Building a Rlaan base...

[Ares]

No picture yet but I'll start this with a question. i was going to do the asteroid rlaan mining base just to have something for the game since i figure respectively it would be cheap to produce and the most common. I was also considering a standard-esque military installation type thing what do people suggest?

EDIT
Made a big ol' boulder
http://vegastrike.sourceforge.net/galle ... alNumber=1
my wierd little extrusion on the bottom looks kinda.. wierd.... should i re-do it?

[Fendorin]

Hello Ares

If you want a idea because Rlaan is insect-like why not try to render the Rlaan base on asteroid like the socialized insect-house:

See below for inspiration:

http://en.wikipedia.org/wiki/Termites
http://fr.wikipedia.org/wiki/Image:Ant_mound.jpg
http://fr.wikipedia.org/wiki/Termitiere

check yourself i think Rlaan could have a construction like termite : fine and large layer

[Ares]

I LIKE that idea gigantic rock termite-esque mound it is!

[Ares]

SO whats considered too high-poly for a base? my NEW asteroid itself consumes 600+ and i'd like to aim for the 10K range but i know VS is rendering alot now...
And stupid question #2, VS renders in poly's right? so 10K quads would be 20K polies.

[Fendorin]

Hello

About polly i m not sure but i see on a externe tutorial when you make object for game you should think about the balance between good visual aspect and the performance for computer ( my english is so bad for explain you better than that , check yourself on web for some recommandation)

but for exemple this ship :http://vegastrike.sourceforge.net/forum ... hp?t=11884

of course is ULN ship normally "Boxy" but is under 500 poly after be triangulate

begin by more simple staff than a massive asteroid

and remenber more you have poly more is difficult to unwrap and texturing

A lot of contributor propose a wings "model without texturing" and a lot of this proposition never appear with texturing after a very long time.......

thank

[chuck_starchaser]

The new word from Klauss is that there's a limit of 64K vertices per object (per mesh), beyond which point there's a serious degradation of performance. If a station needs a lot more vertices, one should export it as a bunch of separate .obj files, none exceeding 64k vertices. Might even be a good idea to keep the number of vertices per mesh below 32k, in case some older videocards had one less bit than current ones for vertex indexes.

[Ares]

so verticies is the important number? ok them i'm nowere near going over...
Always though polies was what you had to watch...

[klauss]

Indices, not polies or vertices. Indices are kind of a hybrid. For pure triangles, indices are 3 x triangles. For triangle strips and fans, indices are 2+triangles. For quads, indices are 4 x quads.

Clearly it's complicated, so you're safest if you assume there are no triangle strips and limit 3 x triangles to below 64k or 32k.

And I believe the limit is more related to the vertex cache or something like that more than the number of bits anywhere. In fact, the limit is queryable through the OpenGL API, and some docs state it to usually be 4096 (4k indices), but that number seems to no longer apply (ie: the GF6 series has that limit at 64k).

[chuck_starchaser]

Gottcha, so a super-safe-side rule of thumb would be 10k triangles per (sub-) mesh? Just curious: Where is the mesh "stripified"? Is it on load, like by the driver or GPU?, or does mesher do some kind of stripification? And is there a way to put stripification "hints" in a mesh, short of splitting vertices?

[klauss]

It's a good question. I had coded stripification in the obj->bfxm module, but I believe I left it disabled because, lately, a well organized triangle list is better than a series of triangle strips (because of API call overhead - each triangle strip has to be drawn with a separate OpenGL command).

So, in the end, there are NO triangle strips if you import from an obj.

But I'll have to confirm that. And, independently, in the few cases where stripification reduces the index list back into performance levels (<32k), it would perhaps be preferrable. So I guess a commandline to enable it (where the default is having it disabled) would be good.

Otherwise, let me assure you, VS does NO stripification of any kind.

[loki1950]

BTW klauss there has been a bite on the maintainer for mesher thread he states that he not that comfortable with Python but pretty solid on C++.

Enjoy the Choice

[klauss]

BTW klauss there has been a bite on the maintainer for mesher thread he states that he not that comfortable with Python but pretty solid on C++.

Enjoy the Choice

Well, mesher is C++ (the ++ used loosely)

[loki1950]

But the current wee GUI is Python if i get it to work for me got to get nvcompress first as it also does the DTX compression before it exits your bfxm.

Enjoy the Choice

[Dunkelheit]

Now that you talk about vertices and related, I dont understand why there are ships in VS that have a lot of dobled vertices, is not this more work for the GPU?. for example I loaded the Taizong in Blender, and saw that it has almost 3000 doubled vetices!!, and when I removed them the model in fact shade better, rouded things are really rounded, is there some reason for this? why not to fix ships like that?



___________________


And this ideas for Rlaan are good for me too, because Im working on Rlaan planet, and talking about that where is Jacks?, I pm'ed him a weeks ago with the pre rendering pictures for his aproval, and at this time have no reply.

[loki1950]

for example I loaded the Taizong in Blender, and saw that it has almost 3000 doubled vetices!!,

Looks like we need a model clean-up might even speed the game up any takers for a survey of our existing assets so we can at least have a to do list.

Enjoy the Choice

[chuck_starchaser]

Now that you talk about vertices and related, I dont understand why there are ships in VS that have a lot of dobled vertices, is not this more work for the GPU?. for example I loaded the Taizong in Blender, and saw that it has almost 3000 doubled vetices!!, and when I removed them the model in fact shade better, rouded things are really rounded, is there some reason for this? why not to fix ships like that?

Vertices should be doubled along sharp edges separating smooth groups. But if things that should be rounded were not looking rounded, that probably means that the modeler was using flat shading; --probably wasn't even aware of smooth shading. That's a... disaster. Then again, it wouldn't surprise me. I saw many a horror when working with the mining base.

Klauss, I was asking about stripification because the author of a mesh optimization tutorial I read recommended to avoid fans. He said that GPU's only support triangle strips and quad strips, in the hardware, but not fans. So he recommended to NOT cap the ends of tubes the way most 3D tools do, like on the left of the pic below, but with parallel quads, as on the right.



But that should be easy for the stripifier to do corectly, because the cap is a separate smooth group, so the vertices are split.
But what if I'm doing a radar dish, where the whole thing is one smoothing group? Would the stripifier correctly go around, then do the center like a vertical strip? Or will it go horizontally and do 5 strips of only 3 quads, and then do two more single quads?

[klauss]

Honestly, I don't know where he got that. Modern vertex caches can handle triangle fans IIRC, perhaps he's talking about really old hardware. But I dunno.

Anyway, as chuck said, hardness requires doubled vertices, but if smooth edges are using doubled vertices then we need to clean them up.

Anyway, a little ago I made VS automatically optimize meshes on load, removing unnecessary doubled vertices.

[chuck_starchaser]

Based on angle? That's NOT right, Klauss. That's punishing good modelers to reward careless modelers. I have a cow with Blender, because it has a fixed limit of 80 degrees for smooth rendering. That's a KILLER. It means that cubes with solid bevels show sharp in renders, as well as in AO bakes. That's absurd!
Solid bevels are the best trick in modeling, to get nice, round edges.
I use solid bevels at even greater than 90 degree angles. If they are small, they look great, anyways.
At the opposite extreme, there are gazillions of cases when you want almost imperceptibly angled surfaces to show a crease.
Just because some modelers are sloppy we shouldn't castrate all modelers.

EDIT:
Besides, every 3D modeling tool has such basic functions as to automatically crease things for you based on the angle, so, to have it in mesher would be a duplication of an existing functionality... and without the option to turn it off?!?! That would be a disaster.
What you could do, if you want, is first to count how many vertices are split. If over half are, then you can be sure they are split by negligence, and use a default algorithm.

[Ares]

I was just looking for a number. lol. so much information.

[klauss]

Based on angle? That's NOT right, Klauss.

Now why would you assume that?
Based on nothing. Truly truly doubled vertices, same normal, same position, same tangent, binary equality.
You'd probably think that would never happen, but it soooo does. The ox, for instance, had like three times the number of vertices before optimization. That's a lot. I grant you, it would be unnecessary if modellers were careful, but thinking would be unnecessary if computers did it for us - wishful thinking man.

[chuck_starchaser]

Ahhh, gotcha! Sorry. You once talked about an idea to have mesher determine creasing on angle, and I thought you were still thinking about it.
Yeah, that would be awsome!
For another idea, mesher could also make sure there are no edges shared by more than three polygons, and print out an error message if there are.
In Blender you can check for that with Select -> Non-Manifold, but you have to remember to do it; and it's a type of mesh corruption that's very easy to incur inadvertently.