The Implementation of LODS

[Rolf Dew]

Hi guys!

I would like to ask a question concerning engine features. How hard would implementing LODS (level of detail) be? Would it require changes to the engine itself or can it be added in another manner? Do you have a rough estimate of the amount of work and time that would be required to implement it? Finally would it help increase the performance of the game? (I know most of the lag is caused by the pathfinder and ai).  

Edited November 1, 2018 by Rolf Dew

[Stan`]

26 minutes ago, Rolf Dew said:

How hard would implementing LODS (level of detail) be?

Well it depends on who is going to do it. There are a few things to take into account.

One code. There is no code that switches objects depending on graphic options.

Two actors. It's still code but it will have to be done by artists by either adding model variants to every actor files.

Three models. In general LODS are made manually by artists. That's a time consuming process. The only automatic LoD system I saw was taleworlds.

30 minutes ago, Rolf Dew said:

Do you have a rough estimate of the amount of work and time that would be required to implement it?

A few years I guess assuming we start now.

31 minutes ago, Rolf Dew said:

? Finally would it help increase the performance of the game? (I know most of the lag is caused by the pathfinder and ai).  

It would only help on older computers. I thought of making a low polygon mode with the old units old horse old camels etc but it's a lot of work.

[asterix]

28 minutes ago, Rolf Dew said:

Hi guys!

I would like to ask a question concerning engine features. How hard would implementing LODS (level of detail) be? Would it require changes to the engine itself or can it be added in another manner? Do you have a rough estimate of the amount of work and time that would be required to implement it? Finally would it help increase the performance of the game? (I know most of the lag is caused by the pathfinder and ai).  

moderately difficult (multiplatform game) 

   and look at other posts from this guy and @vladislavbelov I wanted previously to use the system for LODS from supertuxkart engine that with small modifications could be applied.

[Rolf Dew]

I was mostly asking, because perhaps if the code was in place then it can be used for the Hyrule Conquest mod as I believe that they have low and high detail models. I understand that maintaining features that the Base game doesn't use can be burdensome though. Yes I agree for 0 AD it would be kind of pointless and the base game is not to graphic heavy to run. Although some mods have very detailed units and that can cause a slowdown. Although I just thought I should ask, but it is not a request or anything.

I will take a look at that thread thanks.

Edited November 1, 2018 by Rolf Dew

[Guest]

LOD would have a non-negligible impact on performance. At max zoom, there can be several hundreds of models on screen which could add up to thousands of polys. Even though performance is most heavy in pathing, making rendering cheaper would help as well. Less time spent on one thing would give more time to spend on something else.

[Stan`]

1 hour ago, (-_-) said:

LOD would have a non-negligible impact on performance. At max zoom, there can be several hundreds of models on screen which could add up to thousands of polys. Even though performance is most heavy in pathing, making rendering cheaper would help as well. Less time spent on one thing would give more time to spend on something else.

I can't remember but I think the rendering is like 20% of the frame time time. Pathfinder goes way higher than that. Doing instancing might help more than LODS.

I was thinking of stripping every prop from the structures so that multiple meshes can be reused all over.

[vladislavbelov]

As Stan said, for most lagging moments the rendering doesn't spend much time in comparison with the pathfinder. LOD can be supported on the engine side, but the instancing already has prototype. LOD is best when all models are different/some models are highpoly, Instancing is best when all models are the same. I.e. LOD would be better for buildings like CC or Wonder with many polygons. instancing - for many similar trees. But again quoting Stan, LOD requires an artist work.

21 minutes ago, stanislas69 said:

I was thinking of stripping every prop from the structures so that multiple meshes can be reused all over.

That's the good idea, reusing may save memory and performance.

[Stan`]

3 minutes ago, vladislavbelov said:

As Stan said, for most lagging moments the rendering doesn't spend much time in comparison with the pathfinder. LOD can be supported on the engine side, but the instancing already has prototype. LOD is best when all models are different/some models are highpoly, Instancing is best when all models are the same. I.e. LOD would be better for buildings like CC or Wonder with many polygons. instancing - for many similar trees. But again quoting Stan, LOD requires an artist work.

That's the good idea, reusing may save memory and performance.

That's gonna be a pain, but that will be good for artist work. Cleaning up will be nice. Some models might get some fixes as well I count on @Alexandermb to make some anims, and I have a new boar ready, and enrique's fox to commit.