I haven't played it much, but I've always felt it would make an interesting iPad game. Seems like it wouldn't be too complex to put an objective-c wrapper around the code.
The problem is simple: The iPad lacks the horsepower to run the game at approaching a playable rate. Now if you were running it on another machine and playing by remote desktop.... Maybe. But to me this game simply demands a keyboard.
My 3.2 GhZ Dual Core system chokes at times with it, so yes. The biggest catch is that the core of the game is not multi-threaded (though graphics may be depending on config). The throughput of a single processor core is the most important factor in the simulation's performance. The iPad, while nice for the rather passive computations it's expected to perform, just doesn't have the muscle.
Now assuming the game could be ported to iOS (wrapper/emulator or whatnot), 1 GHz would allow you to generate a small world, and maybe play for a year or two at a low FPS. (I used to play the older version on my 1.2 GHz AMD "netbook," so it's certainly possible.) But in fortress mode, when you've got a great many dwarfs doing pathfinding (say, 80+) and a whole bunch of other stuff being tracked by the game, the simulation can be extremely taxing on system resources. Ultimately, most forts die from what is called "FPS Death," when the simulation frames completed per second reaches an unplayable level.
I wonder how much of the simulation code could be appropriately ported to some sort of GPGPU (CUDA,OpenCL, etc) code. There'd be something nicely ironic about running an ascii-frontend game which required a serious GPU to play :)
That's been proposed in the bay12 forums several times. While I don't think that Tarn ever officially responded, the reason that's not going to happen seems to boil down to it requiring a complete rewrite of all the code behind the game. Which isn't going to happen any time soon - I recall that, a while back, he was using implementing new features as a reward to get him to do bugfixes; putting off the fun things until after he'd done the unfun things. Rewriting all of the code would be extremely unfun, take forever, and prevent implementing new features.
You would probably want to run it on something like a POWER7 for best possible performance then, which have fewer but faster cores than comparable x64s, if there is an AIX port.
