LordSloth -- it's possible that there is a hit to performance from calculating where the lines begin and end, which is part of the simulation cycle, that's true. I wouldn't think that math for that few lines would impact things all that much, but you never know. More likely, it's a factor of those detailed stats not being as reliable as I would like. In SlimDX they were very reliable, but in Unity I don't even really look at them anymore, to be honest, after I found a number of discrepancies with them. But point taken.
orzelek -- we haven't done a whole lot of exact testing on that side by side, but we've definitely seen a generalized improvement by shifting them. The reason it matters enough to notice with AI War is HOW MANY loops and such there are, with so many ships. If there are 10k ships all moving around, and all of them have a couple of division operators that gets run 20-60 times per second per ship, then shifting that to a shift operation can make a noticeable difference. Nothing enormous, but a few milliseconds shaved per game turn (four game turns per second), which is always a good thing to accomplish, because milliseconds shaved here and there are always a good thing.
Our performance improvements are often measured in milliseconds shaved, honestly. If something takes 3ms under light load, and 45ms under heavy load, and we can get that heavy load case down to 20 or 17 or whatever, then the performance impact is immensely huge. If we can then further reduce it from 17 to 16 in those heavy cases by using shift operators instead of multiplication in a few cases, then that's a lot less notable but still something like a 6% speed improvement on that operation. Enough 6% operations, and suddenly we're at 10ms instead of 17ms, and we've really got something.
Anyway, so my point is that a lot of the optimizations we make only matter because of the scale of the game. When the game is in really low-scale mode, with only a few thousand ships (such as in the tutorial), all the processing for each cycle might be taking 1-3ms max. In those situations, there's no real point to optimization because even a 50% improvement in speed, if we could find a way to get that, would have no effect on anything, literally (because the game is framerate-limiting itself, below a certain ms it doesn't pay any benefits to make it faster, it just gives the CPU more idle time).
So we always look for those extreme cases, and optimize with them in mind. That's usually big battles or really high ship count galaxies, and in those cases the economy of scale kicks in and we can make some "small" optimizations that pay big dividends in aggregate. But for your run of the mill other computer game, or applications other than games, that would be utterly pointless to do. Like with the tutorials in AI War itself, as an example.
