The arrow of time

It can be downloaded here.

The XXTEA algorithm is a bit peculiar for two reasons: firstly, it's tiny when compared to others (see also the original TEA algorithm) and secondly, it has a sort-of inbuilt chaining mode enabling it to process arbitrarily sized blocks (though their size needs to be a multiple of 4 bytes) without explicit chaining mode like CBC etc.

This implementation is a bit more than 2x slower then the Python Blowfish implementation, though it could be made more speedy by introducing precomputed tables at critical points.