Inspiring Photographs For Missions

[Dram]

the lights in the first screenshot wouldnt slow down the game to a crawl if they are spaced out enough so that they dont light up the same surface anywhere.

 

i have no idea how many polys and all btw. if you have doom3 you can try i guess, just use the doom3 chairs

[Gildoran]

I've heard that 50,000 to 70,000 triangles in a view is usually fine, but unlike most engines (e.g. previous quake engines or source), the majority of the workload isn't in triangle transformation, but rather fill-rate. So carefully planning out how you're going to set up lights is at least as important for optimizing a map as limiting on-screen triangles. A good way of thinking of it, is that a triangle is drawn at least once for each light that hits it (and if that triangle has a specular map, it's drawn twice for every light).

[bardic]

Thanks for the replys. I am going to fill a room, just as soon as I figure out where my Doom3 cds went and clear off some space.

[sparhawk]

Don't know if a triangle has to be drawn for each light, I doubt it, but it sure has to be calculated once per light.

[grim]

I watched a movie recently that reminded me of thief: Lemony Snickets.

It had nothing to do with the story itself, but rather the colours used. I thought that scenes from that movie would lend themselves rather well to a thief map. There are also a lot af anachronistic elements in the movie.

Anyway, just a little idea of mine that I think is worth sharing.

[sparhawk]

I also said the same way back, when I saw a trailer of this movie. The scenery would fit quite well to a Thief style map.

[sxotty]

I started screwing around on the castle in the pic, and if it gets good enough I will send youguys a copy maybe you can use it somewhere, or if someone makes the terrain to sit on it will look nice. All I have that I really like now though is the stairs in the guard towers (the ones along the wall) I finally got everything to fit just right so they can be stacked.

[OrbWeaver]

Don't know if a triangle has to be drawn for each light, I doubt it, but it sure has to be calculated once per light.

 

From my limited understanding of OpenGL, the difference between drawing and calculating is whether the draw-to-frame flag is enabled or not.

[Gildoran]

Er, by "drawn once for each light", I'm refering to doing a (fill-rate draining) render pass where you calculate the lighting for each pixel for each light, not triangle transformation.

[OrbWeaver]

Er, by "drawn once for each light", I'm refering to doing a (fill-rate draining) render pass where you calculate the lighting for each pixel for each light, not triangle transformation.

 

That's what I was referring to as well

[bardic]

Well, I finally have the game loaded and the editor working. Man, this is a bit different from T3ed.

 

2000 chairs in a giant room with 1 giant light (since there isn't an ambient light setting) gave me 7fps, which is a little better than the 5fps I get from T3ed. No idea how chairs in here compare to chairs in T3ed poly wise but I think lighting in here will lend itself to huge outdoor missions better. I had a little worse performance with 36 individual lights because I allowed the to overlap a bit.

 

All in all, kinda fun.

[sparhawk]

Can't you load the chairs in a 3D app? They usually can give you statistics like Polycount, vertices, and so on.

 

2000 is not so bad though. Which ones did you use? Are the chairs fully modelled or defined with normal maps?

[Dram]

i think he used the vanilla d3 ones

[bardic]

Yeah, it's a rolling chair movable_chair5.

[sparhawk]

I think using static meshes might have better performance, but I can be wrong. Maybe it doesn't matter as long as they don't move.

[OrbWeaver]

It depends if the giant light is a shadowcaster or not. 2000 chair-shaped shadow volumes would slow performance quite a bit.

[bardic]

It was just the default right-click light, so no shadows.

 

My light was 1000x1200x200 units, it looked OK until I changed the texture to a rotating yellow alarm type. I'll work on this more and get some screenshots of some stuff. I'm also taking into account that a big outdoor mission will have more polys used in terrain, and plants will probably have less complex meshes than chairs.

[OrbWeaver]

The default light does cast shadows, unless you uncheck "Cast shadows" in the light editor or select a texture that prevents shadows such as one of the ambientlight ones.

[Ombrenuit]

For the first picture, does it really have to be a dynamic lights? After all, couldn't they be static lights like in the thief games?

[Dram]

unfortunately doom3 does not have lightmaps ala thief 1/2. Dunno why not but it just does'nt. you can fake it though by making lighttextures to suit. This looks better then lightmaps too (not blocky at edges) but takes a much longer time.

[Gildoran]

All lights in Doom3 are dynamic. However, if the light doesn't move, Doom3 can make certain optimizations for shadows of non-moving geometry.

 

Edit: On the bright side, since the lighting is all dynamic, Doom3 maps compile VERY quickly (smaller maps can compile in only a few seconds), whereas HL maps can take hours to compile, due to calculating the lighting. It can be tedious to tweak lighting in HL because even on a smaller map, you have to wait a couple of minutes to see your changes. Whereas in Doom 3, you can see your changes in the editor in real-time as you adjust the light.

 

Another reason why you might not want to include a lightmap, is that lightmaps usually only carry information about surfaces, not 3d spaces, so although they may tell you how to light static geometry realistically, they won't tell you how to light any movable objects inside a room, so you usually need to fall back to some other type of lighting to decide how to such light objects - this often leads to inconsistant lighting, where the room may appear bright, but the objects inside it appear too dark. (I've noticed this atop the train-cars in the train map for HL1 counter-strike)

[DopeFishhh]

My favourite setting would have to be The City, the twisting of the streets and the general crowdedness and make shift nature of it all appeals to me. Breaking into mansions or museums or banks is cool n all but they tend to be the same, the street can have alot of surprises.

 

My favourite setting would be something thats Grand Theft Auto like, the free roaming nature of those games, thats why i liked T3 in some ways. Obviously that would become difficult to do in the doom 3 engine.

 

There used to be a trend to upping the polygons in a scene, surface mapping wasn't used. But this has changed due to animations and the shear effort of having to translate the polygons in real time. There have been several systems for optimising on animations but along with optimisations you have problems. Ragdoll bodies can't really be optimised much. Surface mapping techniques can optimise alot on rendering highly detailed models.

 

dynamic shadowing also takes a performance hit from abundant polygons and the majority of situations detail is just wasted over surface mapping. Considering that the silouette of the shape is what needs to be found and not everything inbetween then when you go to render if the object has alot of polygon detail that can be replaced by surface mapping then the time spent processing those polgons to determine the shadow will be wasted.

 

I believe in Quake 4 they added in lightmaps (or i guess re-added). It's almost the same engine too...

[Ombrenuit]

For interest, another picture of the castle I presented in the first post.

 

Edited November 6, 2005 by Ombrenuit

[Dram]

thats a pretty nice castle. Wish i was still in Europe