adler32 optimization

Hello again:

Yesterday I wrote a function that would perform the old adler32 hash on given data. My function works perfectly well and so far always has found the correct hash. The problem is big files. I made a file exactly 1 MB and hashed it with various commercial software. They took no more than 5-10 secs. Mine takes 1hr and 5mins....

Since this is a fairly simple hash algo, I was hoping that someone here would recognize and optimizations that could be made to my code. I wrote it from scratch and hopefully right off the bat I am not too inefficient!

Code:

unsigned int adler32(unsigned char *data, unsigned int len)
{
	unsigned int A = 1, B = 0;
	unsigned int i, j;

	for (i = 0; i < len; i++)
	{

		A += data[i];
		for (j = i + 1; j > 0; j--)
			B += data[j-1];
		B++;

		A %= 65521;
		B %= 65521;
	}

	return (B * 65536) + A;
}

http://en.wikipedia.org/wiki/Adler-32
Your inner loop is a killer.

Ya, I got the specifics of the function from wikipedia. I don't want to copy that function verbatim though. I will try to work on my inner loop.

> data[j-1]
If you think about it, you've already summed these (except for the [j-1] entry itself) the last time you were round the loop.

Yeah, dealing with pointers is a bit quicker than running threw indexing the array. Also bit shifting is faster than multiplication as used for your return value. The wikipedia posting would be the way to go.