[Gambini]

Diaz, on 27 October 2012 - 02:50 AM, said:

Gambini, pressing "1" on a spotlight sprite (to make it one-sided) disables shadows now. It's great because we can now project textures without killing performance, and the spotlight shape is more realistic in many cases.

When taking lighting into account for performance, you must keep in mind that each light will cause every object and map geometry affected by it to be redrawn once. So basically, if you have 5 lights affecting 20 objects and complex geometry, those objects will be drawn 100 times and the geometry will be drawn 5 times, and that's what kills performance. Maybe the mechanisms at work aren't that simple but that's it roughly. If you want to keep a reasonable framerate, you can have a large amount of lights affecting simple geometry, or fewer lights affecting more complex geometry. Lowering shadow quality also helps a big deal with spotlights.

I've made a small app for my mod that uses sliders to allow the user to change r_pr_maxlightpasses, r_pr_shadows, r_pr_shadowdetail and r_pr_maxlightpriority in the .cfg, so performance is scalable. Full-blown lighting on my maps means both complex geometry and a large amount of lights seen at once, which requires a modern computer, so I had to do that.

There also seems to be some sort of bug/memory leak in Polymer that causes performance to drain as time passes. A restartvid in the middle of a game can sometimes double performance.

Thanks for the explanation. But if you disable spotlight shadows, then the lighting doesn´t look cool anymore. I´m not an expert but I think Polymer will always have a hard time when dealing with performance, because all lighting calculations are done on the fly. There should be a prebacked light system for static objects, on map startup (well nothing is 100% static in duke, but you know what i mean).

[Kyanos]

Great timing on this subject. I've been testing with pointlights in what will someday be a massive tower. It will have four different color themes converging into a dark ominous central section. This is just some "mapster sketching" as I develop some ideas/concept.

[Fox]

Gambini, on 27 October 2012 - 04:46 AM, said:

Thanks for the explanation. But if you disable spotlight shadows, then the lighting doesn´t look cool anymore. I´m not an expert but I think Polymer will always have a hard time when dealing with performance, because all lighting calculations are done on the fly. There should be a prebacked light system for static objects, on map startup (well nothing is 100% static in duke, but you know what i mean).

If that is the case, I suppose the engine should check whenever a sprite or wall coordinates has been changed since the last tic, and only then recalculate it... but I don't really know if it is possible.

[Diaz]

Gambini, on 27 October 2012 - 04:46 AM, said:

Thanks for the explanation. But if you disable spotlight shadows, then the lighting doesn´t look cool anymore. I´m not an expert but I think Polymer will always have a hard time when dealing with performance, because all lighting calculations are done on the fly. There should be a prebacked light system for static objects, on map startup (well nothing is 100% static in duke, but you know what i mean).

Well I believe the conic shape of a spotlight looks much more realistic (see attached image). Sure it looks better when it's casting shadows, but now we can have more lights shaped like that without making performance go down the toilet Not to mention texture projections can look smashing (see the other attached image)

About static lighting... well, generating lightmaps is a slow process, and it's done at design time on every engine for a reason. It would take really long to load a map if lightmaps were baked at load time.

Attached thumbnail(s)

• 
• 

[Gambini]

Well I´m not talking about lightmaps like in Source, I´m pretty much saying the same than Fox suggested. AFAIK Polymer projects all visible lights on the affected surfaces with every frame. When it would be much faster to do it only once at map startup, and only update them once a change in the geometry has been made (example, a door that opens). And of course, it´d take the same time a frame needs to be drawn. Lightmap based engines take a lot of time to compile because there are much more calculations involved, like light bouncing, semi-darkness, ambient light, etc.

[Diaz]

Ah ok, I took it when you said "pre-baked" you were talking about lightmaps, as in Source, or Unreal's Lightmass...

I'm not sure how lights are done though. Maybe the way they are coded the objects they affect need to be redrawn once for each light every frame . I'm suspecting lights don't "project" on the scene, but rather, the scene is redrawn many times and the resulting frames are blended somehow. Only Plagman can tell us.

[Micky C]

When I added lights to Spiker's Project Zero map, I did so without regards to the framerate, and it ended up playing at about 2fps. When I brought it up on the forums, Plagman said something like it wouldn't be a problem when polymer's optimised. I'm pretty sure that the lights can be done much, much more efficiently than they are now, but Plagman did say "it's much easier to get it working right, then to get it working right and efficiently". So one day we should be able to make some pretty sweetly lit maps and not have to worry about the fps counter.

Also, suburbs. It's mostly copy paste work but I made sure theres plenty of variety in texturing and building angle, as well as some more unique buildings to take away from the same-ness. If you look at the actual Serious Sam maps, even SS3, you'll find that a lot of the content is just the same stuff pasted over... and over... and over... and over.. and is much more obvious than most other games.

[Mike Norvak]

It's just me or those houses look giant

[Fox]

Imagine a horizontal line in the middle of the screen, that would be the height of player eyes. So yeah, these are Godzilla's house.

That looks really good. That sand is too much of an empty space, but I guess you are going to add some detail and furniture there.

[Micky C]

To be fair I was looking down a bit at the time. I agree that the houses are oversized. I kinda wanted them to be a bit oversized because everything seemed so big in the original game, but I'm not sure if I've overdone it here. I guess I can easily lower the ceilings but then I'd have to vertically shrink all the textures which will be a real pain..

@Fox, yeah I'm going to add more trees, bushes and props. Everything was fairly flat in the original game so I won't bother much with height variations of the floor.

[Fox]

Why don't you use Mapster resize function? Wouldn't it work?

[Micky C]

The mapster resize function resizes everything down which I don't want. I would only want to shrink it vertically.

[Plagman]

Sounds like a job for m32script. I bet just a few lines would solve that no problem.

[Diaz]

Everything was oversized in Serious Sam anyways

[Micky C]

Discovered another little neat thing with the new light options; negative spotlight + no-shadow option = picture below.

The orange light flickers by default, now if the shade (arc angle) of the spotlight flickers at roughly the same rate (and I suppose the hitag should change as well or it could look weird), it would look very cool.

[thatguy]

Micky C, on 29 October 2012 - 04:26 AM, said:

Discovered another little neat thing with the new light options; negative spotlight + no-shadow option = picture below.

The orange light flickers by default, now if the shade (arc angle) of the spotlight flickers at roughly the same rate (and I suppose the hitag should change as well or it could look weird), it would look very cool.

Okay cool. I did that sort of thing with the Unreal Engine 1 and Morrowind Editor back in the day to accomplish some sick stuff.

[Gambini]

@Micky: Not only those houses are oversized but the roof is not convincing. Make them again with a little bit more realistic measures and make the bottom part of the roofs match the slope of their top. Since you said it´s all copy paste, it may not take more than a few minutes.

Something like this:

[Micky C]

I'm not sure what you mean, can you be more specific? I don't think there's anything wrong with the roofs, and besides, if I make it any more complicated than it is now it would probably be a lot more work and take up more walls due to how they're all constructed. I'm not too concerned if an area doesn't look all that realistic or detailed.

[Gambini]

I don´t know how to explain myself better about this. Right now your roofs look like triangles, when they should look like upside down V´s

That would take maybe one more sector per house. That isn´t too much detailing i guess.

[Micky C]

Ah, now I know what you mean. No I can't do that. Well I could, but it would become 10x more complicated, and I wouldn't be able to copy-paste, so it's not worth the effort. Right now everything's nice and simple.

[Gambini]

I spent more time resaving these images in jpg and uploading them on photobucket than the time i spent building this:





How that can be complicated? And I´m sure that the upper TROR layer only needs another slope in those triangular sectors.

[Micky C]

Ok, ignore the lines on the bottom layer, those don't need to be there (I've scribbled them out). As you can see, for the ease of copy/paste the roof doesn't actually take up the entire area, there's a thin sector surrounding it. If I were to do as you say, I'd have to manually insert points to all the buildings, then create 4 new TROR scissor sectors for each building, and I think I'd have to do them all by manually as well. It will more than double the current number of sectors and walls for each building, and I have to take performance into account as well. Several areas are quite large and can potentially have a lot of enemies so I need to keep things simple to boost the framerate anyway.

[Mark]

I cheated and just created a roof model and put it on top of each slope.

[Diaz]

^ That is the most efficient way of doing it.

[MetHy]

MetHy, on 26 October 2012 - 06:53 AM, said:

Some of you might remember this. It's the first thing I worked on in 2005 when getting back to Duke Nukem 3D after discovering AMC.

I just realized it's Halloween next week so I got back onto it and I'm gonna try to finish it by then.

*foggy forest image*

It definitely WON'T be done for tomorrow. I'm not even half-way through the map.
I thought it would be a quickly made map by putting together two old maps I started and linking them, but the project keeps getting more ambitious every day. I keep getting new ideas as I'm mapping, but as long as the inspiration and motivation to build is there I won't complain.

The wait will be worth it though : it will be unlike any other usermap you've ever played. You will either love it or hate it!

[Cage]

Fog in software mode opens some cool possibilities!

[Loke]

Is that a mixture of Redneck Rampage and Half-Life textures? Looks very cool and eerie.

[gibfrag]

Working on an enemy,

If said enemy was wearing a uniform similar to the one in the image below, do you think that when killed there should be gore in the death animation, or do you think it should have NO gore, but when it reaches its final frame (when it would be laying on ground) a puddle of blood appears (like when enemies die they spawn that blood pool from time to time)?

Spoiler

[Jimmy]

It's only leather. A little bit of blood would appear natural to me.

[Mark]

I guess it might depend a little on what type of weapon he is being hit with.