Print

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

[Neonivek]

The populations arn't that bad.

Technically if a DF large world is supposed to be as large as our own world (which it isn't by both square measure and travel time.) then several of these animals are supposed to be more numberous. There are around 7 million horses.

There was more of a deficiency of mosquitoes... 24 million is actually a pathetically small amount.

300k Whalemen isn't all that bad.

[dirty foot]

He can still have killed 8000 elves.

But the game treats all those 8000 elves seriously when really it could just be extrapolated.

It's that extrapolation that would rob the game of the personality that many of us now expect.

1 of those 8000 could be a 3 hour detour when looking into legends, because we noticed that he or she also had a history.

[NW_Kohaku]

He can still have killed 8000 elves.

But the game treats all those 8000 elves seriously when really it could just be extrapolated.

It's that extrapolation that would rob the game of the personality that many of us now expect.

1 of those 8000 could be a 3 hour detour when looking into legends, because we noticed that he or she also had a history.

You're missing the point - you can still have that character get killed at that point in time, while abstracting the whole rampage point.

The problem with having literally several million werecreature attacks recorded in history, each one of which is several dozen rolls long is that it can almost single-handedly destroy a worldgen. 

Meanwhile, if you simply make worldgen have the same general results (dinging a village a chicken and a smallish chance to knock out a civvie and maybe spread the disease), or generally just extrapolating from the probabilities (assuming that if the werewolf is going on 3 rampages a year, he will get 12 chickens per year).

Abstraction is simply finding the integral of what is going on, and using that to far more quickly extrapolate the result for any variable you plug into the equation.  Abstraction does not need to any loss of the resolution of the data. 

[dirty foot]

The populations arn't that bad.

Technically if a DF large world is supposed to be as large as our own world (which it isn't by both square measure and travel time.) then several of these animals are supposed to be more numberous. There are around 7 million horses.

There was more of a deficiency of mosquitoes... 24 million is actually a pathetically small amount.

300k Whalemen isn't all that bad.

To be honest, it would make sense if things like mosquitos were were not tracked at all, and simply "spawned". They are so erroneous, it seems like one of the few things that can be manifested whenever the need arises.

[Rapozk]

I tried medium history once..

OVER 4 000 000 events at the year 300 something 8 hours later..

It was biggest possible world, with highest amounts of sites and civilizations...

I had to make it crash using taskmanager to get rid of it.

[Biopass]

So I'm currently 16 hours into a vanilla large world/long history generation. No crashes, astoundingly. As of the year 385, which has been genning now for close to 15 minutes, there are 672K historical figures and 9.1 million events. Currently taking up 2.5 gb ram. Got to give Toady massive kudos on the current stability, although I am of course using the LAW exe.

Is there a more elegant way to stop lagging worldgen than putting something heavy on the enter key? I may cut this off at 400 years (or tomorrow morning, which is the more likely) in case it does crash. I'd love to see someone with better skill at reading legends than I report on it.

[xaldin]

So I'm currently 16 hours into a vanilla large world/long history generation. No crashes, astoundingly. As of the year 385, which has been genning now for close to 15 minutes, there are 672K historical figures and 9.1 million events. Currently taking up 2.5 gb ram. Got to give Toady massive kudos on the current stability, although I am of course using the LAW exe.

Is there a more elegant way to stop lagging worldgen than putting something heavy on the enter key? I may cut this off at 400 years (or tomorrow morning, which is the more likely) in case it does crash. I'd love to see someone with better skill at reading legends than I report on it.

Post it up, I'm sure few of us will read it if you like. I suspect if you didn't 0 out the were curse though I know what at least a 3rd to half of the events are...

[Biopass]

I did not zero out the curses, no. So yes, that is most likely what a large portion of it is.

I have to wonder, would running a smaller world to thousands of years result in everyone in the world being a lycanthrope of some sort?

[Feb]

If we can find out the difference between Masterwork and vanilla, we can probably pinpoint the problem.  I've noticed masterwork's worldgen runs a LOT faster, although I've no idea why.

[xaldin]

I did not zero out the curses, no. So yes, that is most likely what a large portion of it is.

I have to wonder, would running a smaller world to thousands of years result in everyone in the world being a lycanthrope of some sort?

Well I tried that as part of my testing to see impact of the curses. The system still typically bogged down around 400-500 to point of near non progression. Being somewhat impatient I'm not inclined to let something keep running indefinitely so once it hits the 5+ minutes per year I call it.  Interestingly enough I had to turn off the abort due to beasts killed cause after first 17 stopped before year 200 from all the FBs dieing off.

[dirty foot]

I did not zero out the curses, no. So yes, that is most likely what a large portion of it is.

I have to wonder, would running a smaller world to thousands of years result in everyone in the world being a lycanthrope of some sort?

With no cure for it but death, we're looking at a constant exponential growth.

We may have figured out exactly why world gen is scarfing up computer resources.

[Biopass]

Werewolves, and the new trade interactions, yes.

Sadly, the world I was genning simply couldn't be stopped. Spamming enter, holding the key down... nothing. So I'm going to gen a 400 year large world (hit 400 on the other one about 18 hours after it started) and a 1050 year large world with no curses. We'll see what happens.

EDIT: Looking through worldgen, saw "max regional interaction types" mixed in with the vampire/werewolf stuff. Anyone know what that is/does?

[Eagleon]

EDIT: oh and just in case that wasn't bad enough, nvidia doesn't currently support openCL, and meanwhile CUDA is a proprietary platform. so people with AMD cards would be unable to make use of any accelerations in CUDA code, unless Toady learned ANOTHER API, and rewrote the gpu accelerations AGAIN in ANOTHER form.

Happily, you're three years out of date. nVidia released drivers with OpenCL capabilities long ago. Any card that supports CUDA now also supports OpenCL, retroactively to the first released. It's also become a lot more straightforward to learn, with much more example code and tutorials floating around. With a little bit of work you can develop some custom functionality that selectively makes use of any available OpenCL devices - things like working through a container of numbers to tweak are particularly good targets - and use it at will later without even having to think about anything but your own code/language.

Also, which is the gamer that you know that doesn't have a dedicated video card? The average consumer, in this case, is not a good representation of what DF players will have in their box. Hell, even some integrated devices support it nowadays, especially AMD boards.

EDIT: Looking through worldgen, saw "max regional interaction types" mixed in with the vampire/werewolf stuff. Anyone know what that is/does?

I believe those control the rain/cloud/animation interactions. It's set seperately so that you can have a smaller variation in one, but a different number of discrete regional effects - combinations of rain/no rain, animation names, enthrallment, cloud types, and possibly activation times, though I'm not sure that varies yet.

[Biopass]

Did another large world, had it stop at 400 years. Took a little over 10 hours to complete. The 1050 year large world with no curses crashed out, I didn't get to see what year it was on.

Link for the 400-year large world: http://www.mediafire.com/?s08mvsb0cncllja

With this one taking 10 hours to finish and the other one taking about 16 to reach the same point, I'm thinking the geography has a lot to do with it. I think trade interactions also cause a lot of slowness in worldgen, and geographical proximity of trading (or warring, or both) nations has a major affect. Populations were also unusually low; whereas the old one had a population of 700K+ when I killed it, this world only has 343K historical figures in legends (6 million events). So population also has a pretty massive affect on speed as well.

Prepost edit: Scanned the world map for the large world. Civs are pretty split up, and there are a lot of really large mountain ranges that definitely divided the world up more than normal

[NW_Kohaku]

Yeah, the two major things that cause slowdowns are trading and werecritter rampages.

I've had worldgens that were very slow with just one single civ before, just because they expanded unchecked and traded between their cities. 

Meanwhile, worlds where most of the civs are isolated and left with no room for expansion, or where the populations drop very low and never recover are relatively fast (and also leave few targets for werecritters). 

Werecritters, meanwhile, are the other major thing, which of course bloats up your legends with a million beast rampages of no importance.