Announcement: SEED system

[Tels]

Also made two new falloff maps "lode_cloud" and "lode_sinus":

 

 

Both were made quickly with Gimp (one with the "Sinus pattern plugin", the other with some plasma cloud and then blackening the border areas).

 

In the game they can be used to create more natural distributions of vegetation, here the "cloud":

 

 

Code-wise, I also fixed a problem with imagemaps: an imagemap did not modify the density, so if you had a map with 50% black and 50% white, you would still get 100% of the entities, but distributed over 50% of the area => twice as dense forest f.i. Now the average image map density is computed and used to correct the density before computing the amount of entities to generate. This way you can swap the image map and the most-dense area still is not denser (or less dense) than before.

 

Here is an overview of the testscene with some explanations:

 

 

Without the bugfix mentionsed above, the second and the last LODE area would have had two respectively four times as many entities crammed into the same space.

[Fidcal]

Sorry had no time to look at this Tels. It needs quite a bit of study and a day or two experimenting but I'm racing with my own FM and reluctant to break away. Seems every other day there is something to try to divert me from it so I'm trying to ignore everything else.

[Tels]

Sorry had no time to look at this Tels. It needs quite a bit of study and a day or two experimenting but I'm racing with my own FM and reluctant to break away. Seems every other day there is something to try to divert me from it so I'm trying to ignore everything else.

 

No problem, we should have plenty other mappers to look at it right? Right!?

 

Anyway, I am also short on time and did not have time to look at my planned mission at all, and everytime I do, it seems more model or more code are needed first *sigh*

[ocn]

Incredible work there, Tels. Compliments.

[Tels]

More work done today:

 

The image-based density map is nice, but havign to draw a new map everytime is a bit cumbersome, so I thought about adding options to it. First option is:

 

* "map_invert" (on LODE) or "lode_map_invert" (on Target classes)

 

This inverts the map before using it, meaning you can now add f.i. two targets, let one use the image map and one the inverted version, and the first model will only be where the second model isn't and vice versa.

 

Fixed a bug with random skin, too. Also, based on this 1919 drawing:

 

http://chestofbooks.com/flora-plants/flowers/British-Wild-Flowers-2/Flowers-Of-The-Heaths-And-Moors-Plate-LVII.html

 

made a calluna plant (18 tris, only one material). Here is a screenshot showing both (new model with pink and reddish skin) and using an inverted map - you can see the calluna only "grows" where the grass doesn't:

 

 

Closup:

 

 

Think I know how to improve that model while increasing the tris from only 18 to 22, but had no time today anymore.

 

As always, my textures look like crap, so if somebody wants to improve them, be my guest.

[Aprilsister]

A big fan of this as I've said. Would love to see it work out and become the thing that pushes TDM into ModOfTheYear for 2011 -- which, if it works out, it would certainly qualify as such a (revolutionary!) feature that there would be no denying the MOTY award (no matter what actually happens)... and if it can be expanded to work with AI models? The award goes out that very day.

 

But, here's a bit of criticism (just the kind of thing that bugs):

 

If you don't want it getting confused with LOD... don't call it LODE. Seems like one of those things that started off being somewhat associated (or maybe not) and the point was to get it done which was the focus and the idea has become completely separate in the meantime but the step back was never taken to go ahead and call it something unique... as often happens with many things in the tech-world(s). There is no reason to stick with arcana.

[Tels]

But, here's a bit of criticism (just the kind of thing that bugs):

 

If you don't want it getting confused with LOD... don't call it LODE. Seems like one of those things that started off being somewhat associated (or maybe not) and the point was to get it done which was the focus and the idea has become completely separate in the meantime but the step back was never taken to go ahead and call it something unique... as often happens with many things in the tech-world(s). There is no reason to stick with arcana.

 

Yeah, it starts to bug me, too. I think it might be possible to rename it for v1.04 because nobody has used it yet, but afterwards it might be tricky. And proposals for a new short name?

 

Procedural Content Generator (RCG) doesn't sound as sexy...

[Aprilsister]

How about what you've already described it as? It is your baby after all.

 

I'll just tweak it a bit to sexy the acronym:

 

Procedural, Semi-random Iterator

 

PSI

[Fidcal]

Well I was going to suggest the very same thing but I usually get opposed when I start talking about misuse of words like brushes, layers, overall, etc.

 

Now I can't remember what alternative I thought up but I think it was some variation of REG or ERG or RMG for random entity generation something or other.

 

The calluna looks OK to me at least from the screenshot. More bushy plants the better say I. Though I don't think I know the word nor does Opera spellchecker. Ah I see it's like heather.

[rich_is_bored]

SEED

 

System for

Environmental

Entity

Distribution

[Aprilsister]

Well I was going to suggest the very same thing but I usually get opposed when I start talking about misuse of words like brushes, layers, overall, etc.

 

Well, I was happy that the suggestion wasn't taken the wrong way. And yes, if there is a great example of the kind of thing I was talking about in my rambling bit, it is "brush"... admittedly, when I said "kind of thing that bugs", I meant the "kind of thing that can get me more likely ranting than just rambling" in the case of something like "brush", which isn't just a possibly confusing acronym but an all-out what-the-fuck.

 

I like SEED. It sure makes a good acronym for the function of this module as it stands; but, Tels, can you envision this working on AI as I've mentioned? That it might just work on NPCs or other more "advanced" Entities (heads + bodies + weapons etc) might be semi-randomly generated?

[Tels]

I like SEED. It sure makes a good acronym for the function of this module as it stands; but, Tels, can you envision this working on AI as I've mentioned? That it might just work on NPCs or other more "advanced" Entities (heads + bodies + weapons etc) might be semi-randomly generated?

 

I like SEED, too, because I tried to fit "natural" words like GROW, POD, STEM etc to find something but failed, so SEED is cool.

 

Any other opinions?

 

As for AI, the module doesn't care what entities it generates, I tested it with picture frames, and they work. AI do have some problems, but I guess this is fixable. Would that impact the name choice? I don't think so.

[Fidcal]

Yes, SEED is good. I was trying to think up something like ERG or GRID but couldn't make up anything that actually fitted. SEED is a natural.

[Fidcal]

Hooray SVN is workin again. So when can I nab your new heather model Tels? It is presumably going to be committed? What was it called again? Catalan or something. I want to plant some in my garden.

[ocn]

I think SEED is a great choice and natural.

[nbohr1more]

This wouldn't be a universal thing so I'm not so sure how practical it is but when I looked at your swamp map I notice that the plants extend in detail all the way till they are barely distinguishable from within a mass of plant matter on the horizon (which looks awesome I must say ).

 

 

It struck me that if you have a homogeneous or random plant distribution you could just have a decal that looks like that random mass of plants always hover x-distance away and have the other plants emerge from it. Then, if you actually get close to the end of the plants, you could replace the patch with with smaller patches then remove them altogether (yeah, it would be tough to minimize pop-in and Doom 3's blasted lack of fade-in doesn't help but still... ).

 

(Yeah, probably too hacky but I thought I would throw it out there.)

[demagogue]

Any other opinions?

 

I was going to say PEG, Procedural Entity Generator (or Procedural Environment Generator).

[Tels]

Hooray SVN is workin again. So when can I nab your new heather model Tels? It is presumably going to be committed? What was it called again? Catalan or something. I want to plant some in my garden.

 

Calluna Vulgaris http://en.wikipedia.org/wiki/Calluna

 

I was actually trying to make an "erica" plant, but the plate had Calluna as first and I startd with it. Now seeing how they are barely indistinguishable, I might not bother with it

 

"found widely in Europe and Asia Minor on acidic soils in open sunny situations and in moderate shade. It is the dominant plant in most heathland and moorland in Europe, and in some bog vegetation and acidic pine and oak woodland. "

 

Might still make more bushy plants, as this might not gfrow in damp conditions, or in deep shade.

 

Will check it in today.

 

Edit: It is in.

 

@nbohr1more: What do you mean by "not an universal thing"? The name or the function?

[nbohr1more]

Sorry.

 

I was referring to having a floating patch mimic a distant mass of vegetation (eg: my crazy idea). Since there would be so many different possible variations it wouldn't be practical to make every possible mass to match every possible configuration of plants but having a couple standard "masses" available might be another (easy) performance option.

 

I like the name SEED very much but I did see a "seed" spawnarg:

 

seed

Default: 0. The random seed. If not set or set to 0, a random value will be used (this is the recommended way, as it generates different results every time you start the map anew, but still keeps the entities when you f.i. save and then load a game).

 

If set to > 0, the random number generator will always generate the same sequence, so you get still randomly distributed results, but they are always the same.

 

 

...so that may make things a little confusing.

 

I liked the name LODE anyway so what do I know

 

 

I've brought this up before but we definitely should has new TDM v1.0x feature contests to get mappers using the new stuff as soon as possible and drive awareness. This probably sounds crazy but we could have a LODE mission contest where the entrants would be aware that this is still experimental in v1.03 but have the deadline for completion be sometime (one week?) after the release of 1.04. If we posted the contest right on Moddb it would get all the Doom 3 mappers out of the wild (or other mappers who are brave enough to venture a new engine.).

[demagogue]

I like the name SEED very much but I did see a "seed" spawnarg: ...so that may make things a little confusing.

 

That looks like a spawnarg for this very system, though the description could be more explicit about that. So if it wasn't before, it's SEED now. It sort of fits since you can imagine physically strewing a mass of seeds on the ground to have an idea of what it does.

[nbohr1more]

Yep. That is from the LODE Spawnargs wiki article. If all the spawnargs use "seed" in their names it could be confusing to have this specific function named "seed". None of the spawnargs would need to be renamed to change the name of the whole system though

[Springheel]

This probably sounds crazy but we could have a LODE mission contest where the entrants would be aware that this is still experimental in v1.03 but have the deadline for completion be sometime (one week?) after the release of 1.04

 

It's probably not a great idea to encourage people to use a system that makes missions un-savable. What happens if the bug isn't fixed by 1.04? The system needs some heavy-duty beta-testing first.

[nbohr1more]

Yep. That's the crazy part.

 

I didn't know how wide-spread that problem was or whether the fix was pretty well understood:

 

1) Does it only happen on maps with extreme amounts of entities?

2) Is it intermittent and thus somewhat difficult to distinguish from other save crashes?

 

If it's completely up-in-the-air right now then that idea would be very crazy.

[Springheel]

As far as I know, the system has had no testing at all beyond what the developer (Tels) has done himself. We need people to start playing with it in order to find out how stable it is.

[Tels]

That looks like a spawnarg for this very system, though the description could be more explicit about that. So if it wasn't before, it's SEED now. It sort of fits since you can imagine physically strewing a mass of seeds on the ground to have an idea of what it does.

 

Hm, yeah, I see that I can do the rename pretty soon. Its a lot of search&replace, tho.

 

@nbohrmore1: The saving/restore crash stems usual from two places:

 

* wrong save/restore order. The code saves anint, then a float, but restores a float, then an int. (or misses restoring a variable). If they are not in the very exact order, it will crash. Easy to make mistake but hard to catch, and at least on Linux, hard to debug, you see it crash, but have no idea where. Basically only fixable by trial-and-error. The easy thing is: it crashes always, so it is repeatable.

* wrong memory management. I did muck with some internal ptr that in D3 usually point to objects created by the engine, but I created them myselves. This means it sometimes can try to free things twice (where I forgot to NULL out a ptr) or it doesn't free things or whatever. Even harder to debug.

 

I see that I can a grip on this, maybe it is easily found.

 

One more thing that needs fixing before a contest: The system uses way too much CPU when you crank up the entity amounts - things like 10000 entities take long, but 50000 take forever. And the latter can easily happen if you just mistype the density (10 instead of 1). Also, I need to make more ready-to-use entities, f.i. I forgot to add spawnargs to our grass so it doesn't grow on stone/metal - these things should be ready-to-go for mappers, to make their life easier. I just didn't get around to it and help never arrived. (So far I don't think anybody EVER has even tried the system. Must be me.. maybe my code smells funny?)

 

 

So I think a contest should not be done before v1.04.