Bay 12 Games Forum

I like the explanation of the constant-duration first step being "on credit" so that the time costs are assessed after travel... that makes good sense.  Now that we have a reasonable explanation for that phenomenon, it looks like the steady-state results confirm the Pythagorean worldview.

(Out of curiosity, Toady, are you using a true SQRT function for the Euclidean distance, or did you use an approximation like this one to save CPU cycles?)

Lectorog, this is awesome stuff.  I can see that you did two rounds of tests, with dwarves going out-then-in, and I can see that you broke down the results by direction.  I'm a little puzzled by the fact that the first orthogonal round took the longest, and the first diagonal round was the shortest.  Can you think of anything that would explain that?  Maybe observation bias? 

If we assume some sort of measurement bias in the first round -- even a fencepost or counting error could account for it -- then it seems reasonable to conclude that dwarves take about 1.4 (probably √2) steps to move diagonally.  I'm hosting my weekly D&D table tonight, and I've got to do some hardware work on my wife's PC, but I'll try to reproduce your results soon using your arena file. 

If enough of us can reproduce your results, I think we should post them on the Wiki somewhere for all to see, e.g.

"Dwarves take about √2 times longer to traverse a diagonal tile than an orthogonal tile (Lectorog, 2011) which means that -- just as in our universe -- the corners of a square room are 41% further than the midpoint of its walls from the center point."

OK, I did a full round of testing. Here are the results:

Code: [Select]

ROUND I
Out-In
W:12-12,10-11,11-11,11-11
N:11-12,11-11,11-11,10-11
E:11-11,10-10,11-10,11-11
S:11-11,10-11,11-11,11-11
----
NW:11-11,14-15,15-14,14-15
NE:11-11,14-14,14-14,15-14
SE:12-11,15-15,15-14,14-15
SW:12-12,14-15,15-14,14-14

Something to take note of is that a square is "technically" round in Dwarf Fortress, since it takes the same amount of time for something to go diagonally as it takes for it to go in any other direction.

Which must mean that circles are squares.

But squares are also circles.

Spoiler (click to show/hide)

Wrong -- diagonal movement costs 1.4 squares of movement.  A nice circle of diameter 15 is rows of 5,9,11, two rows of 13, five rows of 15, two 13s, 11, 9, 5.  It's 7 tiles flat-to-flat, and 5*1.4 = 7 tiles diagonally.  I use it for my cylindrical layouts.

Woodcutter skills up very quickly, and carpentry really only gets you high-quality beds, barrels, and bins... so make your axe-wielding dwarf a 5 Teacher / 3 Axedwarf / 2 armor-user, and then let him go practice chopping trees with his axe.  (If he gets attacked by wildlife this also gives him a fighting chance to survive!)  Around the same time he has cleared the map of trees, you'll have a migrant wave.  If you've planned carefully, his skill points in teacher will help all of your novices level up to the point where they can start instructing each other much more quickly.

Your leader (Appraiser/Judge/etc.) should instead have a smattering of administrative skills.  One point each in Appraiser, Negotiator, Judge of Intent, Organizer, and Record-Keeper; with his remaining five skill points he can be a novice doctor or he can be a 5-point expert in an economic skill like leatherworking.  Since leatherworking can create valuable bags *and* useful armor items, it's a good safe bet.

Since you'd have to dig out the levels below to get falling damage, why not just have the FB fall into a pit and then bring the the ten floors above down on it? Or make a Dwarven Machine Gun, bringing water and magma together on a hole between two hatches, so the obsidian is unsupported by solid ground and immediately caves in? Really, if the game's going to throw indestructible FBs that throw out cave-in dust at you, why not throw several tons of obsidian a second at them?

Mostly for lack of time and resources.  This was an early-game fortress death that I'd like to include in my "what not to do" standard operating procedures.  I'm looking at setting up a one- or two-layer trap system that can hold me until I crack through to magma. 

I think my problem this most recent fortress was that I had already denuded the outside world of trees, and so I thought that I needed to send a fairly aggressive lumberjacking force into the caverns for wood.  (My rationale being that wood would allow smelting, which would allow armor, which would make me mostly safe vs. caverns.)  The conservative play would have been to sprint for the magma and come back for the wood later.  Lesson learned!

I lost my most recent fortress to a Forgotten Beast made of glass.  Steel axes appeared to fracture-but-not-destroy his body parts, and steel hammers bounced off harmlessly.  He made a mockery of my iron-armored dwarves -- punching them through the head, ripping off limbs, and basically decimating the fortress single-handedly.  Given his ability to shrug off steel weapons, I can't imagine my dwarves would have lasted much longer in steel than they did in iron.  I know mineral FBs are fearsome for exactly this reason... but I had always gotten the easy ones (e.g. salt) so I never had to prepare for them before.  What is the preferred approach for dealing with them?  Do knockout dust traps still work or do I need to do actual crushing damage?  Can an atom-smasher bridge do the trick?

I believe it's also the significant sugar draw of the brain.

I know, at least, that when I do math 7+ hours a day I have to eat 5 meals just to get enough sugar to keep my brain still working.  I suspect he has the same issue, because in my experience it's kind of a universal problem among people solving complex problems.

Stimulants -- caffeine, cigarettes, and worse -- and poor (i.e. high-sugar, low-nutrient) food are ubiquitous in those fields... so it's no surprise that Toady has a bit of that in his life.

The game is fantastic.  I won't throw stones.

New question - Will walling off caverns from my fortress speed up FPS? Or making a door and forbidding it, and then setting the entire cavern to 'Restricted' traffic zone.

Short answer: YES. 

Medium answer: Don't bother designating large swaths of restricted pathing; it doesn't buy you much.  Designating high-traffic paths through large open spaces saves you far more FPS.

Long answer: A* (basically!) works by extending a large number of invisible tentacles out from the dwarf, like a root system seeking water.  Each invisible tentacle represents one possible path the dwarf could take while trying to get to the object in question.  The algorithm maintains a set of tentacles of equivalent cost-length, and increments the tentacles' cost-lengths at the same time.  When a tentacle hits a restricted area, that tentacle incurs the entire cost (say 25) at once, and stops growing until the other tentacles have all incurred 25 steps of cost as well.

So if your dwarf is standing in my hypothetical 15-tile-cube fortress, and all the travel costs are either normal or high-traffic, then the very worst case travel distance is 35 and the worst-case cost is probably 60 or 85.  If you have a single-tile shaft leading to your exploratory mining areas / caverns, you can place four consecutive Restricted tiles (cost = 25 each) and make any trip to the caverns more expensive than the longest possible trip within the fortress.  There are upsides and downsides to this.  The upside is that a dwarf inside the cube will never have to search outside the cube for anything he can get within the cube; conversely, any dwarf in a cavern smaller than a 2x2 embark (!) will almost certainly search the entire cavern area before he tries to search the cube.  Again, this is accomplished by locking down four tiles to raise their cost to 100.

If you wall off the caverns, no tentacle can ever reach them... but if you're taking my advice from above, no tentacle should ever need to search those areas unless there is something that is uniquely available in the caverns.

If your dwarves do need to scour the caverns you can use burrows, base camps, and high-traffic paths with restricted gutters (basically three parallel paths of RES/HT/RES) which will shuttle dwarves to the most useful areas like mineral veins, water sources, or flat areas with lots of silk and cavern life.  Using traffic zones to build "rails" in this way can save a bunch of pathfinding costs when you want a dwarf to take a predefined path through a large open area.  This is because the tentacles that try to go off the path are severely penalized, and so the more productive tentacles can look 25x further without the CPU bothering to expand the nonproductive ones. 

(EDIT: cleared up some vague language)