Sauerbraten engine development

by Aardappel on 03/03/2004 05:18, 1571 messages, last message: 03/14/2008 18:53, 1701007 views, last view: 05/19/2024 12:52

This thread is for discussion of Sauerbraten coding issues / implementation ideas etc.

   Board Index   

#261: heightfields

by nieb on 09/21/2004 12:37

i was thinkin about some how implimenting a terragen or the farcry sandbox type thing into sauer so u could make random heightfields that look nice or make it so u can import heightfields from files .BMPs or something

and maybe make it so its selectable like u can do a heightfield generation on just a selection instead of the whole map

just a thought donno if anyone has mentioned this or not

reply to this message

#262: So you're thinking the full terrain triumvirate?

by Pxtl on 09/21/2004 15:22

Heighfield + map geometry + map models? On one hand it would be nice to have the full mapping toolset of other engines (combined into an in-engine mapper tool) - but I think Aard was shooting for simplicity of design here, sticking to a single geometry principle (hell, just use invisible sauer octrees for the bounding boxes of map models).

Plus, landscape isn't scaleable the way octrees are. Unless you want to make quadtree landscape (which could be an option). Octrees give the mapper the abliity to apply very high details to some portions of the map and very low detail to others, and have the engine reflect that detail level. Landscape is a fixed detail level of information (though LOD makes it scale with rendering). If you wanted to make a mega-map, you'd have to have very low-res landscape or you'd use up too much memory. That's not as much of a problem in Sauer's octrees.

Still, for the sake of completeness, it would be cool to have.

reply to this message

#263: Angular resolution

by ajb on 09/27/2004 01:30

I was wondering how precisely Sauerbraten's cubes can approximate arbitrary surface geometry. Since edge spans have a resolution of 1/8th of an edge, the angular resolution is fairly coarse: steps of roughly 5 degrees.

In principle a span scale running from 0 to some number N other than 8 would work as well right? Must N be a power of two? If so, why? Why was 8 chosen?

reply to this message

#264: Quick patch for y'all

by Peter on 10/03/2004 20:21

tools.h line 22:

#include <new.h>

becomes

#include <new>

This fixes a large number of deprecation warnings when compiling with GCC 3.3.3

reply to this message

#265: ..

by shahar2k on 10/13/2004 19:20

any progress?

reply to this message

#266: Re: ..

by Gilt on 10/16/2004 00:55, refers to #265

no not really. haven't made time yet. but I'll try to get some work done on the weekend.

edge spans: yeah it could've been pretty much any size. 8 is a nice round number though, and it makes it easy to line up neighbouring edges of cubes. Also, the higher the number, the more you have to scroll the wheel. anyway, the engine isn't aimed at producing doom 3 levels of detail so higher levels aren't really needed.

reply to this message

#267: Re: ..

by Aardappel_ on 10/16/2004 19:37, refers to #266

well, geometrically its nice to have it be a power of two, and not too high for scrolling through. 0-8 is 9 values thus takes up 4 bits, so an entire edge fits in a byte. If you'd do 0-16, it would already cost twice as much space... 0..8 is pretty much ideal. If you need more granularity than that, you can always split cubes up.

reply to this message

#268: Saur on X

by Louis! on 10/17/2004 03:12

I'd love to play around with a little mapping in Saur on osX.. is there am X build floatin' around? The concepts used in Saur and Cube are really meat.. They seem kind of outdated in a sense.. but on the flip side they are wicked for performance! All vertex lighting! ANY card can do THAT! ^_^

reply to this message

#269: ..

by staffy2000 on 10/17/2004 06:41

a Sauerbraten powered rts game would be awesome. :)

I am sure it could create terrain like this:
http://www.pandemicstudios.com/games/dr2/Images/shot_jungle_001.jpg

The A.I, scripting, models and radar would not be overly fun to do though.

what do you think?

reply to this message

#270: staffy -

by shahar2k on 10/17/2004 11:27

good choice of game (I love pandemic) but the terrain in that picture is a height map... sauerbraten could technically terrain JUST as good and even better (caves, overhandgs, so on) but at the same time it will have truble with smooth slopes because they will be required to fit on the grid... mess around with editing you'll get it

reply to this message

#271: ..

by shahar2k on 10/17/2004 11:27

Personally I'd like to see a sauerbraten TBS :) the grid makes it convenient

reply to this message

#272: Re: ..

by ajb on 10/17/2004 18:28, refers to #267

> geometrically its nice to have it be a power of two

OK, I can see how it allows for a better matchup between different size cubes: more of the stops along an edge will match those of larger-sized cubes.

> If you need more granularity than that, you can always split cubes up.

Positional, yes. But better angular resolution cannot be gotten that way. And the current angular resolution is coarse enough to be quite noticeable.

> If you'd do 0-16, it would already cost twice as much space...

Well, if stored as packed bit sequences it would cost 2 times 5 bits per edge span. But indeed, it would be desirable to stay within a single byte for an edge span. Actually, that should be possible since there is an additional constraint: the begin of the span is less than or equal to the end of the span. That gives a total of 153 valid edge spans for 0-16. These pack into a byte by making the byte an index into two lookup tables that return the corresponding begin and end stops. The tables could even be ordered such that they are backwards compatible with 0-8 geometry.

reply to this message

#273: *

by ^gilt on 10/21/2004 15:51

so yeah, I did do some work on it on the weekend, and everything is fairly smooth now. The player truly is an ellipse now (the older version felt more rectangular), and the response is nice. But I still have to work out some kinks (ex: since you're elliptical, when you bump into a wall that is shorter then half your height, it makes a half-hearted attempt to crawl up it...).

reply to this message

#274: ???

by Pxtl on 10/21/2004 17:23

Why an ellipse? I'm afraid I don't understand. Oh, and how rotatable is this ellipse (eg. flight sim banking?)

reply to this message

#275: Re: ???

by ajb on 10/21/2004 23:23, refers to #274

> Why an ellipse?

Gods create in their own image: it's because of Gilt's girth.

reply to this message

#276: ..

by staffy2000 on 10/22/2004 08:10

:)

An eclipse should do the job well. Just as long as it is fairy skinny.

reply to this message

#277: ..

by ^gilt on 10/22/2004 14:34

it makes movement smoother when hitting corners. it's a similar effect to using clipping planes around walls that jut-out slightly, so you don't get caught on them.

they're not rotatable, but they can be stretched horizontally, and vertically.

reply to this message

#278: quick tips on CVS?

by Pxtl on 10/23/2004 03:14

I'm using Tortoise CVS. First tried straight-out, but that gave me a login error, so then I logged in from the command line. Now I'm trying to update my copy of the Sauer source that I got before I wiped my disk, and it just isn't happening. TortoiseCVS just hangs with the following text:

In D:\My Documents\code\sauerbraten: C:\Program Files\TortoiseCVS\cvs.exe update -d -P
CVSROOT=:pserver:anonymous@cvs.sourceforge.net:/cvsroot/sauerbraten

reply to this message

#279: Re: ..

by Pxtl on 10/23/2004 03:16, refers to #277

Oh, and re:ellipses. You realise that now that I'm visualising elliptical bounding boxes, I'm thinking of a Zeppeline combat mod don't you? Or are they only stretchable in both horizontal axes at once?

reply to this message

#280: Re: quick tips on CVS?

by Pxtl on 10/23/2004 05:13, refers to #278

Oh, nevermind - got newest version of Tortoise and its good now.

reply to this message

   Board Index