Merge pull request #173 from delroth/movbe Optimize memory access on Haswell by using MOVBE when possible.