Thread: Need help with Bit manipulation programming

Indeed. Unfortunately, the instructions can only work with data sizes 1, 2, 4 or 8 bytes. Processor instructions cannot work with other sizes, so what you propose will be cumbersome.
One thing that "might" work is that you treat one byte as one bit. That way you can allocate as many bytes of memory as you need for the number of bits and set them accordingly.
But then using that array will be cumbersome.
Perhaps you should explain a little more on what you're trying to do.

You put bits into an int in multiples of 20, and take them out in multiples of 8.
If any bits are left afterwards, you flush out the remainder.
Like this:

Code:

unsigned int bitsHeld=0, bits=0;

pack20BitsOf(int val)
{
    bitsHeld += 20;
    bits = (bits << 20) | val;
    while (bitsHeld >= 8)
    {
        bitsHeld -= 8;
        writeoutByte(bits >> bitsHeld);
        bits ^= (bits >> bitsHeld) << bitsHeld;
    }
}

flushBits()
{
    if (bitsHeld > 0)
    {
        writeoutByte(bits << (8-bitsHeld));
        bits = 0;
        bitsHeld = 0;
    }
}

or if you want to pack the bits in the other order (which may be a little bit simpler), then like this:

Code:

unsigned int bitsHeld=0, bits=0;

pack20BitsOf(int val)
{
    bitsHeld += 20;
    bits |= val << bitsHeld;
    while (bitsHeld >= 8)
    {
        bitsHeld -= 8;
        writeoutByte(bits & 0xFF);
        bits >>= 8;
    }
}

flushBits()
{
    if (bitsHeld > 0)
    {
        writeoutByte(bits & 0xFF);
        bits = 0;
        bitsHeld = 0;
    }
}

Why do I see missing return type?

Rats! - Guess whether I tried compiling it or not.
void should do for the return type. However if writeoutByte were writing to disk and itself could return failure codes, then I'd make the return type mirror that of writeoutByte, with the failure codes being passed back in the same manner.
It's the algorithm that I was more worried about. Make sure you take the time to understand how it works, as you'll have a better idea how to do such things in future.

If you would like the corresponding "unpacker" and have trouble implementing this yourself, then let me know. Or if you have trouble using the code I've posted, just ask.
flushBits is to only be called once, at the end, after all numbers have been processed.