optimizations have much better floating point performance. [1] http://elinux.org/RPi_Distributions#What_is_armhf