HI,
I found this post

https://groups.google.com/forum/?fromgroups=#!searchin/comp.graphics.algorithms/new$203d$20voxel/comp.graphics.algorithms/UqHsc-DVwO8/y17WuZE3xfAJ

and I like to see what does this guy, the webpage not have the exe anymore.
I write to her but not response.
but with wayback site a get the demo

http://web.archive.org/web/19980712165309/http://faza.com/spec.htm

But I don't know how dissassebly a dos4gw .exe.

any hint to get the assembly source of the exe?

thank's

pabloreda,

Probably you can supply that .Exe because wayback page contains links to HTML content only. Anyway, you may traverse to real LE (skip to overlay, it should start with 'BW' signature, use relative offsets at +20 to skip all of them to 'MZ', that's all).

thank's baldr

I have the exe and run ok in dosbox, it's a watcom compiled program.
I try the wdis(watcom disassembly), but not admit .exe, tell about coff or something

pabloreda,

I can't say much about wdis behavior. Maybe you can make that .Exe accessible somewhere? Then I'll try to prepare it for IDA disassembling.

thank's a lot baldr!
I do the dissasembly at last, IDA work like a charm, I have a ASM and a C!!
The reason for analyse this program is for try to guessing the Unlimited Detail Format, (the guys of euclideon), some points in common with this code.

thank's again

The video demos running directly off USB are quite impressive. Seems clear that some kind of spacial compression is designed into the algorithm. Also, Euclideon has said the data sources need to be preprocessed by them.

Here is how I imagine it: The binary of the coordinates define the partition address. This address could be masked and found in several hash tables (depending on LOD). For example, there would be only one point cloud representation of a grain of sand, but several hash entries which point to that description.

That would reduce it to a couple 100 instructions per pixel, regardless of point count.

Here is their latest video:
http://www.youtube.com/watch?v=Irf-HJ4fBls

I had to stop after 6 minutes - the narrator's voice is awfully annoying, and (up to that point, anyway) it contained nothing but their usual marketing bullshit superlatives and all-fluff-no-content hype.

Their previous "infinite detail" stuff was also in no way impressive, as it consisted of extremely repetitive data.

So, is there anything interesting and/or substantial after the 6-minute mark? To me, the company has all the marks of a "We'reReallyJustInItToGetBoughtOutAndLaughAtTheSuckers" scam.

there are a very curios thing about this subject, the amount on point are implosible on current technology. I found a similiar?? proyect from 1997

you can look in
http://chorasimilarity.wordpress.com/tag/unlimited-detail/

Yeah, I mute the videos - he has always been annoying as fuch. From what I've gathered, he already got funding for this Geoverse app and are working on some government projects? I don't know if there is any novelty to his specific algorithm, but it doesn't seem to merit reverse engineering this DOS exe when similar public methods can achieve the same?

“Once you eliminate the impossible, whatever remains, no matter how improbable, must be the truth.” Arthur Conan Doyle


Lets start with some facts extracted from S. A. Janczewski (1999):

• 4000 x 4000 x 2000 points possible
  
• concern for instruction count
  
• different color/attribute from each of 6 directions
  
• not cubical -- can have sloped surfaces
  
• six colors are not at all stored together as a unit
  
• non-coordinate granular data organization

Then we have some facts from Euclideon (2013):

• billions of points (up to 140TB)
  
• real-time dual screens
  
• preprocessed point cloud
  
• loading time is <5 sec
  

Many techniques can be avoided as they would never be fast enough. Assuming, 2^25 pixels per second means 128 cycles per pixel on single 4Ghz processor - very conservative resolution at 20 fps.

http://www.ausgamers.com/forums/general/thread.php/3260288

An impressive voxel engine:
http://youtu.be/VYfBrNOi9VM

...and some research papers:
http://maverick.inria.fr/Members/Cyril.Crassin/