Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Nil Eyeglazed]

That design won't quite work-- pressure plates send opens and closes, and you've got to build some logic to make them work like you want, logic that will slow down activation time (to the point that the system won't keep up with the dwarves).  But something like this, maybe:

Code: [Select]


 ######
 h^#^h^
 ##h###
  #s#
  #b#
  #o#
  #e#
  ###


Each pressure plate is linked to the hatch behind it (to the left of it, in this diagram), guaranteeing only one-way passage.  Dwarves won't be able to enter from the right, nor leave to the left.  The central hatch is the dropchute.  (Dwarves won't get scared too early-- confident about how LOS works-- but if they drop too late, that'd be a problem, as it'd block off the lure tile.)  Really slow dwarves might occasionally dance a bit, but that'll straighten itself out.

What would be ideal would be to have some kind of bolt waster as an enemy, something like a bronze colossus.  Of course, those are hard to get your hands on.

You can see that I've also put a raising bridge in there, so that if we do get a dwarf that somehow drops on the lure tile, we can block off sight to clear it out.

EDIT: I've got some tests (in 34.02) going on now.

The hatch+bridge undump is working fine for furniture.  Plating a lot of goods, but nothing making it through to lure.

The scaredorf isn't working at all.  The dwarves can see the bad guy, but they go one tile closer to him anyways before cancelling the task.  I'll probably try to build it one hatch longer, see if that works.

EDIT: Stuff makes it through to lure with hatch+bridge undump.  It's less frequent, but it happens.  I've designed a version that so far is causing no problems, but it's more mechanics than most people want to mess with.  It uses a hydromechanical latch to keep the hatch open until the dwarf leaves the undump.  Design is at http://dwarffortresswiki.org/index.php/User:Vasiln/Undump#Latching_Dropping_Undump.

[Kaos]

That design won't quite work-- pressure plates send opens and closes, and you've got to build some logic to make them work like you want, logic that will slow down activation time (to the point that the system won't keep up with the dwarves).  But something like this, maybe:

Code: [Select]


 ######
 h^#^h^
 ##h###
  #s#
  #b#
  #o#
  #e#
  ###


Each pressure plate is linked to the hatch behind it (to the left of it, in this diagram), guaranteeing only one-way passage.  Dwarves won't be able to enter from the right, nor leave to the left.  The central hatch is the dropchute.  (Dwarves won't get scared too early-- confident about how LOS works-- but if they drop too late, that'd be a problem, as it'd block off the lure tile.)  Really slow dwarves might occasionally dance a bit, but that'll straighten itself out.

What would be ideal would be to have some kind of bolt waster as an enemy, something like a bronze colossus.  Of course, those are hard to get your hands on.

You can see that I've also put a raising bridge in there, so that if we do get a dwarf that somehow drops on the lure tile, we can block off sight to clear it out.

EDIT: I've got some tests (in 34.02) going on now.

The hatch+bridge undump is working fine for furniture.  Plating a lot of goods, but nothing making it through to lure.

The scaredorf isn't working at all.  The dwarves can see the bad guy, but they go one tile closer to him anyways before cancelling the task.  I'll probably try to build it one hatch longer, see if that works.

also try placing the scaredorf above the target drop tile using floor bars or floor grate, the new versions are supposed to allow LOS through these.

[micha]

Hey, thank you for turning an accident into something productive viz science!

i started a new fort last night where i spent some hours experimenting with the design, placing 11 cells north and 11 cells south of 11 up down stairs spaced on apart.   i made the cells have a door in the end so i could disable them individually, e.g. they are  PpPHdP  pile/pad/pile/hatch/door/pile.  all the cells converge into a simple 2 wide hallway in which i used doors to set up paths and experiment with the performance.

to make a long story short, i found specialized piles unsuitable for general use and wound up turning my aparatus in one 22 cylinder undump-everything-stockpile. this works. *

i completely floored by how well the fps in this fort has been holding up too. i've dug and hooked up a river filled cistern entry flusher and other atrocities and the fps is where it was at the completion of me having set up the stockpiles. now there are only 22 empty stock pile cells looking for work .. perhaps this is what's behind the radical [for me] speedboost.

does this tie in with forum experience?


i've not done much butchering beside the 10 dogs and 2 horses i had when arriving in 4 butcher shops set up for a test. due to the bandwidth of the 22 cylinder engine, i got no miasma, but it wasnt a fair test cause anything near those 11 staircases is practically sitting in the pile. 

anyway, this thing already looks like a train-station and if it could slaugher and disappear 100 war elephants in reasonable time you might have given us the holy grail of fpsdom here. after the last of hamlet wallshadow's illustrious denizens finish slaughtering each other due to my complete neglect of them, i'll sick some crafters at the machine to see how it works as an everything dump.

[micha]

this is a pic of my 22 cylinder undump everything prototype.  my fps has remained at where it was when i setup the stockpiles.  that is if there is no work in the system.  if i have more open slots than there idlers available fps drops from a normal '50(45)' all the way down to '18(12)' until i close some of the cylinders. 

with more idlers than open slots this seems to work well. you can see in the img that i experimented with multiple open slots at the top as well as feeder piles. neither prooved helpful. now there's room for tracks and one of those new and coming push-piles; realistically though one could build this without the extra 3 spaces for lower lag.

in the pic too you see that there are 4 overruns blocking a bit of capacity. these happen. and they do not stop the everything dump, depending on how many idlers you have, they may even be desirable if inefficient. with an evolution of the pile-tops they could be freed by a the lever-pulling shut-in.

this is easy to build and macro. in case you've not seen this before, an undump cell consists a stockpile having 3 open spaces in the space of 6. build the channel, hatch, door and pressure plate, then overlay the stock-pile. stuff stacks after the plate, right before tha hatch its connected to. here there are 22 identical 'everything' custom piles.  the doors at the extremes segregate the 'should always be empty' stock-pile ends so that there is only one path to get to them - which the doors on the pile control.

lets hope this keeps on working! enjoy.

edit: this is after adding some 4000 items.  the fort survived with 6 battle hardened vets and there's 16 folks now, 10 idle, 5 getting clay, the fps sometimes rapidly dropping to '50(34)' but always creeping back to the '50(45)' where i started.  can someone else please give this design look-over?

edit: scratch that idea with the lever-puller touching the piles. just realized i can put all 22 open spaces on bridges to have them droped from outside, requiring even less space.

[Veylon]

I've given this piston method a try and, for the most part, it works just fine. The one niggling problem I have is that dwarves will often, instead of dropping the object, try to bring it to another stockpile. And another, and another, and another. This happens regardless if each cell has it's own pile or if all of them are tiny fragments of one big one.

[micha]

Hi Veylon, i saw them try to path into multiple pistons until i locked the doors between the pistons.   for curiosity, the top right 3 cells share a stockpile and they appear to work just the same as having them all be different piles; seems. i'm going to make all 22 use one pile (with 22 isolated single tile lures) and report back if that remains true.


meanwhile here's an update from the guys:

as you can see they are all busy right now so i'll do the talking. the hauling actually comes in bursty waves. there'll be this activity and then a moment of pause .. which makes me think that if not for the shorter paths, this could be slightly slower.



but lets back up. i had dead bodies galore and brute forced a test by setting up butcher shops and refuse dump-zones, then autodumping 600 some items from the surface, including maybe 30? butcherables from the last siege


i also dumped 300 wood just to confuse them them all. disposing the butcherables netted 512 tallow. there were 5 'overruns' and 4 times that i saw an item get dropped.

It does seem like Hamlet Wallshadow is not going to die from item or pile proliferation. But I'm an optimist. Apologies for that. If you have real world experience with this .. please chime in.. since..

i'm going to take what i learned about the design an replicate it 3 lvls down with DOUBLE the spaces between the pplate and the hatch. having two spaces could maybe drop the overruns from ~1% to ~0.1%, but we'll see.

as is, i would cautiously endorse the undump engine for use in a real fort.

(made pngs this time and now i cant see the images. aaargh. here are the links:
http://imgur.com/8uSg5
http://imgur.com/k8fF9
http://imgur.com/GFJS5
http://imgur.com/XSamb)

[micha]

http://imgur.com/1u6aX

ok, here is a quick update on the undump engine, 2nd revision. as of the time of picture each cell had about 20 pages worth of goods. its up to about 30. there have been no broken legs, noone's fallen or gotten any items placed in the lure square. 

the addtl square really works and the presence of the second engine, some 3000 more items and 6 new immigrants has not hurt the fps 'as much as it would have had i continued making piles and spaces for piles'.  it took a bit longer with siegers and remains but it got back to '50(45)' before i gave the command to move everything but the food to the new pile. there's a lot more to be moved, so i've moved the dwarf idle zone between the stacks. i think i want the living area there and workshops in the other two floors.

drinks dont spoil, fortunately. i don't know why or how, but drinks don't stack that fast. that's to say they seem to get put away last and then not even quickly. haven't milled anything so i dunno if that's a 'we're storing more than the barrel-setting would allow' or a food-related thing. seeing it 'not work' on the barrels was a bit troubling, but it cleared up. no miasma at all yet.


to make: dig the room, channel for the grate, build 22 doors, 22 pplates, 22 grates. hook up the grates to the plates (citizens,all weights), make the entire room a stockpile, 'x' out the stairs. adjust stockpile settings: no bins, max barrels.

todo: if anything should get to the lure, adjust your build to put 3 spaces before the hatch instead of building a recovery mechanism right away. cause if the 2nd square wasnt the answer to the problem, the 3rd sq should be it until then, autodump.


Thank you Nil, your observation will add years to my forts!

[Veylon]

I've been having each piston with three doors (one before the plate, one between the hatch and lure pile, and one at the end), with the third one locked at all times unless I need to pull something off the lure pile, which only happens when I screw up.

I found a partial workaround in having my haulers all share a burrow that touches the actual piles, and then making a dozen burrows that each touch one lure pile and contain 1/12th of the workers. This means that each hauler can only bring objects to one of the pistons (thus preventing the "bouncing" from one to the next), but only a handful of haulers will share a piston. The downside is that this requires a ridiculous amount of micromanagement.

I'll try the new revision and see if that helps. It looks a lot cleaner, if nothing else.

[micha]

Veylon, hey thank you for writing.  i'm sitting here again staring at >6k food and >1k of each craft item, >2k bars, >2k stones, i have 30+ siegers on the map and am waiting in disbelief for the other shoe to drop because the fps still goes back to where i started.

I kept it at 18 dorfs and dug just enough to add 36 3x3 bedrooms on the level between the stacks, plus putting a hospital/well where the meeting area used to be. to compensate i made all empty spaces restricted traffic zones and zoned the staircase and engines to be high traffic zones.  i'm at more than double the items i had in my last report. i just looked, 12 of 18 idle, fps '49(44)'.

my experience want me to belief that i'm doing something wrong, that this can't be, that i will wind up with eggs on my face and my foot in my mouth, but i'm not doing it wrong!  i've debunked the idea that item count kills fps. for me at least,

WITH AN UNDUMP ENGINE ITEM COUNT DOES NOT SEEM TO AFFECT FPS AT ALL
(yes, i know how insane this sounds)

my original pile is about to get split into food and refuse. i like having the butchershops and kitchens on that level too. i had miasma once, from one item, as i had crafting and smelting and clay and building all scheduled at the same time.  that made me smile because i thought i might have broken something  

the new pile is very cool and i've removed the doors with which i blocked the exits on the level below. i had hoped to catch fallen animals or dwarves. there have been none. no accidents. and with two spaces before the hatch nothing at all winds up on the lure tile.

food/drink barrels are not the problem i thought they might be. that was me not resetting dump flags or something.

if i was building from scratch, i'd use 2 floors with 22-engines each, use one for food and one for everything else. with a few floors above/between/below into which to put related workshops right by the stairs.  one floor i'd use for tracks cause this concentration of stuff will surely make for an ideal rail station.  it is already super-fast to get stuff made because anything and everything is always within 10 or so spaces from where its needed.

so, yes, please try the zero-micromangement, one pile, 22 unreachable lures design.

note: in the r2 design i have door which became superfluos with the second tile before the hatch. you can leave them out and save the space. if i counted this right one could place the r3 undump engine staircase 12 tiles from one another and _then futz around with quality settings or other pile specializations, suffering the normal fps penalties for digging of course.

let me sum this up once more, i have never had a 7 year old fort with more than 867 listed dead, pages and pages of masterwork meal and bone craft announcements and insane amounts of stuff that would come back to show roughly the same fps it had before the killing and cooking and crafting. 

i fear my optimism is staring to border on excessive so could we please have someone cynical step up to shoot this down?

(i'll process a few more sieges now and post something to the front page later)

Thanks Nil, Thanks Veylon!


edit: p.s. an hr after the post one item made it to the r2 lure pile. i saved and autodumped all the items from the pile to get a count. there were 11637 items in the pile vs 1 mixup. i dunno that this justifies a 3rd space or a door or an access mechanism.