Daimonin TileEngine V2

[polyveg]

moin,

coming soon... tileEngineV2
more detais but faster than the old one.
here is a screenshot from the early techdemo.
http://img216.imageshack.us/my.php?imag ... d01sk2.png

- 2 tiles in front of the hero you can see the blending of the 2 ground textures.
- The dark triangle on the tiles will be the terrain shadow.
- 16 groundtile-textures and 256 blending-filters can be used (for every mapset).
- the fps its on linux, ti4200, 1800+ using a 21*21 tiles map.

I am very proud of the framerate. So don't dare to say you have seen something faster than that in an ogre project

[SiENcE]

http://www.iris2.de/index.php/Image:200 ... 39_187.png

our tileengine uses:
- 16 blendings
- texture atlas with currently 9 different terraintypes
- fixed function (without shaders)
- tilemapsize 768*8x512*8=25165824 tiles (but can be unlimited)
- screenshot was taken with an ati x1600pro at 800x600 + 2xAA (with 36864 groundmap tiles)

but i like the shadowmap of your tileengine

[polyveg]

Hi sience,

ok, i accept the battle. Its both gpl, lets see who will steal the first line of code...
- with 16 blendings you mean 16 filters for the blending? we have 256 (mirrorable on x/z = 256*4)
- you are using 9 terraintypes - we 16 (the only limit is the gfxmem here)
- we both use no shader
- tilemapsize can be unlimited here too. texture blending is so or so only done in the near range (atm 21*21 tiles). that means fps depends only on the polycount.
- afair x1600pro == gf6800gt its a bit hard to compare this with a gf4200

I like your work, but its not really playable on my ti4200 system.
your turn

[michtoen]

And i am the server coder which
- will only full stream the visible map area
- only stream by moving the parts which comes in new
- let the client buffer & show the already transmitted parts

and you must collect and show that by yourself... No client stored map data.

I marry the winner...

[SiENcE]

hi .

hopefully we both can optimize our engines.
It's good to hear you make progress on your daimonin 3d client.
With blendings i mean unique blendings (without rotations).

http://zwischenwelt.org/trac/iris/brows ... n.hard.png

We draw this using multitexturing in one pass. Here is the material:
http://zwischenwelt.org/trac/iris/brows ... n.material

The only limit are the multitexturing capabilities to the graphiccard. Currently we only use 4 units. Thats ok i think.

Do you work on precalc. shadowmapping for the map?

[polyveg]

you have given up very fast

I also use 4 texture units, but at the moment it looks like some ati cards needs to get it in 2 passes.
I cant precalc the shadow, there is no time for it.
i will use texture unit 4 to draw it in realtime. my artist told me that it will look realistic enough when he can have half of the filters.
texture unit 4 will be used as "shadowmap", "lightmap" and grid.

when i look at my free time, i think i will need 1 or 2 month for the first demo release.

btw, I will setup the tileengine without daimonin dependencies and throw out the environment stuff like walls (atm, i am not sure if grass, small rocks, etc will come in here or will be renderered by an environment class). so it will be very easy to test it in other projects.

[SiENcE]

@michtoen: nice idea. i like your mapmaker and serverside map storage very much. We tested a fully dynamic worldmap and it's not playable on our environment. To many items on one block. too many lags with more than 100users. we only receive dynamics (houses aso.) and buffer them clientside.
using this depends greatly on the gameplay speed.

@polyveg: i don't gave up, we already have an working client developed with 3 people in 1,5years. As far as i can see, your first checkin of the daimoin 3d client was May 2006. We started the development 24. August 2006 . But of couse i know you not only coding on the 3dclient. You are also involved in other daimonin parts.

More interesting to me is the techniques used to build a large dynamic 3d tileenvironment. Both Systems (daimonin & uo) are basically the same and we both have the same problem/solutions.

Why ati cards ? I have only ati cards and i know no problems.

[polyveg]

well, its more a problem of spare time than too much daimonin tasks.
but back to the topic... your tileengine is too slow to run on a ti4200, mine will do the job.
its because i have invested much time in this research. so you have a working client, i have a speed optimized tileengine.

and of course, i am also interested in a brainstorming and source code pool, so i accept your offer

[SiENcE]

i can't really believe it, that our new tilerendering is so slow. it runs fine on my geforce2mx card. Also on an onboard intel chipset the terrain rendering works very well. i don't think there is much difference in polycount, batches and textures.
But nevertheless i take a look at the current status of your client .

[polyveg]

well, you can't take a look at the current client, because the new version is not in svn yet.
give me some time to finish the shadows...
but first of all i will be away for about a week.

i have tested a version of your client right after you'd declared that the shaders were removed.

cu,
polyveg

P.S: this looks to me like nobody else is interested, so we should continue this on our/your irc soon

[KungFooMasta]

I've been reading it, depends on what level of "interested" you refer to. I don't really have a use for a tile based system, but you never know if somebody might jump in and say something more useful than I have. Your call!

[MadLion]

Same for me as for KungFooMasta. I was reading this but have no use for it for now.
I also follow news on iris2 develeopment, because i like UO and like the idea of an open source client. ^^

[Lord LoriK]

There's a lot of people interested in this. I'm one, of course, but I can't add anything remotely useful to the thread, so I rather stay quiet until there's something to test.

That said, carry on. But please keep the updates comming.

[SiENcE]

We keep on updating this talk, and maybe should describe the techniques we are currently using.

[sybasesql]

well, you can't take a look at the current client, because the new version is not in svn yet.
give me some time to finish the shadows...
but first of all i will be away for about a week.

i have tested a version of your client right after you'd declared that the shaders were removed.

cu,
polyveg

P.S: this looks to me like nobody else is interested, so we should continue this on our/your irc soon

No, it's very interesting. Keep going pls!

[polyveg]

looks like i wont find much time the next days -but some people here seem to be interested- so i put a demo on the daimonin sourceforge dl page and the source into the daimonin svn.
both are a bit outdated, because my current version is optimized too much (it crashes)

- the tile-textures are from my last holiday, and i am thinking of giving them to ogre.
- as soon as i have a cleaner version, i will remove all daimonin depencies (and i am thinking of putting it under lgpl, but not sure right now).

btw, the initial tiles are coming from server, but after a move over a tileborder they are undefined.

[Lord LoriK]

and i am thinking of putting it under lgpl, but not sure right now

Why not? Too liberal or too restrictive for your liking?

[Brainshack]

What is a tileengine for?

[freegamerblog]

Taking a tile based map and making it look nice and not-so-tiled, I imagine.

[SiENcE]

Normally you don't need a TileEngine for 3D, because it's more practically to create a nontile 3d Engine. If you build a very big tilebased world, you have to handle all ressources of the Tiles.

3D is not designed to have many small objects and textures. If you look at commercial 3D Games you can see some real big Objects and some real big textures.

A TileEngine only makes sense if you:
- make an 3D-Client for a 2D Tilebased Game (as Daimonin and Iris2)
- create a small levelbased game (jump&run for example)
- want a small and easy editable levels + editor

You can minimize this huge overload of meshes and textures if you Batch it and use an texture Atlas.

One advantage of a tile Engine in comparsion to real 3D Engines is the case, that you don't need a complex collision detection, because you can check against the tiles ... if it's a wall or not. But this way you can only walk tilebased.

Thats why we added for Iris2 a interpolated walkingmode. Non Tilebased walking in a tilebased world .

here is a video: http://de.youtube.com/watch?v=1DZzDUoSQME

[sayad]

A TileEngine only makes sense if you:
- make an 3D-Client for a 2D Tilebased Game (as Daimonin and Iris2)
- create a small levelbased game (jump&run for example)
- want a small and easy editable levels + editor

Yeah thats right. Saves some RAM too.


Sayad

[polyveg]

After rewriting the tileengine again, I get this:

http://img59.imageshack.us/img59/3082/client3d02yy1.png

The filters are not very good yet, it was just a 10sec job with gimp. But I think you can get the point...

You can see a blue circle at the mouse pointer - that is an editor function:
you select 1 edge of the tile and change height or texture.
No more need to select from 100s of filters.

This version uses 4 texture units (nr 4 is reserved for fow, etc)
The horizontal lines are filtering errors. I must sacrife at least 4 of the src textures to switch it off. That means we have 28 textures in the atlas left.

P.S: I got informed that a new private msg arrived, but there wasn't any. If you send me one, just do it again.

[SiENcE]

This looks more nicer.

I hope the development goes now faster for Daimonin 3D Client.

[rogerdv]

Polyveg, what GUI system are you using?

[polyveg]

@SiENcE:
This is not about nice looking terrain, but about fast terrain rendering.
The right look comes when the engine is ready.

@rogerdv:
Its my own GUI System.