Bay 12 Games Forum

Please login or register.

Login with username, password and session length
Advanced search  
Pages: 1 ... 4 5 [6] 7 8

Author Topic: Curb treesplosion by gradually reducing tree maturation as tree density increase  (Read 14721 times)

Bumber

  • Bay Watcher
  • REMOVE KOBOLD
    • View Profile

That simply is not true.  Modelling even one day's worth of tree growth on a large forested biomes taxes the RAM (or is it processing power?) of most modern computers, as I know from bitter experience.  What you are proposing is that during embark the RAM cope with the whole growth of a forest over years in a single moment!  There is a way around it however, the way round is to take 'time out' of the embark (or when loading a new area of the adventurer map) and dedicate the whole memory to filling the loaded area with trees that did not need to exist before.  At the end of tree-gen the trees are simply saved into the save-game file and can then be loaded into the game since static trees take far less memory to load than the process needed to grow them does.
That's very doubtfully due to just simulating tree growth. There's all sorts of background history stuff going on, pathing, tree litter tracking, and growing branches off the existing ones. The problem is that trees are not static after they've been loaded into the game, but must continue to grow. (Also, as PatrikLundell stated, that's CPU, not memory.)

I think you'd be hard-pressed to argue that a set of Cartesian coordinates, a bounding box (length & width,) and a value for age take as much memory/CPU as these same things plus a physical layout. It is more optimal to do everything up front with the former (and then turn them into physical forms,) than it is to have the physical trees and model every physical stage of their growth while also making sure the player didn't go and build something there.

Quote
They have clearly rejected the tree-farm model, which even if randomised can still only manage a very limited array of shapes compared to the dynamic and competitive growth model which makes very interesting trees indeed.
No, it's actually not clear at all. While I'm not sure if one would actually notice the difference, the trees would start to grow through each other post-embark anyway. (I can only assume this is what you think makes them interesting. Everything else boils down to placement, which you wouldn't be able to discern from a proper pre-embark algorithm.)

Quote
None of those are DF pictures so I have no idea what the final situation in those biomes looks like.
Me neither. Unfortunately I don't have a whole lot of spare time to find out at the moment. Word of mouth says low tree density isn't enforced. Edit: Vattic posted pics below.

Quote
I do not see much of a cluster effect at all, there a small amount of clustering only because the dots are small and the number of them is small.  If the dots were large and they were free to multiply across the board, placing new dots at random locations in the white area it would not take very long until we end up not with a distribution of dots but a black sheet of total uniformity.
Try imagining the dots as tree trunks, not leaves. Compare with your ground-level picture.

Quote
That isn't sufficient proof if it's randomized. Suppose you have a twenty-sided die. How many rolls does it take to land on every number at least once?
400 rolls, is this supposed to be a mathematics puzzle?
Statistics trick question used as an example. There's no guarantee you won't roll 400 ones, however unlikely.

The metaphor extends as thus: Each embark tile is 48x48 = 2304 game tiles. That's 9216 tiles on the default 4x4 embark. If the game attempts to place saplings randomly without respect to existing trees (my assertion based on Occam's Razor), then for each sapling placed there is a 1/9216 chance per viable spot to place it such that it will grow. As the trees fill the map, it gets less likely to find a remaining good spot. I assert that this is why we see these 'clusters'. While I have not sufficiently proven this assertion, neither have you disproven it. It exists as an alternative explanation.

Quote
If the distribution of something is indeed truly random then as the number of the item placed increases, the clustering in the placement of the items *must* decrease or else the situation is not truly random.  That there are, in an sapling-less area after 7 years of uninterrupted sapling placement then something is clearly going on.
Eventually, yes, but there are no guarantees as to when that will happen. It depends on how frequently placement occurs for those 7 years. We lack the information for statistical analysis.

Quote
I have already identified such a zone Bumber, there is clearly one to the west of the lake in the south-east of the map; I could draw them if I wished by drawing around all the saplings and tree trunks on the map.  At the end of the day it does not matter, because I do not have to prove a negative, you have to prove that treesplosion exists and not I the reverse, especially given how extraordinary a claim it is to have the devs make such a detailed tree-growth model without putting in place any checks on tree growth even in arid treeless areas.

If your argument is simply that 7 years is not long enough for random increase to result in uniformity then come up with a 20-30 year map that shows the resulting uniform distribution of trees with no clusters in site.  Mere hearsay that such a thing as treesplosion exists is not sufficient.
What I meant was do these zones persist? If you clear the map of trees and vegetation, will no trees grow in that zone? If it happens once, your assertion is disproven. It's easy enough to test, and I would expect at least one of your zones to get a tree within the first year or so without the interference of existing foliage. It's not practical for me to test this because I don't have your map, nor a complete understanding as to what you consider a zone.

I have to prove treesplosion occurs on arid embarks (which I might do during winter break, if someone else doesn't Edit: Vattic did below.) You have to defend the existence of clusters, as well as your "clearly working as intended" assertion.

It's definitely false that new saplings do not appear right beside existing trees. I've manually (using dirt roads) killed off all unwanted vegetation in an orchard on a different embark from the tree farm above. Saplings have appeared right next to existing trees.
The question was whether or not they could grow anywhere away from other trees, with regards to the statement:
Quote from: GoblinCookie
What random chance? The trees are supposed to exponentially spread across the map, without any checks or limitations...
(Emphasis and misunderstanding mine.)

It was clarified that he meant that there are areas where saplings simply won't grow. Spread quickly, not spread starting from other trees.
« Last Edit: November 25, 2015, 05:51:56 pm by Bumber »
Logged
Reading his name would trigger it. Thinking of him would trigger it. No other circumstances would trigger it- it was strictly related to the concept of Bill Clinton entering the conscious mind.

THE xTROLL FUR SOCKx RUSE WAS A........... DISTACTION        the carp HAVE the wagon

A wizard has turned you into a wagon. This was inevitable (Y/y)?

PatrikLundell

  • Bay Watcher
    • View Profile

My statement about saplings growing next to existing trees was in response to this:

Quote
I am talking about how the growth of trees are regulated within a tree-growth zone, where saplings are free to be placed randomly.  Within the tree growth zone the regulatory factor is the proximity of adult trees, new saplings do not get placed (and existing saplings get snuffed out) if they are too close to a grown up tree, the bigger the tree the further they have to be away from it. 

The bold marking is my doing.

I should be better at using quotes...
Logged

Vattic

  • Bay Watcher
  • bibo ergo sum
    • View Profile

The following thread starts with a screenshot of a savannah biome after a few years. It is suggested that felling existing trees might be making space for all the embark generated saplings to grow. It also has a number of other people reporting the same occurrence.

Edit: Decided to let a fort run with the fps uncapped to get some quick comparison shots. The embark is a "sparse" savannah.

Spoiler: year 5 (embark) (click to show/hide)

Spoiler: year 8 (summer) (click to show/hide)

Spoiler: simplified comparison (click to show/hide)

Sorry about the poor image quality. I neglected that I had a high resolution tileset installed and the final images would have been massive without a resize.

I think they give a fair idea what people are talking about.

« Last Edit: November 23, 2015, 11:12:55 pm by Vattic »
Logged
6 out of 7 dwarves aren't Happy.
How To Generate Small Islands

Bumber

  • Bay Watcher
  • REMOVE KOBOLD
    • View Profile

Edit: Decided to let a fort run with the fps uncapped to get some quick comparison shots. The embark is a "sparse" savannah.
Thanks. This illustrates the issue perfectly. It doesn't even seem to be respecting minimum distance. Shade-less trees?

What's the vegetation supposed to be on that embark? Is it equally a problem?
« Last Edit: November 24, 2015, 05:07:47 pm by Bumber »
Logged
Reading his name would trigger it. Thinking of him would trigger it. No other circumstances would trigger it- it was strictly related to the concept of Bill Clinton entering the conscious mind.

THE xTROLL FUR SOCKx RUSE WAS A........... DISTACTION        the carp HAVE the wagon

A wizard has turned you into a wagon. This was inevitable (Y/y)?

Button

  • Bay Watcher
  • Plants Specialist
    • View Profile

I have definitely observed and been annoyed by treesplosions in my sparsely-wooded embarks. The most absurd for me was when I had a marsh embark, and then retired after 10 years and brought an adventurer in to visit. Tromp tromp tromp through the mostly-open marsh dear Armok that's a wall of trees.

As for how we know this behavior isn't intended - well, the difference between a marsh and a swamp is that a swamp is dominated by trees, and a marsh is mostly open. That's literally the only difference. A marsh covered in trees isn't a marsh anymore, it's a swamp.

That's how we know that the forested marsh isn't intended, GoblinCookie. A wetland is a "marsh" only if it has few to no trees. If trees were intended on that embark, it would have been called a "swamp."

Now, maybe swamps don't have enough trees. I don't know. But marshes, after 10 years of play, definitely have too many.

Now ideally, in a fully-dynamic world, it should be possible for biomes to change. For a marsh to grow up into a swamp, and for swamps to burn down or have extreme flood events, turning them into marshes. That would be great, but it's not here yet.

I second Bumber's quick fix:

Easiest solution is a cap on the number of trees per biome type. It even allows for a more random distribution, as opposed to trees being evenly spaced apart.

I assume you mean "trees per embark tile, varying by biome".

It's simple, it achieves the desired goal, and it would add almost nothing to FPS - saplings already repeatedly attempt to grow or die, this would just modify the probability each uses based on the trees already on that embark tile.
Logged
Sort of back maybe?

BAMM! lets you switch up your graphics without destroying your mods. Currently in v0.1: The Graphicsenating.

Worship!

Vattic

  • Bay Watcher
  • bibo ergo sum
    • View Profile

Edit: Decided to let a fort run with the fps uncapped to get some quick comparison shots. The embark is a "sparse" savannah.
Thanks. This illustrates the issue perfectly. It doesn't even seem to be respecting minimum distance. Shade-less trees?

What's the vegetation supposed to be on that embark? Is it equally a problem?
No problem :), it's not too lengthy a test with FPS uncapped and NO_EAT, NO_DRINK dwarves.

I didn't think to check vegetation before embark because of a myopic focus on trees. Figured the bright display was fairly representational of arid climates when in bloom, but it would be interesting to know if vegetation does the same as trees.


Logged
6 out of 7 dwarves aren't Happy.
How To Generate Small Islands

Bumber

  • Bay Watcher
  • REMOVE KOBOLD
    • View Profile

I didn't think to check vegetation before embark because of a myopic focus on trees. Figured the bright display was fairly representational of arid climates when in bloom, but it would be interesting to know if vegetation does the same as trees.
Just retire and check a tile from that same biome.

But first check the z-level above. I'm curious if it's only shade from leaves that's stopping forest embarks from being even worse. Could just be that "two saplings growing at the same time" thing either way.
Logged
Reading his name would trigger it. Thinking of him would trigger it. No other circumstances would trigger it- it was strictly related to the concept of Bill Clinton entering the conscious mind.

THE xTROLL FUR SOCKx RUSE WAS A........... DISTACTION        the carp HAVE the wagon

A wizard has turned you into a wagon. This was inevitable (Y/y)?

Neonivek

  • Bay Watcher
    • View Profile

Here is something. Why don't we just have the game store the different tree states in its memory at all times and switch between them during the seasons instead of simulating the entire process?

Boom fixed.
Logged

PatrikLundell

  • Bay Watcher
    • View Profile

@Neonivek:
The FPS effect of season changes is a bit outside of the scope of the original suggestion, but since other tree related effects such as pre embark tree generation have been discussed I guess it's not totally off.
Switching between different tree state sets would sort of work if trees were static, but since they grow the state from the previous year for the following seasons may no longer be correct. Of course, tree growth could update all sets at the same time. However, set changes would behave differently from the current model where trees change individually, unsynchronized (as an analogy, change to a snow covered winter state would immediately change the ground from clear of snow to completely snow covered from one tick to another rather than by falling snow gradually covering the ground). I would guess most players would care more about the FPS than the looks (full blown graphics whores aren't playing DF anyway, after all), but I think it goes against Toady's ambitions of a gradually more detailed and complete system/simulation.
Also, I suspect most of the FPS effect comes not from the trees themselves, but the stuff they drop.
Logged

Dozeb˘m Lolumzalýs

  • Bay Watcher
  • what even is truth
    • View Profile
    • test

GoblinCookie, why do you persist in stating that nothing is wrong with the current tree growth model? Savannas look like forests after several years. That. Is. Wrong.

And you evidently cannot grasp probability theory, so stop using it to support your argument. "400 rolls," my dwarven butt. We have much evidence that treesplosion is real. You need to support your argument with actual evidence.
Logged
Quote from: King James Programming
...Simplification leaves us with the black extra-cosmic gulfs it throws open before our frenzied eyes...
Quote from: SalvanÚ Descocrates
The only difference between me and a fool is that I know that I know only that I think, therefore I am.
Sigtext!

GoblinCookie

  • Bay Watcher
    • View Profile

That's very doubtfully due to just simulating tree growth. There's all sorts of background history stuff going on, pathing, tree litter tracking, and growing branches off the existing ones. The problem is that trees are not static after they've been loaded into the game, but must continue to grow. (Also, as PatrikLundell stated, that's CPU, not memory.)

I think you'd be hard-pressed to argue that a set of Cartesian coordinates, a bounding box (length & width,) and a value for age take as much memory/CPU as these same things plus a physical layout. It is more optimal to do everything up front with the former (and then turn them into physical forms,) than it is to have the physical trees and model every physical stage of their growth while also making sure the player didn't go and build something there.

That is already how the original trees are placed. 

No, it's actually not clear at all. While I'm not sure if one would actually notice the difference, the trees would start to grow through each other post-embark anyway. (I can only assume this is what you think makes them interesting. Everything else boils down to placement, which you wouldn't be able to discern from a proper pre-embark algorithm.)

Everything boils down to generation vs scripting, the devs tending to prefer the former to the latter as their ultimate goal. 

Quote
None of those are DF pictures so I have no idea what the final situation in those biomes looks like.
Me neither. Unfortunately I don't have a whole lot of spare time to find out at the moment. Word of mouth says low tree density isn't enforced. Edit: Vattic posted pics below.

Try imagining the dots as tree trunks, not leaves. Compare with your ground-level picture.

What controls the growth of new saplings is the blob of the tree canoply above not the tree trunks. 

Statistics trick question used as an example. There's no guarantee you won't roll 400 ones, however unlikely.

The metaphor extends as thus: Each embark tile is 48x48 = 2304 game tiles. That's 9216 tiles on the default 4x4 embark. If the game attempts to place saplings randomly without respect to existing trees (my assertion based on Occam's Razor), then for each sapling placed there is a 1/9216 chance per viable spot to place it such that it will grow. As the trees fill the map, it gets less likely to find a remaining good spot. I assert that this is why we see these 'clusters'. While I have not sufficiently proven this assertion, neither have you disproven it. It exists as an alternative explanation.

If things worked that way the devs could simply have placed as many trees as there were good spots to grow on.  The reason I think that clusters are in effect rather than dots is because saplings tend to appear in proximity to existing trees even when they are a completely different type of tree while ignoring the far more inviting open areas of the map. 

Eventually, yes, but there are no guarantees as to when that will happen. It depends on how frequently placement occurs for those 7 years. We lack the information for statistical analysis.

When I have cleared areas of the map, saplings popped up pretty fast in large numbers. 

What I meant was do these zones persist? If you clear the map of trees and vegetation, will no trees grow in that zone? If it happens once, your assertion is disproven. It's easy enough to test, and I would expect at least one of your zones to get a tree within the first year or so without the interference of existing foliage. It's not practical for me to test this because I don't have your map, nor a complete understanding as to what you consider a zone.

I have to prove treesplosion occurs on arid embarks (which I might do during winter break, if someone else doesn't Edit: Vattic did below.) You have to defend the existence of clusters, as well as your "clearly working as intended" assertion.

I have shown you the clusters, you have to prove that the clusters are the result of an insufficiant length of time rather than being simply how the mechanics grow the forests in clusters.  Your position is unfalsifiable because it could always be that any observable cluster pattern however was 'just an accident' however improbable. 

The following thread starts with a screenshot of a savannah biome after a few years. It is suggested that felling existing trees might be making space for all the embark generated saplings to grow. It also has a number of other people reporting the same occurrence.

Edit: Decided to let a fort run with the fps uncapped to get some quick comparison shots. The embark is a "sparse" savannah.

Spoiler: year 5 (embark) (click to show/hide)

Spoiler: year 8 (summer) (click to show/hide)

Spoiler: simplified comparison (click to show/hide)

Sorry about the poor image quality. I neglected that I had a high resolution tileset installed and the final images would have been massive without a resize.

I think they give a fair idea what people are talking about.


I fail to see what anyone is talking about.  As already mentioned, the total number of trees on the map to start with is less than the total number of trees that there *should be* according to the mechanics of tree growth seperately in the game.  That is because the original trees are scripted but what the devs ultimately wanted is a generated trees rather than scripted and randomised trees on the map; but the CPU power did not exist to generate such a forest automatically on embark. 

Your savanna map shows the cluster effect far more strongly than my own screenshot did, although it is only 3 years rather than 7 years apparantly.  This confirms what I have been saying, the map is randomly divided up into forests/nonforested zones, the less forested the map is supposed to be the larger the nonforested zones are.  While the non-forested zones are small on my map, the non-forested zones are huge on yours, the whole eastern section of your map is such a zone. 

It also confirms another thing I have been saying.  The 'problem' appears to be that there is no difference in the kill zone of the trees on the different biomes, this means that the forested zone on a savanna is ultimately equally dense to that in a forest.  That however is likely because the mechanics of nutrients and water are intended to ultimately be implemented in the game, the devs have not wasted their time coming up with complicated placeholder mechanics to control tree growth until the time those things are added in.

GoblinCookie, why do you persist in stating that nothing is wrong with the current tree growth model? Savannas look like forests after several years. That. Is. Wrong.

And you evidently cannot grasp probability theory, so stop using it to support your argument. "400 rolls," my dwarven butt. We have much evidence that treesplosion is real. You need to support your argument with actual evidence.

'Treesplosion' is the result of hearsay and rumour not evidence.  The game has quite solid mechanics in place to control tree growth and no savannas do not look like forests, forests do not look like forests; nobody can drive a wagon through an actual forest.
Logged

Bumber

  • Bay Watcher
  • REMOVE KOBOLD
    • View Profile

That is already how the original trees are placed.
Which begs the question: Why, in your opinion, would they not all be placed up front? We've just established it's less CPU/RAM intensive, so your previous theory falls flat on its face.

Quote
Everything boils down to generation vs scripting, the devs tending to prefer the former to the latter as their ultimate goal.
I'd ask for clarification, but I really don't care anymore at this point.

Quote
What controls the growth of new saplings is the blob of the tree canoply above not the tree trunks.
What happened to given distances / kill zones? It's just tree shade killing saplings at that point.

Quote
If things worked that way the devs could simply have placed as many trees as there were good spots to grow on.  The reason I think that clusters are in effect rather than dots is because saplings tend to appear in proximity to existing trees even when they are a completely different type of tree while ignoring the far more inviting open areas of the map.
"Good spots", as it stands, is everywhere. No, not from what I've seen.

Quote
Eventually, yes, but there are no guarantees as to when that will happen. It depends on how frequently placement occurs for those 7 years. We lack the information for statistical analysis.
When I have cleared areas of the map, saplings popped up pretty fast in large numbers.
This is meaningless. Where? Same places? This information only supports the notion that it gets less likely to place good saplings as the map fills.

Quote
I have shown you the clusters, you have to prove that the clusters are the result of an insufficiant length of time rather than being simply how the mechanics grow the forests in clusters.  Your position is unfalsifiable because it could always be that any observable cluster pattern however was 'just an accident' however improbable.
No, you propose clusters, you have to prove their existence is not due to sampling error. Burden of proof is on you.

Quote
I fail to see what anyone is talking about.  As already mentioned, the total number of trees on the map to start with is less than the total number of trees that there *should be* according to the mechanics of tree growth seperately in the game.  That is because the original trees are scripted but what the devs ultimately wanted is a generated trees rather than scripted and randomised trees on the map; but the CPU power did not exist to generate such a forest automatically on embark.
Then maybe you need to get your eyes checked. This is a savanna. That is not what everyone else is seeing in the year 8 map. If you can't understand that, then there's nothing I, nor anyone else, can do for you.

Quote
Your savanna map shows the cluster effect far more strongly than my own screenshot did, although it is only 3 years rather than 7 years apparantly.  This confirms what I have been saying, the map is randomly divided up into forests/nonforested zones, the less forested the map is supposed to be the larger the nonforested zones are.  While the non-forested zones are small on my map, the non-forested zones are huge on yours, the whole eastern section of your map is such a zone.
That's a (mountain?) biome boundary. You can tell because it's on the edge of the map and not even grass will grow there.

Quote
It also confirms another thing I have been saying.  The 'problem' appears to be that there is no difference in the kill zone of the trees on the different biomes, this means that the forested zone on a savanna is ultimately equally dense to that in a forest.
Which ultimately disproves this:
Since all trees must grow a given distance from other trees at the moment and there are variables in a biome deciding how wooded it is; I have yet to be convinced that your suggestion has not already in essence been implemented.
It's exactly the opposite of what you've been saying the entire time. Nice try.

Quote
'Treesplosion' is the result of hearsay and rumour not evidence.  The game has quite solid mechanics in place to control tree growth and no savannas do not look like forests, forests do not look like forests; nobody can drive a wagon through an actual forest.
::)

"To argue with a man who has renounced the use and authority of reason [...] is like administering medicine to the dead [...]" -Thomas Paine

I'm out.
Logged
Reading his name would trigger it. Thinking of him would trigger it. No other circumstances would trigger it- it was strictly related to the concept of Bill Clinton entering the conscious mind.

THE xTROLL FUR SOCKx RUSE WAS A........... DISTACTION        the carp HAVE the wagon

A wizard has turned you into a wagon. This was inevitable (Y/y)?

TheBiggerFish

  • Bay Watcher
  • Somewhere around here.
    • View Profile

*sigh*
Can we at least agree that there should be a cap on the number of trees per tile in a biome, regardless of how much that number is, and debate balancing later?
Logged
Sigtext

It has been determined that Trump is an average unladen swallow travelling northbound at his maximum sustainable speed of -3 Obama-cubits per second in the middle of a class 3 hurricane.

Bumber

  • Bay Watcher
  • REMOVE KOBOLD
    • View Profile

*sigh*
Can we at least agree that there should be a cap on the number of trees per tile in a biome, regardless of how much that number is, and debate balancing later?
No, this is exactly what GoblinCookie disagrees with because something, something, tree farms.

Edit: He said it was too artificial, not a realistic depiction, and/or already exists.
« Last Edit: November 28, 2015, 11:29:29 am by Bumber »
Logged
Reading his name would trigger it. Thinking of him would trigger it. No other circumstances would trigger it- it was strictly related to the concept of Bill Clinton entering the conscious mind.

THE xTROLL FUR SOCKx RUSE WAS A........... DISTACTION        the carp HAVE the wagon

A wizard has turned you into a wagon. This was inevitable (Y/y)?

TheBiggerFish

  • Bay Watcher
  • Somewhere around here.
    • View Profile

... Dwarves planting trees is different from natural tree growth.

And I must disagree with the assertion that anything like a tree cap exists, based on what this thread has provided.

... Too artificial?  The entire game is artificial.  The realism issue is disproved by even a cursory look at the evidence.  Again.
« Last Edit: November 28, 2015, 11:41:17 am by TheBiggerFish »
Logged
Sigtext

It has been determined that Trump is an average unladen swallow travelling northbound at his maximum sustainable speed of -3 Obama-cubits per second in the middle of a class 3 hurricane.
Pages: 1 ... 4 5 [6] 7 8