Bay 12 Games Forum

Please login or register.

Login with username, password and session length
Advanced search  
Pages: 1 [2] 3 4 ... 21

Author Topic: Will we ever get to a point where forts don't die FPS deaths?  (Read 61694 times)

Ailure

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #15 on: June 17, 2010, 04:07:39 am »

Seeing how computing is more focused on multi-threading, it's probably a necessity at some point for DF. The problem is finding parts of the game logic that can be parallelized, and I'm quite sure there is a quite few things that can be.

Of course, optimizations helps as well and might be easier to start with for Toady.
Logged

Leperous

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #16 on: June 17, 2010, 04:20:47 am »

I think it depends on whether or not Toady wants to go for a simpler Dwarf simulation - for example, overtly complicated temperature calculations are clearly a huge drain on CPU, turn it off on your 100 dwarf fortress and see - and past updates don't point to this being the direction he wants to go. Certainly end-game needs urgent fixing for this very reason.
Logged

Maggarg - Eater of chicke

  • Bay Watcher
  • His Maleficent Magnificence of Nur
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #17 on: June 17, 2010, 04:30:38 am »

Take a few fps-saving measures in the init file or by making traffic zones.
Logged
...I keep searching for my family's raw files, for modding them.

Qiu

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #18 on: June 17, 2010, 04:57:38 am »

Will having a really powerful computer help?
Well, when your fort is at 5 FPS, having a computer 4 times as powerful will still be really slow. :(

And the time where single core power was doubling every few years seem long behind us, so as long as DF doesn't become completely multicore (more than just having a few % on another core) i think more powerful computers aren't going to run DF much faster.
Logged

fanatic

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #19 on: June 17, 2010, 05:06:27 am »

I was thinking just now.

I NEVER saw someone  on the forum say something boiling down to "Hey, screw the perfs i'd really rather have feature X in ASAP" . Maybe forum veterans can prove me wrong, but what i see more often is like "try turning off temp" "try taking a smaller embark zone" "you have to adapt your fort to your hardware" " we're only in alpha".

Given this, does that strike anyone as weird that we have a ton of new features coming in with the new version an only mild optimization?

Oh well, nothing new i guess...
Logged
fanatic cancels play DF : gone berzerk at framerate.                                                  x1000
------------------------
Pour magma first - ask questions later!

Urist McDepravity

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #20 on: June 17, 2010, 05:35:26 am »

Quite simply, I know people looked into understanding how DF manages Dwarf AI and pathfinding -which remains the single biggest FPS killer. The more dwarves, creatures, paths, and items for them to interact with = worse fps.
Well, yes, i guess its pathing too, but it should not work this way. Even worst naive tile-by-tile search does not care about map size if path is quite direct.
Esp. when most of movement happens in repeated paths over small part of map.
And item count would matter only in one case (assuming temp off) - if dorf checks distance to each and every item of needed type. Altho didn't Toady say that item distances are just linear, w/out actual path-finding? So even if you have 10000 stone, it should not take long to find 'nearest'.
Quote
I NEVER saw someone  on the forum say something boiling down to "Hey, screw the perfs i'd really rather have feature X in ASAP" .
You are wrong. Check ESV, there are alot of features above pathfinding in top. So yes, people care about FPS, but would prefer some automation over it.
Logged

Jimmy

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #21 on: June 17, 2010, 06:04:47 am »

Hot tip: I see many players using stacked up/down staircases. FYI, these kill your FPS. Ramps are extremely FPS friendly on the other hand. It's why the outside map, even with steep slopes and tons of ramps, doesn't cause lag. Design your fortresses with ramps instead of stairs and traffic zones and you'll see your FPS drastically improve.

Item count is another big thing, and your veterans will have their own methods of dealing with it. Atom smashing for stone is a popular method, or crafting for trade value. Remember that traders don't just bring items onto your map, they take them off too.
Logged

Grimlocke

  • Bay Watcher
  • *kobold noises*
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #22 on: June 17, 2010, 07:23:04 am »

Figures stairs would be a drain, but never realy checked it. I guess ill put my spiral ramp staircase to use for next fortresses.

And lets hope that once Toady gets around to writing a more efficient pathing code he throws in some multithreading as well.
Logged
I make Grimlocke's History & Realism Mods. Its got poleaxes, sturdy joints and bloomeries. Now compatible with DF Revised!

Jimmy

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #23 on: June 17, 2010, 07:52:21 am »

If I may be forgiven a personal plug, feel free to explore my Canyonlance map for an example of aesthetically pleasing and efficient ramp designs (link). With 7 years of history, 50 dwarves and close to 20,000 items I still get 100FPS.
Logged

huhu

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #24 on: June 17, 2010, 08:39:00 am »

And item count would matter only in one case (assuming temp off) - if dorf checks distance to each and every item of needed type.

One solution I can think requires a little managing by players for a lot of FPS gain. Simplest form of it is that the only dwarves checking every single item around the fort are those doing hauling jobs that end in a stockpile. Everyone else can only search stockpiles for items, which can be optimized in quite a few ways. This would probably be initialized once a manager is appointed from the nobles menu.

If an item needed for a job is not in any stockpile, then they abort that job with an error message.

Edit: It probably requires some method for dwarves to carry at least 10 stones in a single trip.
« Last Edit: June 17, 2010, 08:44:23 am by huhu »
Logged

Rowanas

  • Bay Watcher
  • I must be going senile.
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #25 on: June 17, 2010, 09:05:33 am »

2.2 laptop
Two cores, but unimportant
40 FPS

Seven dwarves? 40.
200 dwarves? Still 40
What is going on?

FPS cap set
100 FPS max
but only 40.
Logged
I agree with Urist. Steampunk is like Darth Vader winning Holland's Next Top Model. It would be awesome but not something I'd like in this game.
Unfortunately dying involves the amputation of the entire body from the dwarf.

Djohaal

  • Bay Watcher
  • [PREFSTRING:Utter Insanitiy]
    • View Profile
    • My deviantart
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #26 on: June 17, 2010, 09:09:48 am »

Yes, sadly the lack of multithreading/paralelism in DF is a big problem if we consider it is the prevalent trend in modern computing, and it would require a massive code overhaul... but it is toady who we are talking about, so who knows  :P

Temperature simulation seems to be a major performance hog too, and I think it could be optmized quite a lot considering how most of the time it is just a simple temperature transmission.
Logged
I really want that one as a "when". I want "grubs", and "virgin woman" to turn into a dragon. and monkey children to suddenly sprout wings. And I want the Dwarven Mutant Academy to only gain their powers upon reaching puberty. I also have a whole host of odd creatures that only make sense if I divide them into children and adults.

Also, tadpoles.

Motigg

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #27 on: June 17, 2010, 09:27:21 am »

Yes, sadly the lack of multithreading/paralelism in DF is a big problem if we consider it is the prevalent trend in modern computing, and it would require a massive code overhaul... but it is toady who we are talking about, so who knows  :P

Temperature simulation seems to be a major performance hog too, and I think it could be optmized quite a lot considering how most of the time it is just a simple temperature transmission.
The temp calcs are pretty big.  I tested it on a 120 dwarf fort that I was getting 30 FPS, turned it off, jumped up to 50.  Theres something wrong with that.
Logged

Rowanas

  • Bay Watcher
  • I must be going senile.
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #28 on: June 17, 2010, 09:30:25 am »

Bug fixing, then speed.
People whine if no features,
but whine if not quick.
Logged
I agree with Urist. Steampunk is like Darth Vader winning Holland's Next Top Model. It would be awesome but not something I'd like in this game.
Unfortunately dying involves the amputation of the entire body from the dwarf.

Dr. Hieronymous Alloy

  • Bay Watcher
    • View Profile
Re: Will we ever get to a point where forts don't die FPS deaths?
« Reply #29 on: June 17, 2010, 10:09:39 am »

Hot tip: I see many players using stacked up/down staircases. FYI, these kill your FPS. Ramps are extremely FPS friendly on the other hand. It's why the outside map, even with steep slopes and tons of ramps, doesn't cause lag. Design your fortresses with ramps instead of stairs and traffic zones and you'll see your FPS drastically improve.

Item count is another big thing, and your veterans will have their own methods of dealing with it. Atom smashing for stone is a popular method, or crafting for trade value. Remember that traders don't just bring items onto your map, they take them off too.

Yeah, massive item count is a bigger killer of FPS than pathing in old forts.

Pathing is a big deal, though. As you say, you can basically design your fort for one of two kinds of efficiency -- dwarf-time efficiency, with lots of up and down stairs everywhere, or CPU efficiency, with (ideally) only one given path from any location to any other location. For CPU efficiency, a central spiral ramp is going to be more effective than stairs everywhere.
Logged
Pages: 1 [2] 3 4 ... 21