will you ever optimize the game, so that, say, i can run a 200 dwarf fort at 100 fps for 20 ingame years?
"Premature optimization is the root of all evil" (Donald Knuth) Remember, DF is in early Alpha, with a multi-decade possible release window. A very vague estimate is 1.0 in around 20 years; the computer you'll be running it on then might be expected to be as much more powerful than a good Core i7 as the Core i7 is more powerful than an 80486. And quite possibly with an even more drastic change in architecture.
At this point in time, it seems moderately likely that future computer developments will be in the direction of "wider", more parallel, architectures. You can get computers with four CPUs each with 15-cores / 30-threads *today*; that's 60 traditional cores, none of which is individually faster than a desktop core. Put in one to four Kepler card for an additional 2,880 micro-cores each if you want. Today, taking serious advantage of that sort of width requires a fair amount of effort; in twenty years, that might be the sort of hardware you have in your phone to run retro games on, and compilers will handle it all automatically.
It's not out of the question that by the time DF is closing in on 1.0, the logical optimization might be to use one traditional core per dwarf, and one micro-core per chunk of map to handle its fluids and temperatures. People might end up saying "Well, I only have an old 96-core, so I try to keep my forts under 80 dwarves".
On the other arm, if we get real quantum computing, programming for that will be a different world. As I understand it, pathfinding would be one of the things it would be better at than traditional computing; perhaps most of DF will run on the deterministic side, with pathing offloaded to the "QPU" card.
The point is, there's no way to tell that far out; it's better at this stage to concentrate (as Toady generally has been) on the actual world generation and gameplay; and worry about serious optimization in another decade or so as the game closes in on Beta status, once it becomes more clear what to optimize *for*.
(Side note: On these time scales, it's not out of the question that some kid playing DF now will grow up to work for Intel and design a CPU *specifically to run DF-like tasks*. Think big!