.
Also, I rather not compress everything into one loop as that slows down
the encoding significantly. The compiler can't unroll the loop since it doesn't
know the value of size.

Code:

#define SHIFT_BYTES 3
#define READ_BUFFER SHIFT_BYTES
#define FIRST_SEXTET 26
#define SECOND_SEXTET 20
#define THIRD_SEXTET 14
#define FOURTH_SEXTET 8
#define SEXTET_MASK 0x3f
#define SEXTET 6

static inline void    encode64
    (char *encoded, const char *stream, unsigned int size, unsigned int padding)
{
    unsigned int    shift_count;
    uint32_t        word;

    while (size >= sizeof(word))
    {
        word = *(uint32_t*)(void*)stream;
        word = reverse_byte_order(word);
        *encoded++ = b64_alphabet[(word >> FIRST_SEXTET)];
        *encoded++ = b64_alphabet[(word >> SECOND_SEXTET) & SEXTET_MASK];
        *encoded++ = b64_alphabet[(word >> THIRD_SEXTET) & SEXTET_MASK];
        *encoded++ = b64_alphabet[(word >> FOURTH_SEXTET) & SEXTET_MASK];
        stream += READ_BUFFER;
        size -= READ_BUFFER;
    }
    word = 0;
    shift_count = SHIFT_BYTES;
    while (size--)
        word |= *stream++ << (shift_count-- << 3);
    size = (shift_count != SHIFT_BYTES ? 1 + (SHIFT_BYTES - shift_count) : 0);
    shift_count = FIRST_SEXTET;
    while (size--)
    {
        *encoded++ = b64_alphabet[(word >> shift_count) & SEXTET_MASK];
        shift_count -= SEXTET;
    }
    while (padding--)
        *encoded++ = '=';
}

Code:

        *encoded++ = b64_alphabet[(word >> FIRST_SEXTET)];
        *encoded++ = b64_alphabet[(word >> SECOND_SEXTET) & SEXTET_MASK];
        *encoded++ = b64_alphabet[(word >> THIRD_SEXTET) & SEXTET_MASK];
        *encoded++ = b64_alphabet[(word >> FOURTH_SEXTET) & SEXTET_MASK];

With

Code:

for ( int i = 3 ; i >= 0 ; i-- )
        *encoded++ = b64_alphabet[(word >> (8+SEXTET*i)) & SEXTET_MASK];

At first sight I was going to accuse you of not reading my PS, but then I thought
"Come on, my man Salem is better than that" and on second look I see a compile
time constant, loop unrolling achievement unlocked!
Seriously, how long did it take you to conjure that up?

Although I'm not allowed to use "for" loops, I can always define away that whole
line and guard the definition with with an #if clause -little trick someone named
after a witchy town taught me, wink, wink- and the school bot will be none the wiser.
Stupid rules deserve no respect.

XoXo

> Seriously, how long did it take you to conjure that up?
Couple of minutes - tops.

> Although I'm not allowed to use "for" loops
for loops are just while loops with some decoration.