This is currently an attempt at getting the compiler to auto-vectorize the heaviest parts of the decoding process.
bschwind
