You have to be careful with lookup tables. They can be impressive each at a time when microbenchmarking a single loop, but in the aggregate hurt performance by crowding other useful data out of limited cache space.