sloppy performance lessons


but why is full blown on one and sloppy on the other side even slower?

theory: instruction cache trashing.

lesson: pedro-sized code helps

needless to say, my autosloppy code went directly to /dev/null