GeoControl 2 and World Machine 2

Started by reck, November 11, 2007, 07:22:26 am

Previous topic - Next topic

Sethren

I tend to agree there.

He should team up with Fairclough and throw that plug-in into Terragen. One can dream i guess.    ;D

Oshyan

Making erosion work procedurally is a *lot* harder than anyone realizes. The normal erosion processes that World Machine, Geocontrol, and other systems do operates on raster data, at a pixel level. Procedural data (as in TG2's procedural terrains) would need to be rasterized before applying such an effect. Rasterizing an entire planet to even a low resolution would be tremendously CPU and memory intensive. This is why *nobody* does it. Localized rasterizing is a possibility, but is extremely tricky to implement. You can rest assured that the reason these things are not being done isn't for lack of desire. Procedural and massive scale erosion is one of the hardest things to tackle in landscape modeling.

- Oshyan

Sethren

November 18, 2007, 01:24:57 am #17 Last Edit: November 18, 2007, 01:27:15 am by Sethren
That makes a lot more sense now. We will have to wait for faster computers then, that or some genius coding work.

cajomi

That will not help. As a "raster" developer I would say, that with faster pcs the algortihm could be more accurate, use more informations. And so, a raster erosion will always be faster.
Also: What sense does it make to apply one erosion to a complete planet? This would only be interesting, if you would develope a "total" erosion, which takes into account climatic zones, the material at different places and the vegetation (which has great influence to erosion).
Developer of GeoControl

latego

I don't see the point for planet wide erosion. Erosion is a very local process so, as soon as you are considering areas larger than, let's say, 100km, what you see is the general structure of the terrain and not the erosion effects. Just have a look at pictures taken from astronauts in low earth orbit.

As Cajomi wrote, modelling planet wide erosion would be a dreadfully complex task and then you would not be able to see the effects from long distances.

The scale factor is fundamental in computer graphics: if you are modelling the whole Enterprise, can easily omit to create Kirk's coffee cup...

Bye!!!

Cyber-Angel

November 18, 2007, 07:46:56 am #20 Last Edit: November 18, 2007, 08:20:11 am by Cyber-Angel
I am no programmer so tell me if this is not possible could you not have the rasterization divide the planet into tiles then each tile could be assigned a grid that would be the resolution of the terrain then each grid square would be broken down into points and the rasterization would average these points not only per each grid square but also across boundaries and look at its nearest neighbors looking vertically [In The Up Direction] (Look At some thing say a numeric keypad to help visualize) ( Number Combinations: 147, 258 and 369) [the down direction] (Number Combinations: 741, 852 and 963, ) [Horozontal Direction] (Number Combinations: 456 and the reverse 654) and the [diagonal direction] (Number Combinations: 159, 951 and 753 and 357) and calculate the erosion that way, how this would work with micro-triangle displacement I do not know since like I stated I am not a programmer.

_____________________________________________________

Note: Just so that people do not get confused, the numeric keypad idea is just so people have some frame of reference for how to theoretical operation of what I had I mind might search across grid boundaries: nothing complicated inferred or intended.

Regards to you.

Cyber-Angel           

latego

Quote from: Cyber-Angel on November 18, 2007, 07:46:56 am
I am no programmer so tell me if this is not possible could you not have the rasterization divide the planet into tiles then each tile....


That is the way programs like Mojoworld work: you can surely create a whole planet that way. I have seen on the internet tools for role playing games which can start from a whole galaxy down to a single star system, down to a planet down to the terrain at 1ft/square scale! If you want to dwelve into this field of generalized terrain generation, you will find something so vast that it is a hobby by itself!

My doubts start when you finalize your work for rendering: provided that when we render the image we have to "cram" everything into 1000-4000 (or a bit more) pixels on each side, we have no possibility to display anything outside (at most) 4 orders of magnitude of size.

Bye!!!

P.S.: GeoControl and WM are not competitors for TG2, they integrate it as much as the 3D modelling program you use to create meshes to render in it.


Sethren

November 18, 2007, 05:01:05 pm #22 Last Edit: November 18, 2007, 05:07:12 pm by Sethren
Quote from: latego on November 18, 2007, 07:10:19 am
I don't see the point for planet wide erosion. Erosion is a very local process so, as soon as you are considering areas larger than, let's say, 100km, what you see is the general structure of the terrain and not the erosion effects. Just have a look at pictures taken from astronauts in low earth orbit.

As Cajomi wrote, modelling planet wide erosion would be a dreadfully complex task and then you would not be able to see the effects from long distances.

The scale factor is fundamental in computer graphics: if you are modelling the whole Enterprise, can easily omit to create Kirk's coffee cup...

Bye!!!


Well, if there is no point for planet-wide erosions then why bother with planets at all. Even ancient glacial erosion is visible from high orbit. Is this not the level of realism that some of us hard-core realists want to have some day. I am not saying it will happen over night but say 10, 15, or 20 years from now when we have the more powerful hardware and evolved software to achieve such terrains then some like myself would find use for such details. So i happen to be a person that likes fine details and i am sure there are others out there like me as well, obviously. I have seen plenty of photos from low orbit and fluvial, sedimentary and all sorts of erosion is clearly visible from high mountain ranges to the deep deserts. Obviously it is a dreadfully complex task, that is why we wait until the power is there some day in the future. I will wait because in the meantime there are plenty of local-terrains to be challenged.

It is true that much of the larger terrain structures would mask smaller scale erosions to a certain degree. These larger structures are shaped by glacial erosion from eons back, then you have plate tectonics that distort and stretch the terrain which could be achieved with certain displacement and turbulence functions, not only in Terragen but in World Machine and GeoControl. I have done this already but on a local scale for now. All of these larger scale structures can be seen from high orbit easily. This is the stuff that shapes continents and our oceans.

I don't think no one here ever said that GeoControl and World Machine are competitors for Terragen 2. I myself had said they compliment Terragen's work flow.

Back in 2004 i had spoken to Fairclough about glacial planet/continent wide erosion. He said he was aiming for the idea at some point and i thought that was a great idea and it is a wonderful way of realistically shaping continents. Now does that sound crazy?

Cyber-Angel

I shouldn't think that it would be more than 10-15 years before desk top computers have the processor capability to do many of the things people want in CGI, in that timescale desktop Terra-scale computing should be available, earlier this year Intel(TM) showed an eighty-core experimental wafer which shows potentiality where things are heading for the next generation of computers. There are many challenges ahead in terms of cooling and so on but these in time can be over come.

Yes scale is important in terrain visualization, but any decent terrain visualization software would have the necessary metrics in place to deal with it; scale falls off (De cresses from a viewers perspective) relativistically proportional to distance, this should be easily handled by the Level of Detail system.

If one single method dose not work then a combination of methods might, should current methods and techniques fail then new ones should be devised that is how after all progress and advancement are made.

Regards to you.

Cyber-Angel                   

Sethren

Yeah, i think i was thinking a bit to far ahead. That 80 core sounds fantastic, this should bring the expensive processors out now way down in there price range in due time. Cooling is definately an issue here but someone will likely solve that sooner then later like you said.

LOD, yes a level of detail system would work well for this. Say you are out in orbit and you want to see large scale erosion, well the camera will only see so much of the planet details per pixel but this system can render out the large scale effects only and not the small scale as is only has to calculate what is visible from the camera's point of view. So when you zoom in closer, the LOD and the camera will just have to account for picking up the smaller scale details as they show up but again only with what is visible as the rest concerning the tiny non-visible stuff is ignored in the rendering process.

efflux

You can have a fractal that has different basis functions at different octaves of scale. Mojoworld actually has this.

Sethren

Quote from: efflux on November 18, 2007, 07:14:12 pm
You can have a fractal that has different basis functions at different octaves of scale. Mojoworld actually has this.


Huh, that is interesting.

efflux

However, I don't remember actually using it much. It think it might be quite system hungry but it's certainly something that TG2 could have in future.

Cyber-Angel

The Eighty Core I mentioned was only intended to be a proof of concept and is not intended to be taken into the development stage that would be needed to make a production chip, as it is right now todays software barely makes use of the computing power with dual and quad-core processors let alone eighty; with that many cores you are getting into the realm of parallelism which is in the super-computing domain.

At the Terra-Scale for desk-top computing Applications (T.D.C.A.) there would be a para dine shift in the desktop computing space proving there is also the allied memory increase (My preference would be memory in the Petra-Byte range) then theoretically with the right pedigrees you could take Terrain Visualization to places that have not yet been realized (3D Animations of Two Ice Bergs Colliding with chunks of ice braking off automatically, with real time calculation of mass, kinetic energy and other forces by way of example).

This new age of computing will herald new possibilities that aren't on anyones horizon yet.

Regards to you.

Cyber-Angel             

Sethren

I think i do recall hearing about that 80 core being only a proof of concept. It was in some article i had read a few weeks back. It is true most software does have to play catchup with the quad core systems out there.