This allows easier plugging of optimized processing functions in the future, like for SSE or AltiVec.