Print

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

[A Dwarven Smokeologist]

So a hobby of mine recently has been to tweak worldgen files & entity population limites and then attempt to gen the largest most lively worlds possible just to see what happens.

Finalizing Gen 1 -

Spoiler (click to show/hide)

Testing some settings for civilizations (maximum number for 257x257 world) 600 civs in the world gen init is almost the maximum I have been able to fit in with out getting many world rejections. I assume I could fit more in there (maybe 50 or so) but never managed to get an unrejected world with 700.

Civilization Map Year 10 -

Spoiler (click to show/hide)

You can try and count the civs there if you want but also keep in mind that many are placed close together and their borders overlap, no animation to show it though.

Civilization Map Year 250 -

Spoiler (click to show/hide)

Here the same civilizations are 250 years later when world gen finally ended. There were many wars fought and blood spilled through those years, and none of it was forgotten by the time year 250 rolled around.

I don't know as much about the worldgen / entity files as many people around here so am wondering now, what is the largest world you can gen? Is there a maximum limit to the size of the world and what it can be pushed to? I will say I've never had a world gen succed if I allowed it to continue to long, if starts to approach 5million total events I've had to end it or it always crashes.

Can the engine handle more than 5million events? How many people can fit into the world at once?

[darius]

I'm guessing it should hang when 4 294 967 296 (2^32 size_t for 32 bit programs AFAK) events are reached. So yeah that's engines limit.

[Hummingbird]

Legends would be real fun to read in that world.

Alternatively, what you can try is to have fewer civs but remove the population cap so that you have epic huge civs fighting it out. Remember to try to make large, continuous mountain chains and forests so that your elves and dwarves have room to expand.

What's sad about world-gen at the moment is that humans, who are guaranteed to be the largest civs in such conditions, will never war with each other.

[Warlord255]

Legends would be real fun to read in that world.

Alternatively, what you can try is to have fewer civs but remove the population cap so that you have epic huge civs fighting it out. Remember to try to make large, continuous mountain chains and forests so that your elves and dwarves have room to expand.

What's sad about world-gen at the moment is that humans, who are guaranteed to be the largest civs in such conditions, will never war with each other.

I have attempted an experiment faintly similar to this; I tried to mod Orcs' entity numbers and that of others to accomodate massive, world-spanning unified empires. I called it the Genghis Khan project. I got it to work at least once.

It should be noted that one CAN limit the maximum number of civilizations in a given entity, but unless your race is so rapaciously invincible that they cannot be defeated, there's at least a 50/50 or less chance of a given civ either dying or being stifled by competition.

[vooood]

I'm guessing it should hang when 4 294 967 296 (2^32 size_t for 32 bit programs AFAK) events are reached. So yeah that's engines limit.

this could be easily overcome with some clever coding..

[lordnincompoop]

I'm guessing it should hang when 4 294 967 296 (2^32 size_t for 32 bit programs AFAK) events are reached. So yeah that's engines limit.

this could be easily overcome with some clever coding..

Not unless that hacker is the great Toad himself. I'm not a programmer, but this is probably source code stuff, and DF is closed source.
Even if some legendary hacker cracks DF's capacious shell open and does the necessary arrangements, that version will be obsolete in a week or two when the new 9008123535465z comes out.
Plus it involves changing the variable type, or even the whole program, and as that size will prolly be 64 bit (which is very rare), it means a lot of work and a large number of players booted for incompatibility (methinks).

[termitehead]

I'm guessing it should hang when 4 294 967 296 (2^32 size_t for 32 bit programs AFAK) events are reached. So yeah that's engines limit.

this could be easily overcome with some clever coding..

Don't really need to move to 64 bit or anything.  You could have 2^32 events for each age as a work-around.  If it ever got to the point where you ran out of event id's you could force the beginning of a new age.

Not unless that hacker is the great Toad himself. I'm not a programmer, but this is probably source code stuff, and DF is closed source.
Even if some legendary hacker cracks DF's capacious shell open and does the necessary arrangements, that version will be obsolete in a week or two when the new 9008123535465z comes out.
Plus it involves changing the variable type, or even the whole program, and as that size will prolly be 64 bit (which is very rare), it means a lot of work and a large number of players booted for incompatibility (methinks).

hmmm ... my response got deleted...

Anywho, if the events are indexed by age index then you could have 2^32 * 2^32 events.  Thus the 1st age (Age of Myth or whatever) would be index 0 with 2^32 events, the next age would be index 1 with another 2^32 events.  It's a workaround and it doesn't require going to 64 bit.

[shadow_slicer]

But you'd still need the source code to make that change. And of course, if you had the source code, this change would be fairly simple, since most compilers support 64 bit types even for 32 bit archs. The only change that would be necessary, is to change

Code: [Select]

unsigned eventidxto

Code: [Select]

uint64_t eventidx in every place that needs an event index.
This would be much simpler than adding an extra layer of indexing for each age or whatever.

Of course I'd be willing to settle for a native 64 bit version of DF. In addition to not having this problem, odds are it would run 2-3 times faster than 32 bit*.

*Note that this is not due to the difference between 64 and 32 bits so much as the significantly increased register count available to 64 bit programs. Tests performed for the pathfinding project show increases in speed of 2-3x simply due to compiling the algorithm as 64 bit instead of 32 bit.

[azurereaver]

nobody seems to have pointed out that 2^32 is 4,200 million, and so not a very good explanation for why it hangs at 5 million (or 4.2 million).  The concept of an overflow event is certainly the right area to be worried about.  Could be a memory limitation - you can monitor the heap and stack usage to get an idea if that is the problem.

Interesting to know about the # of 64 bit registers.

Seems to be a lot of knowledgeable people bouncing through these boards.

[SirPenguin]

The world I generated for the topic in my sig is actually one of the most interesting ones I've ever seen.

It's large (256x256 I believe) and chock full of caves and civilizations.





It not as crammed together as your map, but everything feels so natural. Some groups are isolated, others have a few neighbors, then in the southwest you have a huge clusterfuck of civs vying for control. The Kingdom of Fishes, the behemoth teal nation in the center that spans the ocean, is definitely the coolest dude on the block.

Also, you can't really see it, but in the northwest is an island forest full of elves. One day I plan on swimming the ocean to that island, if only to kill them all

[zwei]

I can't help, but notice that most kingdoms do not really have realistic shape and look more like bunch of overlaping circles.

I'd expect more boundaries based on rivers for example.

[LeoLeonardoIII]

It may be biome based. I see some sharp edges too.

[milaga]

I imagine empires continue to expand until they meet resistance. A river isn't much resistance unless there is an army on the other side of it.

Also, because empires seem to overlap so much, these maps are most likely maps of influence rather than concrete borders.

[quinnr]

Neat project! That's a ton of civs!

BTW, what font are you using the the first screenshot? Looks cool.

[Judas Maccabeus]

Also, because empires seem to overlap so much, these maps are most likely maps of influence rather than concrete borders.

I would be inclined to agree.  Pre-modern societies weren't very good at having honestly defined borders, especially as for most of them any centralised authority would be limited or non-existent outside of cities and towns.*  The odd well-organised empire such as Rome or China would be more the exception than the rule; others certainly claimed large areas of territory but either had delusions of grandeur or had large areas paying lip service and then doing their own thing.  Note that in Dwarf Fortress, this is generally the case as well (the civilisations' areas of control are based around settlements, and dangerous creatures, as well as members of enemy groups, can regularly be found within said areas outside of settlements).

Alright, enough boring historiography, carry on with the fun projects. 


*An even bigger problem considering that pre-modern societies tend to be overwhelmingly rural as well...