Print

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

[BigD145]

I've been doing some experimenting.  Turns out the game doesn't cope well, currently, with 3000 types each of dummy creatures and vermin.

I don't think the Ark would have coped well with all those animals, either. Unicorns, dragons, and gryphons didn't seem to make it on.

[Footkerchief]

Here's the file I used for stress testing.  Use at your own risk -- trying to load DF with this file in the raws is probable to bring your whole OS to its knees (from thrashing, I think) unless you have 2+ GB of RAM.

This doesn't mean the project is doomed, as there are probably many optimizations (both in speed and mem usage) that Toady can make at some point.

Do you know what he can do specifically, or is that merely a reasonable assumption?

Just an assumption.  There's probably been no attempt at optimizing the raws loading process, because it's not an issue with vanilla content.  Additionally, each creature gets full copies of materials, tissues, body structure, etc., even if they're identical in all but name.  So adding a little hierarchical structure to say, tissue storage, would probably eliminate a large chunk of the overhead.

I've been doing some experimenting.  Turns out the game doesn't cope well, currently, with 3000 types each of dummy creatures and vermin.  It's pretty much impossible to load on my machine due to crappy memory (1 GB), although on my friend's much better machine (4 GB) it runs fine after a few minutes of loading (although it takes up 1 GB of memory all by itself).

Does it slow down in-game stuff or is it mainly Worldgen/loading type things? If it bogs down creature creation in arena, we could use the arena_restricted tag judiciously.

It slows down the process of loading the arena itself.  I haven't had a chance to try it with worldgen, since my own computer can't handle it at all (I sent it to a friend for the arena test).  It would be great if someone (with 2+ GB of RAM!) could try worldgen.

If abstract creatures consist only of prefstrings, should they have the [DOES_NOT_EXIST] tag, or will the game figure it out?

I'm not sure precisely what the tag does. If it just removes them from stockpile menus and stops people liking their skin/bone I guess so. Does it have any wierd side effects?

Dunno.  Barring weird side effects, though, it'd be perfect for our needs.  The wiki parser script could probably add that automatically... no point in manually adding it to every single Abstract Creature.

[Zantan]

Turns out the game doesn't cope well, currently, with 3000 types each of dummy creatures and vermin.

We could write a program that takes the massive ark raws and produces a 10% sample of its creatures.  The hierarchical setup would allow the program to make sure the inclusions are spread out relatively evenly among different types of creatures.

This may even make it more realistic, because you wouldn't have a world's worth of biodiversity packed into every biome.  Also, each playthrough will be more distinct, since each would have a unique combination of creatures.

[Lancensis]

Turns out the game doesn't cope well, currently, with 3000 types each of dummy creatures and vermin.

We could write a program that takes the massive ark raws and produces a 10% sample of its creatures.  The hierarchical setup would allow the program to make sure the inclusions are spread out relatively evenly among different types of creatures.

This may even make it more realistic, because you wouldn't have a world's worth of biodiversity packed into every biome.  Also, each playthrough will be more distinct, since each would have a unique combination of creatures.

Oh! That's not a bad idea at all. Certainly a nice interim measure.

[Zantan]

I'd be happy to write the program, once we're further through the project.

[Gorobay]

Dunno.  Barring weird side effects, though, it'd be perfect for our needs.  The wiki parser script could probably add that automatically... no point in manually adding it to every single Abstract Creature.

Actually, do we even need abstract creatures? Couldn't we write, for each creature, lines like [APPLY_CREATURE_VARIATION:FISH_PREFSTRING], where that variation would consist only of [CV_NEW_TAG:PREFSTRING:shiny scales]? If this works I think it would be better, because then someone could "like pirahnas for their shiny scales" instead of just "fish".

[Zantan]

As I recall, the purpose of having abstract creatures was so that dwarves could like fish in general, rather than specific types of fish, to make preferences more interesting. As far as whether we need them, we don't, but necessity is not the driving force behind the Ark Project, or the majority of dwarfy endeavors for that matter.

Assuming it can be implemented easily enough, the only potential problem with abstract creatures is that a dwarf who likes fish will never be pleased to see/own any kind of fish.  On the other hand, the Ark Project will make it incredibly difficult to find specific creatures, so I'm not sure how much that matters.

[Footkerchief]

Turns out the game doesn't cope well, currently, with 3000 types each of dummy creatures and vermin.

We could write a program that takes the massive ark raws and produces a 10% sample of its creatures.  The hierarchical setup would allow the program to make sure the inclusions are spread out relatively evenly among different types of creatures.

This may even make it more realistic, because you wouldn't have a world's worth of biodiversity packed into every biome.  Also, each playthrough will be more distinct, since each would have a unique combination of creatures.

Oh! That's not a bad idea at all. Certainly a nice interim measure.

Yeah, I had thought about picking random subsets, but it hadn't occurred to me that it could actually be a good thing not to have all creature types in one world.  That could also be a solution to the weirdness we discussed a while back, where you'll get Australian and North American fauna in the same biome.  We could include some basic geographical metadata in our raws (XML?) as hints for the program that selects the subset.

As I recall, the purpose of having abstract creatures was so that dwarves could like fish in general, rather than specific types of fish, to make preferences more interesting. As far as whether we need them, we don't, but necessity is not the driving force behind the Ark Project, or the majority of dwarfy endeavors for that matter.

Yeah, the whole point of the piranha example was that it seems a little weird, to me at least, that a dwarf could single out piranhas for their shiny scales when it's a quality shared by hundreds of thousands of other fish.  But liking fish generically is also a little weird... I don't know, I could really go either way on this.

Now, one day when prefstrings make more sense, maybe a dwarf could like piranhas for their shiny scales because that's the only shiny-scaled fish he's ever encountered.  But that requires support for abstract creature classes in the raws themselves -- 31.01 has some creature classes, but they're just simple groupings without the ability to define characteristics.

[Zantan]

We could write a program that takes the massive ark raws and produces a 10% sample of its creatures.  The hierarchical setup would allow the program to make sure the inclusions are spread out relatively evenly among different types of creatures.

That could also be a solution to the weirdness we discussed a while back, where you'll get Australian and North American fauna in the same biome.  We could include some basic geographical metadata in our raws (XML?) as hints for the program that selects the subset.

Firstly, XML will not be necessary here, since we can include any necessary metadata in comment lines.

Spoiler (click to show/hide)

though if toady moved the raws into XML format this would be much easier

Also, geographical metadata would not be enough, even combined with the taxonomic classifications.  For example, we may decide that five frogs in a family are redundant with each other, but the sixth should always be included because of its distinct poison.  We would need a system that would allow a program to identify the five frogs and select two of them for inclusion, and always include the sixth.  Even worse, there will be cases where a group of redundant creatures include some family level entries and some genus level entries.

We could add our own redundancy groups to the raws in comment lines.  If we added a line containing '*normal_frogs' to the raws of the redundant frogs, a program could easily identify them, group them, and randomly choose which to include in the final raws without a geographical identifier.

How does that sound?

[Gorobay]

We could add our own redundancy groups to the raws in comment lines.  If we added a line containing '*normal_frogs' to the raws of the redundant frogs, a program could easily identify them, group them, and randomly choose which to include in the final raws without a geographical identifier.

How does that sound?

It sounds like a very reasonable and useful solution. A similar system could be used for the dialect problem discussed before:

Code: [Select]

[PREFSTRING:coloration] *en-US
[PREFSTRING:colouration] *en-UK
The compiler would include only one of the two lines.

[Dsarker]

As an Australian, I have to ask if you've added the Drop Bear yet.

[Lancensis]

As a tourist, I have to say how frightful that sounds, and ask if there's anything I can do to warn such a creature off?

[Rainseeker]

Yes, I would encourage you to buy this fine repellent for 5,000 credits.  It's guaranteed to keep them away at every tourist location!

[Deon]

A drop bear:

Spoiler (click to show/hide)

Works well with bear cavalry.

[Lancensis]

Yes, I would encourage you to buy this fine repellent for 5,000 credits.  It's guaranteed to keep them away at every tourist location!

Pfft. I could get like, five Magikarp for that.