Yes, the technology page calls out that on average it is 3.7X faster with just LLVM and algorithmic improvements. We are trying to push more value into Vitesse X by adding vector processing, and maybe include some of our threading work in the next version.