How would Skyrim react if it had to display and store individual scars, missing teeth, broken nails, artifact decorations, etc? How could it remember the exact coordinates of each spatter of blood and vomit? How would it deal with injuries being crippling and healing on their own, how would it depict a goblin splatting on the ground while still depicting dwarves falling a z-level and just bruising their chest?
Giving Df good Skyrimesque graphics would probably take as much code as tbe rest of DF, plus several times the graphics coding of Skyrim.
Let's assume for the moment that Skyrim's graphics are roughly half characters and half setting.
The character models would need expansion for various procedurally generated options; things like scars, tattoos, etc are fairly easy, while expanding the head model to model teeth individually would be easy to set up, but slow rendering down unless carefully optimized. Then they'd need some more base models for other races. I'd guess maybe 5 times as much here. They'd also need armor and clothing for other styles, biomes, etc. and a procedural option for handling decorations; perhaps another 5x to 20x depending on how elaborate and specific you get. What that doesn't cover is the anthropomorphic animal races. What you'd really need would be a procedural template to "anthorpomorphize" a beast model; this is fairly doable for anything that's basically 4-limbed, but would require more work for things on radically different body plans. You'd also need fairly decent models for a much wider variety of animals; this could easily be another 10x, 20x, or more.
As far as setting goes, you'd need nearly that much again for every "majorly different" biome you wanted to depict. How many you'd need is a matter of opinion; 5x might do it for a very simple take, 10x would be more reasonable and 20x would allow you to get pretty elaborate. Much of this would be plant models. Code to generate different sorts of procedural terrain would need to be worked on; the existing setup is designed to give a fairly limited number of "looks" optimized for a craggy, Norse sort of feel. New terrain modeling routines takes a lot of time to develop but not much space; the hard part is figuring out what combination of inputs to which generators give you "rolling hills of the Shire" versus "Mordor".
You'd also need a procedural item decoration system, which probably inherits from other procedural systems but ends up being a frame-rate problem if not carefully optimized (and possibly even if it is).
All told, I'd guess that the most minimal version that would meet the description would be roughly ten times the graphics size of Skyrim, and run at similar speeds in wilderness areas but drop to half speed at best (quite possibly more like a quarter) in busy towns, meeting halls, etc. due to having to process a lot more specific info. In practice, you'd probably have a variable-detail system, where the "draw distance" for a lot of details is shortened significantly in the presence of crowds; and some details are abstracted away. (For instance, in a crowd the aforementioned teeth might be simplified from 3d objects to a pair of curved 2d arcs representing only the front teeth and merely tracking their absence or presence via simple rectangles of dark or ivory; 1/100th the rendering cost for a visual impression that's nearly as good for anyone in the crowd not adjacent to you.) A more complete implementation could easily be a couple times that, but probably still less than a hundred times the graphics size.
As for simulation... everyone should read the classic short story
"I don't know, Timmy, being God is a big responsibility" for a starter. (Seriously, go read it now, before the next several paragraphs, which will have spoilers.)
The story generally works if the setting has two criteria: genuinely infinite computing (not currently believed to be possible due to the quantum nature of reality; reality is not analog enough), and a causal feedback loop, which is a neat plot point but I haven't found any indication that it's likely.
Somewhat more generally, to get an infinite stack of simulations as in the story, you need a setting that is mathematically analog (ie, distances, velocities, angles, etc. are true mathematical real numbers with physically meaningful infinite possible precision). This means you can in principle encode anything, or in fact *everything*, onto something like the spin axis angle of a single particle using what is basically a
Hilbert's Hotel inspired encoding scheme with increasingly higher-order functions. (Given the comments in the story, this appears to be the way their infinite computer works.)
We're pretty sure our universe doesn't work that way. However, it's possible that if you have a situation where computational power grows faster than the complexity of the universe, both approaching an asymptote, you might be able to *simulate* the above infinite simulation, by doing so exponentially slower (effectively, trading functionally-infinite calculation time for infinite precision). At one point it was thought by some that quantum computing in a so-called "Big Crunch" scenario at the end of the universe might have this property, which lead to some of the original speculation about the odds of us being in the middle of a stack of simulations. Given current astrophysics a Big Crunch seems unlikely and perhaps impossible; given current information theory, asymptotically increasing computation on quantized finite matter/energy seems impossible.
So, according to our current and likely understanding, it's not possible to accurately simulate a universe like ours with less than a similar universe's worth of stuff. This means no infinite chains of nested simulations (the "turtles all the way down" plan of the short story), which dramatically decreases the odds we're in a simulation.
It doesn't remove them, however. If our universe exists, as some theories seem to think, as a "bubble" of sorts in a much higher-dimensional metaverse; then it is possible (and perhaps even moderately likely) that our seeming quantum reality is precisely the simulation limitations of some higher-order sim running in the metaverse. In the same way that
Conway's Game of Life or of course DF is a highly simplified, far more quantized version of our reality that remains interesting; our reality might be a highly simplified, far more quantized version / simulation / game / offshoot from a "higher order" reality. It seems that an infinite chain of these is unlikely; each layer down requires one (or more) fewer dimensions (or perhaps one cardinality level less infinity in scope); and it seems likely that in a real system, the "starting" situation would have a finite number of dimensions and thus an eventual limit on layers of simulation. This might well be dozens; the original Bosonic string theory posited 26, but is currently thought to not be reasonable, with 11 or 10 seeming more likely if applicable at all.
So, cheer up! Odds that we're the "original" universe and not basically the @s in some higher order setting's version of DF may be as high as 1 in 10 or so!
P.S.: There is an additional eccentric option, which is that if you have a finite universe, but infinitely repeatable time travel, you may be able to simulate infinite matter/energy by infinite time loops. A very limited version of this is used by Stephen Baxter's
Xeelee, who use finite time travel to leverage their abilities; but that setting does not appear to allow infinite time travel and they are thus unable to improve their situation by more than what is effectively a constant. (Spoiler: Eventually, they (and all baryonic intelligences) loose to dark-matter intelligences, and are forced to evacuate to another universe. It's implied that there is a large finite or very-low-order infinite number of alternate universes, and a fairly small maximum number of meaningful dimensions, so infinite-stack simulation doesn't work in that setting either.)
P.P.S: A more sophisticated take on the odds looks at our place in the spatial dimension stack. 0 dimensions can't simulate anything. 1 dimensional simulations do exist (IIRC they were used in early stellar evolution models, as a homogeneously layered sphere can be modeled by a suitably arranged 1-dimensional ray simulation with dramatically less computer power needed). Two-dimensional simulations of various fidelity are common, and we are 3-dimensional. If there are dimensions 4-11 (or some similar number) "above" us, then odds we're the "top" level seem less likely, as each dimension gets dramatically more "interesting". One would have to posit some unique property of 3-space that makes it more organizable than higher spaces despite its limitations; there are some odd mathematical possibilities to start from but that begins to get fairly speculative.
Author
Topic: What would be the file size of dwarf fortress if it had skyrim graphics? (Read 25739 times)