Print

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

[FantasticDorf]

Until the wind and 'power' is streamlined to be less of a magic force (who knows maybe in the future it will be 'magic') a fusion reactor is a well touched upon subject but has no means or basis in DF reality to come into being. What realistically are you going to power on 14th century technology that doesn't stretch further than a windmill?

Question

When the game is more optimised in the future, will we ever see any development on 'the wind' in regular fortress play/greater map extent, since it is mostly defined by static air tiles at the moment and has no basis on the game. Its either in the air tiles or it is not for windmills, thermodynamics would be greatly aided by more hot/cold air gas floating around and weather patterns akin to 'evil rain' would have reason to exist and travel instead of being static to a biome to where even side by side tiles of a different temperature and biome can rain and the other will snow heavily.

In that kind of extent, i guess nether cap wood would be constantly condense ridden in hot humid tempratures from the humid air interacting with the constantly cool surface like a dwarven damp catcher, as well as dusty abandoned fortresses and tombs spluttering when opened. Mmm, theoretically with some more world meddling you could force a ecological disaster (a hot plume of dwarven industries melts away the glaciers flooding sea level tiles/magical event or a world event of a global ice age where temperatures via a super high amount of cold air circulating in override of natural patterns plummet to tundra conditions, destroying agriculture leading to famine and war)

Hot and cold air currents meeting between biomes sounds like a fun recipie for a thunderstorm/extreme weather and a bad time to be wearing metal on high ground and have stubby little legs that can't run that fast when you think walking across the mountains is a cheap shortcut.

[Random_Dragon]

Same reason you'd build a waterfall-based perpetual motion engine. You'd use it to power screw pumps, of course. XP

[PatrikLundell]

So, if you combined a dragon with a magma forge AND a compost heap, would we get a dwarven fusion reactor? o3o

No, you'd at least have to add a cave-in to compress matter sufficiently to induce fusion
(And it would still be orders of magnitudes too cold..)

[Random_Dragon]

Cave it into magma for good measure, because you can ensure it'll be warm there.

[Whatsifsowhatsit]

Question copied over from another thread.

Is there any chance that some of the abstraction that has to be in due to memory/processor constraints and whatnot will be taken out in the future due to improving technology? An example would be how currently, only some fraction of the creatures in the world can be a historical figure that is tracked in any level of detail, because it would be too much to do this for all of them. Similarly, not all items are tracked individually, etc.

I understand that for most purposes, this doesn't matter too much, and it's unlikely that a player actually notices the difference, but I'm the geeky sort of guy who would get pleasure just knowing things are simulated faithfully. (Small wonder Dwarf Fortress already is my favorite game as is.)

[Bumber]

Question copied over from another thread.

Is there any chance that some of the abstraction that has to be in due to memory/processor constraints and whatnot will be taken out in the future due to improving technology? An example would be how currently, only some fraction of the creatures in the world can be a historical figure that is tracked in any level of detail, because it would be too much to do this for all of them. Similarly, not all items are tracked individually, etc.

I understand that for most purposes, this doesn't matter too much, and it's unlikely that a player actually notices the difference, but I'm the geeky sort of guy who would get pleasure just knowing things are simulated faithfully. (Small wonder Dwarf Fortress already is my favorite game as is.)

I very much doubt it if we don't get multithreading. Improving technology has been focusing on parallelism, DF has not.

[Shonai_Dweller]

Question copied over from another thread.

Is there any chance that some of the abstraction that has to be in due to memory/processor constraints and whatnot will be taken out in the future due to improving technology? An example would be how currently, only some fraction of the creatures in the world can be a historical figure that is tracked in any level of detail, because it would be too much to do this for all of them. Similarly, not all items are tracked individually, etc.

I understand that for most purposes, this doesn't matter too much, and it's unlikely that a player actually notices the difference, but I'm the geeky sort of guy who would get pleasure just knowing things are simulated faithfully. (Small wonder Dwarf Fortress already is my favorite game as is.)

I very much doubt it if we don't get multithreading. Improving technology has been focusing on parallelism, DF has not.

And Toady has been looking at multithreading. So...what's the problem?

[King Mir]

Question copied over from another thread.

Is there any chance that some of the abstraction that has to be in due to memory/processor constraints and whatnot will be taken out in the future due to improving technology? An example would be how currently, only some fraction of the creatures in the world can be a historical figure that is tracked in any level of detail, because it would be too much to do this for all of them. Similarly, not all items are tracked individually, etc.

I understand that for most purposes, this doesn't matter too much, and it's unlikely that a player actually notices the difference, but I'm the geeky sort of guy who would get pleasure just knowing things are simulated faithfully. (Small wonder Dwarf Fortress already is my favorite game as is.)

I very much doubt it if we don't get multithreading. Improving technology has been focusing on parallelism, DF has not.

And Toady has been looking at multithreading. So...what's the problem?

DF is the antithesis of a parallel application. Everything effects everything. Every subsystem effects every other. The mechanism system effects the fluid system, which effects the temperature system, which effects the damage system, which effects the psychology system. And all of them have to sync every tick. There's no room for big gains from parallelism even if Toady were an expert in it, instead of just starting to dabble. This isn't even a design problem; it's a project objective problem: everything is supposed to interact with everything to be a complete simulation.

[Fleeting Frames]

There already are some things that act/register with 1-step lag sometimes or always, such as timing dependent on build order or minecarts on pressure plates, and I don't think we'd even notice slight lag in something as random as water motion outside of edge cases.

Making DF parallel would be a crazy difficult rewrite of the game, but I don't think because of this.

[PatrikLundell]

I agree with Fleeting Frames, although King Mir is correct in that DF is very far from s a perfect candidate for parallelization. Since everything has to synchronize at the end of a tick, you can run things in parallel DURING the tick and consolidate the results at synchronization points during the tick. Different methods can be used for different sub problems.
One example is pathing, where you can have separate threads calculate paths for creatures. At a synchronization point you then validate the paths (to ensure they paths calculated are still valid) and recalculate the ones that are not. Validating a path is a reasonably cheap operation compared to calculating the path in the first place, so there's a potential for a gain here, but there's overhead in organizing the farming out of the path calculation, in addition to the rather substantial effort of ensuring the data used for path calculations is stored in a thread safe manner that doesn't slow down things more than you gain.

I'd worry more about the huge volume of work and how to reorganize things such that bugs introduced by unanticipated consequences of parallelism can be identified than the very substantial work of the parallelization itself. In particular I'd worry about the several years of (part time) preparation work where data stores are secured for parallel access and DF gets SLOWER as a consequence.

Parallelism is by no means a free lunch. There's a lot of organizational overhead introduced, so I wouldn't be surprised if the total number of CPU cycles of DF parallelized would be 2 or 3 times the single threaded version. The speed gain would come from spreading the X times as many cycles over (many) more than X cores. All of this is dependent on the CPU being the bottleneck and not memory bandwidth, though. If memory is the bottleneck, the additional administration of threading might actually result in a loss.
Thus, before threading is introduced somewhere, you should try to determine that it will actually result in a net speed gain (of course, one way to determine it is to hack it in crudely, see if it seems to work, and roll back if it doesn't, and do it properly if it does, possibly rolling back in that case as well, if the hack is ugly enough).

[Untrustedlife]

Hey tarn/forums, been awhile since i was on the forum.

What happens when your adventurer dies post worldgen via world-activation, for example your adventurer dies by say dying in an invasion/insurrection or dying from starvation or dying from old age due to world activation while you were playing another adventurer or playing a fort. Does the adventurer get entombed in the crypts underneath the city or do they just cease to exist or does it just drop their corpse somewhere in the city?

[Whatsifsowhatsit]

Thanks for your insights, everyone, and PatrikLundell in particular. I don't have a lot of understanding about these matters, myself, but I think I could follow along for the most part. It sounds like, even if it does ever happen, it would have to be quite a timesink, and to be honest even I don't want that to happen if it keeps development on other areas of the game back. So perhaps it could be something for after version 1.0 is out or something? I dunno. I'd still be interested to hear Toady's thoughts on it. I think multithreading was addressed in a question to him before, but I'm specifically interested to know if that (or other elements of progressing technology) could be used to make the simulation more complete and faithful, with less abstraction, like I said in the question.

[Migrant]

When during play does world activation happen? Does it happen all the time during play or does it catch up at the end of every year?

[Untrustedlife]

When during play does world activation happen? Does it happen all the time during play or does it catch up at the end of every year?

It happens as you play. Otherwise no armies would ever visit your fortress, and no one would come to your tavern, and it would cause weird issues with adventure mode.

http://dwarffortresswiki.org/index.php/DF2014:World_activities

(First section)

[PatrikLundell]

@Whatsifsowhatsit:
1.0 is a decade or two away, and I think DF will have slowed down too much by that time unless parallelism is used to some extent, in particular since most of the computer speed increases lately comes from increasing the number of cores, although a major memory speed/bandwidth breakthrough might help a lot if one would happen.
Farming out work to side threads offloads CPU time for the primary thread, and if the side thread work is CPU bound you might afford an additional level of simulation within the tick time budget, so if the simulation is more number crunching on the same set of information it could certainly be done to some extent, but often it's also a matter of increasing the amount of data crunched, which brings us back to the memory bottleneck. Also, you can't use parallelism to such a level that DF won't work at an acceptable speed on a somewhat outdated (laptop) computer with a limited number of cores when running a reasonable scenario.
Finally, developing a more detailed simulation of some details will be time spent not implementing some other feature... It's always a decision of where to spend the time and effort now, and whatever decision Toady makes most other people would have wanted the time spent differently (namely on their pet features/bugs).