Interesting thread.
I've also been doing FPS tests as of late, primarily after I noticed atom smashing 5k stone and other garbage had a negligible effect on FPS with a ~60 dwarf/10 cat/4x4 embark fort. (In 40d I'd get about 10 FPS back from atom-smashing on average, now, on 31.08 I recover about 3-4.) I'm less interested in FPS lost to pathing (as that should be reasonably constant if the number of dwarfs and animals stay the same), and more curious about the "FPS Drain = [some constant] times (Fortress Age)" effect. (I find the game to be unplayable around 25-30.... That 4x4 embark was down to 15.)
So I embarked on a 2x2 map, set the FPS max to 400 and observed the following over two tests:
[A brief note on the FPS at the start of game: Nothing spectacular, but with minimal changes to the real estate I was getting around 240 FPS unpaused, I think. I didn't make note of it as things were flying by and I often had to drop the FPS manually down to ~100.]
Control Point (start of experiments), Year 3 (3012, for the record): 23 dwarfs, 5 cats (but one pesky immigrant female cat), ~180 FPS average. Invaders turned off. Caverns not breached. Large stockpile of booze and prepared meals, enough to last ~3 years (estimate).
A: Control Experiment: What is the effect on FPS with zero changes to the fort over the course of two years.
Details of Experiment: I stopped producing everything. No new items, no digging, no engraving, no food, no farming. The only "new" things were the crap dragged to my fort by caravans (and promptly dragged back), two new dwarf babies, and seven kitten skulls. Some stone smoothing was going on, but again no engraving. No pathing changes. Nothing. Lots of bored, idle, and drunk dwarfs. No extra untied animals (only 4 war dogs chained at strategic points and the 5 wandering felines).
Result: At the end of two years my FPS was hovering around 95 unpaused, and sometimes dropped as low as 75 (for no observable reason, except possibly early caravan calculations). I occasionally reach 110, but not for long.
B: Active Fort Experiment: Standard 1:1 production items only (ie: rock instruments, rock blocks, etc). No farming. Again, no cooking. No "assembling" reactions where multiple items become one. Otherwise same as the control experiment, except dwarfs are not idle and are pathing to/from their jobs (mostly dumping stone).
Result: Identical response as Control. ~95 FPS unpaused after 2 years. Sometimes 110, sometimes 75. 2 fewer kitten skulls.
Extremely small sample size so far, but I'm going to repeat a few times until I can get a design down that is survivable on my system with more than a few dozen dwarfs. I'd really like to be able to play 120+ dwarf forts again, and unless I can find some "cause" for what's killing my FPS I might just have to mark that off as an impossible dream. I've tried revising my fortress "flow" using "ramps," and other suggestions already covered in this topic with minimal results (My forts were fairly efficient to begin with). Likewise with atom-smashing, as I already mentioned.
As far as I can tell there's nothing we're doing that specifically causes the FPS drain. But again, this is so far an extremely small sample.
(And for the record: After tweaking everything in 2010 for optimal settings I get about 2/3 to 1/2 of the FPS as I did in 40d at ~60 dwarfs and identical fort layouts. Obviously there have been world and code changes, and there's about 400% more matter under the ground, but that doesn't explain the age-related FPS drain I experience. All experiments on a Windows 7 Pro box, 3.0 GHz Intel Pentium-D w/ 4 GB of 533 MHz DDR2 RAM [not that DF sees more than 2 or Windows more than ~3.2], 800 MHz Bus, and an ATI Radeon HD 4800 display adapter. Hard Disks are 7200 RPM of varying structures and formats, DF might be living on one of the drives that's split Linux and, other, but nothing has changed there.)