Picked Up Farcry

[New Horizon]

Edge of chaos has some really meaty shots in their media section. I'm beyond impressed. It's like a completely different game.

[Domarius]

Yes that map is huge, but a huge map with nothing in it other than some simple boxes only proves that the engine is capable of drawing shapes that large, and not that it handles such large spaces efficiently when you start putting in the rest of the game, like AI pathfinding.

 

Also don't forget the issue of needing to section off large areas of high detail with visportals so that the engine isn't rendering things unessecarily.

 

I bet the non-acellerated games from the days of the 486 could also do a map just like that one - it's pretty low detail.

 

 

At the end of the day, any city map will be more efficient if its "corridored" as much as possible, because when you get down to it, Doom 3 is a BSP engine, and they work better that way.

 

So a huge city shouldn't allow you to get onto a roof that can see over a huge portion of the city - maybe only a small house surrounded by taller buildings that don't let you see beyond them.

 

And that big tower mission you have planned Maximus, should be mostly contained inside the tower so that you can rarely see into it from the outside except for some small windows, and they shouldn't really look in onto anything too complex.

[Maximius]

And that big tower mission you have planned Maximus, should be mostly contained inside the tower so that you can rarely see into it from the outside except for some small windows, and they shouldn't really look in onto anything too complex.

 

That makes sense. I was planning on most of the action taking place inside anyways.

[sparhawk]

There is only one rule in computer optimizations. Try it. No amount of theoretical discussions will tell you if a given structure is reall faster or not. So if you want more complex outdoorscenes being viewable through the towerwindows, then just make them and see how they perform.

[Domarius]

Yes yes, sure, try it, but you can't say tips and guidelines mean nothing. The fact is BSP engines work better with closed in spaces so if you design with this in mind you will make quicker progress.

 

Even in the T2 DromEd docs they have guide lines. They tell you to turn on the view statitsics (so you can see how many polys are being rendered at once) and something like: 100 for areas where a lot of action may take place, and 300 for pretty areas where not much action may happen. Of course those numbers are based on older spec computers, but the principal is the same.

 

The "trying it" will come about when Maximus decides having a certain window look onto a complex area is REALLY REALLY nessecary for a certain part of the game, so he should try it and see what he can get away with in terms of how big the window can be and its impact on frame rate.

 

One tip they used in T2 was to make windows looking onto complex areas very very thin windows, so you couldn't see much of the area at once. Which fit in well with the medieval setting since they had a lot of thin windows like that back then.

[Domarius]

So a huge city shouldn't allow you to get onto a roof that can see over a huge portion of the city - maybe only a small house surrounded by taller buildings that don't let you see beyond them.

 

More importantly, for maximum speed benifit, there should be a "roof" over the whole city, in fact idealy it hsould be a seires of "hallways" the way the cities in T2 were designed. This is great for portalisation. Though you can never get onto the tops of the rooves though becaue they don't exist. But in my example there, you could easily have a building that is lower than the others, and climb on top of that.

[sparhawk]

Even if a highjer building is accessible, portalizing should pay off, because even then you wouldn't be able to see all of the other parts of the map, at the same time. In Enemy Territory I think it was done similarily. You could see that by using a cheat and flying up ath thei higher points. Then when you switched back to normal, you could see that, even though you ddid see to the other side, you couldn't reach it.

[Domarius]

Given varying amounts of detail, you can have lots of rooves you can climb on. But if you're plannign on cramming in a lot of polys and detail in your city, you'll probably want to make it using the "hallway" method used in "ambushed" in T2.

 

If you want to climb on a lot of rooves, you'll want to keep your detail down and make it like "Life of the party". Because even though buildings will hide other buildings, it's not quite as performance saving as only being restricted to hallways so that you can only see certain hallways, rather than nearly all of them at once. Actually even life of the party used hallways, just the hallways were very wide.

 

I spent a lot of time examining the original maps when I was trying to learn from them to make my own maps.

 

And I lost count of how many times in the DromEd documentation I saw them use the "Ambushed" map as an example of how to make maps "properlY" in the DarkEngine, they kept saying "the city is actually carved out as halways", using the same example to drive the point home.

[Maximius]

Domarius, when you say that its carved out like hallways, do you mean that literally as in you build a blank series of hallways then paint them with textures and fixtures to look like a city street?

[Fingernail]

yes, that's it exactly. Many authors don't do it that way, though.

[Maximius]

yes, that's it exactly. Many authors don't do it that way, though.

 

is there a difference between carving out hallways and using hallways to block hallways to conserve on processing power, and building discrete buildings that block the views to conserve power? (Assuming that the alternative to the hallway method is the discrete "block" method {my terminology} of building each individual structure or city block.)

 

I understand why rooftop maps demand more power, as the extended view has to be processed simultaneously, but on the ground level is there an important difference between hallway and the "block" method as long as the views are portalized or blocked by other structures?

 

While Im at it, portalization is the process of putting an invisible BSP in each doorway/window/extended viewpoint that forces the renderer to only process what would be visible through that aperature from the players POV at that moment as opposed to having to create the entire possible viewpoint through the aperature at one time. So if I am looking into room X through a doorway w/o the portal, a chair that is next to the door in room X would be rendered even though I cannot see it at that moment. But if there is a portal, the chair will not be rendered until I move into the room or until I move to a POV outside the room that allows me to see the chair. Is this correct?

[Fingernail]

In modern engines, probably not. In the Dark Engine, probably.

[TrevP]

Cry engine 2 tech demo

 

http://s2.putfile.com/videos/26010463625.avi

[Domarius]

In modern engines, probably not. In the Dark Engine, probably.

Hehe, let me throw some facts in to this sea of probabilities.

 

The Doom 3 engine and the DarkEngine are both BSP engines, so while one is aimed at a higher set of system requirements, the principals are still the same.

 

When you have more processing power available, you do more with it, you don't waste it by ignoring the fundamentals of the tools you're using.

 

Both engines have similar performance saving techniques, and to benifit from these you have to know how they work and use them properly.

 

Both of them cull areas of the map that you cannot possibly see, thereby saving potentially huge amounts of not only rendering time, but other calculations such as physics and pathfinding, that can be faked in simpler ways because you cannot see them happening. And of course the huge CPU guzzler, mesh animation.

 

DarkEngine uses portals, Doom 3 uses visportals.

 

In DarkEngine, the portalisation happens automatically, since you make everything out of pre-defined simple shapes that intersect each other. The engine records which shapes can be seen from which other shape, in a data tree. Shapes you cannot see are culled from rendering, and also any AI stuff that is happening in there - their actions are simplified. Which is why on large FMs, you may find that AI on patrol end up getting stuck and bunched up in certain places (especially Calendra's Legacy, 2nd mission) because of innacuraces introduced by the simplified programming. (Fortuantely that rarely happens)

 

In Doom 3, since you can make any sort of mesh you want (using patches etc) you are required to segment things yourself, by placing visportals. The principal is the same. Any segment you can't see into, is culled from all the intensive processing and rendering.

 

 

Now, when you make a city or mansion without carving it out as hallways, what you are doing is making the entire city one big area that cannot be broken up into segments, so EVERYTHING is being processed all the time, and no performance saving techniques kick in.

 

Eg. in the DarkEngine, if you start off with one huge air brush, and then start biulding your city buildings, or mansion inside that, the DarkEngine says "okay, which shape are you in... hm the huge air brush, that contains this and this and this..." etc. everything in the map.

 

And in Doom 3, if you build a place with air above the buildings rooves or mansion roof, you have no useful place to place visportals, since they would have to stretch up to the top of the sky, and you would be able to see them all at once, so that they are all active anyway.

 

The end result is, although ignorant mappers may not notice a frame rate drop on their high end systems, what they don't realise is they could cram MUCH more into their levels (in terms of everything - brush detail, number of AI, etc.) and it is going to chug down sooner for people with slower computers, narrowing down the target audience of the mission completelly unessecarily.

 

is there a difference between carving out hallways and using hallways to block hallways to conserve on processing power, and building discrete buildings that block the views to conserve power? (Assuming that the alternative to the hallway method is the discrete "block" method {my terminology} of building each individual structure or city block.)

Depends. In the DarkEngine - YES - your huge air block that everything is inside will ensure that EVERYTHING is always active.

In Doom 3, no, as long as you have good closed off places to segment parts of the map with visportals, its the same thing. Since in Doom 3 you start off with infinite air (the opposite to DarkEngine) you have no choice but to build this way anyway.

 

So if I am looking into room X through a doorway w/o the portal, a chair that is next to the door in room X would be rendered even though I cannot see it at that moment. But if there is a portal, the chair will not be rendered until I move into the room or until I move to a POV outside the room that allows me to see the chair. Is this correct?

Almost. Assuming we are talking about Doom 3 - If there is a visportal in the doorway of that room, then if you cannot see that doorway, that room will not be rendered. But as soon as you can see into the room, the entire room is processed as if you could see it all, including the chair. Only the Z buffering of the video card prevents you from seeing the polygons of the chair through the wall. There is a performance saving by not drawing the polygons of the chair... but there is a HUGER performance saving by not asking the video card to even think about the room in the first place.

 

That is the important part.

[Fingernail]

All of those considerations are pretty much negligible for the Dark Engine these days

 

Look at the Hammerite Imperium. Basically ALL of their maps are built as giant cubes or a series of them, fill air ones. They run fine, no problem on modern machines, but the limitations of the dark engine are the very number of brushes that it can hold in a single level, not to mention the object and light limits.

 

In fact, Doom 3 is also very limited in terms of entities per map. However, that CAN actually be changed in the SDK IIRC, so it's not so much of an issue in the future. But I think you can only have 4096 entities or something, which includes AI, meshes and lights, as well as script objects.

[sparhawk]

In fact, Doom 3 is also very limited in terms of entities per map. However, that CAN actually be changed in the SDK IIRC, so it's not so much of an issue in the future. But I think you can only have 4096 entities or something, which includes AI, meshes and lights, as well as script objects.

 

Yes. 4096 is the limit, because this is the arraysize. I think it can be changed, because it is completly defined in the SDK. An entity is a map object that can be intereacted with in some way. This doesn't need to mean the user can interact with it, but also the engine can use it for some purposes (like visportals). All the meshdata which makes up the gameworld, can not be interacted with directly.

[Ishtvan]

Just a minor point, but visportals aren't actually an entity in the gameLocal::entities array. They are stored somewhere deep within the render engine, and only accessed by specifying an area number and portal number.

[Domarius]

All of those considerations are pretty much negligible for the Dark Engine these days

 

Look at the Hammerite Imperium. Basically ALL of their maps are built as giant cubes or a series of them, fill air ones. They run fine, no problem on modern machines, but the limitations of the dark engine are the very number of brushes that it can hold in a single level, not to mention the object and light limits.

The entire reason I explained that was to show how this same principal still applies to Doom 3 and will always apply to anything you make something with, which is what I spent that entire post trying to explain. These performance saving techiques are as much a consideration to mappers as programmers choosing bounding box collision when poly collision is not needed.

 

And actually, it does still apply to DarkEngine; a lot of Thief players have low end systems since it is a low end game. It does not take much extra consideration to map appropriately, so when you design poorly, what you are saying is "We are ostricising all the people with slower systems than us because we are too lazy and ignorant to pay attention to mapping properly."

So while I can play an awesomely detailed map such as Calendra's Legacy, the Hammerite Imperium would be unplayable on my P3-450 just because the mappers were too ignorant to learn about the tools they were using.

 

 

 

To avoid this, all that is required is to consider the appropriatness of the tool you're using - sure you can use a hammer to make a hole but it will be a very messy and useless hole compared to one that you would make using a drill.

The fact that you are so strong that you can punch a hole in ANY wall without the use of a drill makes no difference. It doesn't mean it will be as useful a hole as one made by a drill.

[Fingernail]

You're just exaggerating to demonstrate your knowledge of "principles of design". It's not exactly that clear cut.

 

And it's not mapping incorrectly or anything of the sort; as a team, they're just doing more complex things than have ever been done before. Things like the T2X train level. That certainly wasn't ideally optimised, much better would have been a static landscape.

[Vadrosaul]

http://www.tweakguides.com/Farcry_1.html

 

If you have a top of the line system now, you should be able to turn on all the options, wtih full AA and AF and experience excellent framerates on the demoloop.

[Domarius]

You're just exaggerating to demonstrate your knowledge of "principles of design".

*sigh* Yes Fing, that's it. I'm only posting here boost my own ego. What ever I did to earn that level of respect from you, I sure would like to know.

 

The reason I'm posting here is because of the times I'm playing an FM and I'm in a little simple room and the frame rate is chugging like I'm looking at the whole city (because as far as the game is concerned, I am). Or when I'm being chased by a guard and suddenly he stops at the bottom of a stair case and starts cursing me because he doesn't know how to get there (because the mapper ran the "Build Pathfinding Database" on one area of the map at a time, instead of letting it process the entire map).

 

I'm sick of FMs runing slower than they should, and I'm speaking from mapping experience. I map, draw, animate, etc. the way I program - I try different techiques and compare the outputs against each other so that I KNOW for myself what works. In the case of mapping in DromEd, this means comparing the maximum viewable portals and polys. And because I have a slow computer, I can actually see the difference in frame rate. People with fast computers who don't bother with this step just go "yep, she'll be right, looks fine on mine so it must be fine."

 

And it's not mapping incorrectly or anything of the sort; as a team, they're just doing more complex things than have ever been done before.

If you're talking about a mansion or city - what exactly about it has never been done before? There are plenty of good examples of how to best do them in the original T2 missions. I'm not talking about special cases such as the train mission which obviously require totally different approaches.

 

The best maps run beautifully (like Calendra's Legacy) because the mapper knows what they are doing.

 

It's not exactly that clear cut.

No its not clear cut but how often have you heard the phrase "you got to know the rules before you can break them?"

 

In Deus Ex, I played a mission where you could get to the top of a really tall crane and see the top of every building. "Oh noes they broke my rules!", right? No. Every building was a perfect, simple box. And by no coincidence, every door into every building was hidden behind another building or on the opposite side of the building, so that from the vantage point you could not see into a single doorway - so all those portals would be deactivated.

 

This is in contrast to a complex city mission, where you cannot get onto the rooves because the streets and buildings are very detailed and the game would run to slow on the target system if they allowed that.

[Fingernail]

Oh, I didn't really mean it, I was mostly playing "Oddity's Advocate".

 

Nevertheless, you simply can't make a huge cathedral by using the tunnel method. You have to have an air brush to have the cathedral in, and then if you want to build a large cathedral, it has to be a fairly big one.

[Domarius]

Yes you can - they teach you in the DromEd docments to "build in reverse". In real life, you start off with air, and you place solid things to build a structure.

 

In DromEd, you start off with infinite solid, so you have to do the opposite. And the reason the engine works this way is because of what I described earlier - it looks at what air shape your'e standing in and works out what other things inside other air shapes you can see from that one.

 

So to build a building of any sort, including a cathedral, you place air blocks to carve out the outside area around the building till you're left with a solid in the middle - much the same way a sculptor would carve a stone to leave something in the middle.

Then you place air blocks in for doors, then continue to tunnel your way into it like a mole, using just air blocks and no solids. The end result is - the thickness of major walls should actually belong to the infinite solid, and not a solid brush that you placed there.

 

Of course you don't do every detail this way - just like anything, you rough out the main things first and get more detailed as you go. For example, if you wanted to model a piano, you wouldn't gain much performance by modelling it in reverse - just make such a small thing out of solids. The same can apply for some small rooms.

 

But if you at least do this for large areas, you'll have the performance saving of the engine not thinking about what to render in the library upstairs when you are actually in the foyer downstairs. The trick is to play test and pay attention to the stats and plan to make sure high detail areas are seperated from each other using the tunnel method.

I mean at the very least, my computer should not be thinking about rendering the museum on the other side of the map when I'm in some cave underground, sheesh.

 

 

 

But anyway this is not the way to model in Doom 3 because you start with infinite air. Doom 3 is much easier in this respect because all you need concern yourself with is having nice tight places to place visportals.

[god_is_my_goldfish]

Getting back on topic, I installed FarCry on my new lappy, and it looks amazing. I'm now able to have everything on Maximum and it still runs smooth! Images are at my resolution of 1280x800.

http://img45.imagevenue.com/img.php?loc=lo..._FarCry0000.jpg

http://img15.imagevenue.com/img.php?loc=lo..._FarCry0001.jpg

http://img104.imagevenue.com/img.php?loc=l..._FarCry0003.jpg

http://img45.imagevenue.com/img.php?loc=lo..._FarCry0004.jpg

 

EDIT: AAAAAAAARG! Screw imagevenue, I can't link to the god damn files at all now!

[Domarius]

Wow... only a few years ago you'd have to fire up a copy of Bryce and let the computer spend 10 minutes a pop to get images like that.