Jump to content

  • Log In with Google      Sign In   
  • Create Account

Banner advertising on our site currently available from just $5!


1. Learn about the promo. 2. Sign up for GDNet+. 3. Set up your advert!


Beware if you have purchased GPU Gems 2


Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.

  • You cannot reply to this topic
15 replies to this topic

#1 jasjas   Members   -  Reputation: 262

Posted 20 December 2005 - 08:44 AM

Since I want to make an open-source terrain-rendering engine, I began to implement the Geometry Clipmaps technique shown in Chapter 2 of GPU Gems 2. A few months later (I worked on it only 1/2 hour a day) I got everything working nicely except for the frustum culling. To figure out how to do this, I went to Google. When the search results returned, I was in for a rude shock. It appears that the author of that chapter has been granted a patent on geometry clipmaps! This patent has been assigned to Microsoft. So now, I'm not sure what to do. I can rip out all of the clipmap code and replace it with something else (I'm leaning towards SOAR at the momemt), or take my chances and release it. IMO, it's a crappy thing to do when a company submits a paper to a book knowing full well that it has applied for a patent on the technique. This considerably lessens the usefulness of the GPU Gems series; who knows what other techniques have been patented by the authors? With this in mind, I will not purchase any more books in the series, assuming GPU Gems 3+ comes out. Comments, suggestions? -- jas, libnoise project admin [Edited by - jasjas on December 20, 2005 4:05:52 PM]

Sponsor:

#2 Witchcraven   Members   -  Reputation: 564

Posted 20 December 2005 - 08:48 AM

That is lame. I would just jack their patent and run with it. Life is too short to rewrite renderers. Or you could ask their permission to use it.

#3 Khaos Dragon   Members   -  Reputation: 196

Posted 20 December 2005 - 08:48 AM

It may just be a defensive patent, you should consider yourself lucky if your project is successful enough to garner microsoft's attention. In that case, you could base your project overseas to avoid the US patent issues, or are you exempt from them in Canada?

#4 Yann L   Moderators   -  Reputation: 1802

Posted 20 December 2005 - 08:52 AM

AFAIK, Canadian law allows software patents, but only under very specific circumstances. And I think they have to be specifically patented in Canada, a US patent is not valid. IANAL, but that's what I seem to remember.

And while I'm not really familiar with the clipmap patent, it might be a defensive one. Nowadays, it becomes more and more important to simply patent everything you invent, just to be sure that noone is going to sue you afterwards. That's especially true in the software industry (and in countries that allow software patents, of course). And Microsoft is well known to adhere to this principle.


#5 Trapper Zoid   Crossbones+   -  Reputation: 1370

Posted 20 December 2005 - 09:02 AM

I agree that it probably is a defensive patent. Looking through the literature on Google, it seems the patent was filed a few months before they published a paper in SIGGRAPH '04.

Since I don't know much about geometry clipmaps, I don't know what part of the technique is unique (and I don't have the time to study the paper in detail to find out [smile]), but as long as you avoid that part then technically you should be fine; reading through the references of the SIGGRAPH paper there's heaps of background material that they've based it on. Of course, if Microsoft decides to send the lawyers after you anyway then you'll be in trouble no matter what you do...

#6 caffeineaddict   Members   -  Reputation: 633

Posted 20 December 2005 - 09:29 AM

Even so, if the article was out for public consumption on implementing this technique before the patent was issued, wouldn't someone implementing it from this article be within their rights since it was already out in the open for that purpose? Sucks either way though.

#7 jasjas   Members   -  Reputation: 262

Posted 20 December 2005 - 09:32 AM

Quote:
Original post by Trapper Zoid
Since I don't know much about geometry clipmaps, I don't know what part of the technique is unique (and I don't have the time to study the paper in detail to find out [smile]), but as long as you avoid that part then technically you should be fine...


Hi Trapper Zoid, the gist of his claims are:
1: a number of nested regular grids (clipmaps) surrounding the viewer,
2: the contents of each clipmap is cached in video memory,
3: each clipmap contains 1/2 the resolution of the previous one,
4: geomorphing the vertices between shared edges, and
5: when the viewer moves, the clipmaps are incrementally filled with new data using toroidal (wrap-around) access.

My clipmap engine implements all of that.

I suppose I could do geometry clipmapping and ditch the incremental filling of the clipmap. (In order to infringe, your invention must use all the claims). Video cards have enough RAM that I can store the entire height map, normal map, etc. on the video card instead of just storing the clipmap data. This means I can't have a planet-wide area to roam, only a local area. This should be OK since I'm only interested in rendering from the surface anyway.

-- jas, libnoise project admin


#8 Trapper Zoid   Crossbones+   -  Reputation: 1370

Posted 20 December 2005 - 09:33 AM

Quote:
Original post by caffeineaddict
Even so, if the article was out for public consumption on implementing this technique before the patent was issued, wouldn't someone implementing it from this article be within their rights since it was already out in the open for that purpose? Sucks either way though.

That's true, but I think it's based on when the patent was filed. Looking through the dates of the SIGGRAPH article and the publication of the book, both were made after the patent file date.

I'm just not sure what the novel part of this patent is; I've only skimmed the abstract, but surely this technique has been around for a while?
edit:Whoops, there's been a reply on this; I'll have to read through that!

#9 AnonymousPosterChild   Members   -  Reputation: 168

Posted 20 December 2005 - 09:34 AM

MS will patent everything but never really sue over it. Its more a defense mechanism so that if somebody ELSE does it first and patents it, there wont be a hugeass legal battle.

This actually HELPS people against software patents, because MS is essentially cutting off the people who could sue over it.
With love, AnonymousPosterChild

#10 phantom   Moderators   -  Reputation: 8563

Posted 20 December 2005 - 09:35 AM

instead of all this knee-jerk reaction to it why not email the guy who helpped create the system and ask what the score is?

#11 Trapper Zoid   Crossbones+   -  Reputation: 1370

Posted 20 December 2005 - 09:43 AM

Okay; I'm no graphics expert, but I'm not sure which bit of this is the novel part.

Quote:
Original post by jasjas
Hi Trapper Zoid, the gist of his claims are:
1: a number of nested regular grids (clipmaps) surrounding the viewer,
2: the contents of each clipmap is cached in video memory,
3: each clipmap contains 1/2 the resolution of the previous one,

Surely these three steps have been done before, right? I'm not so sure about storing the data in video memory, but using multiple regular grids to represent terrain data must have been used? That seems reasonably obvious to me; similar techniques are used in computer vision for looking for features.

Quote:

4: geomorphing the vertices between shared edges, and
5: when the viewer moves, the clipmaps are incrementally filled with new data using toroidal (wrap-around) access.

I'm not that sure what steps 4 and 5 involve. Is step 4 linking the seams between the high detail clipmap close to the camera with the lower detail clipmaps further away? I'm also not entirely sure what step 5 means, either. It's possible these are new, because I'm not that knowledgable about the hardware side of things, or what's involved with step 4. However, I have seen LOD implemented in terrain engines for ages.

It could be that the novel part is using video memory for blazingly fast performance; I wouldn't know if that was new or not.

Heh, I suppose I'd be more use if I was actually a graphics programmer [smile]. I suppose I'm just curious about everything, especially new algorithms to do cool stuff.

Quote:
Original post by phantom
instead of all this knee-jerk reaction to it why not email the guy who helpped create the system and ask what the score is?

I was going to recommend this as well, but I forgot [smile]. However, a problem might be that the patent is granted to Microsoft, not the researchers involved. Most researchers I know would be thrilled for people to use their algorithms, but the legal department at Microsoft might have other views.

#12 phantom   Moderators   -  Reputation: 8563

Posted 20 December 2005 - 09:50 AM

As I recall the guy who invented/came up with it works for MS in their research division, so in theory at least he might have an idea of the state of play, if not you can always ask if he can pass it up the chain/give you the address of someone to talk to about it..

#13 jasjas   Members   -  Reputation: 262

Posted 20 December 2005 - 09:55 AM

Quote:
Original post by phantom
As I recall the guy who invented/came up with it works for MS in their research division, so in theory at least he might have an idea of the state of play, if not you can always ask if he can pass it up the chain/give you the address of someone to talk to about it..


Hi phantom, I'll email him and ask him if he knows if free implementations are allowed. Although he's the inventor, Microsoft will have the final say.


#14 Conner McCloud   Members   -  Reputation: 1135

Posted 20 December 2005 - 10:04 AM

MS bends over backwards to make game development on their platforms as easy as possible...do you honestly believe they're going to release an algorithm for use in video games and then not let anybody use it?

CM

#15 Witchcraven   Members   -  Reputation: 564

Posted 20 December 2005 - 10:22 AM

I am sure its fine. Just use it.

#16 phantom   Moderators   -  Reputation: 8563

Posted 20 December 2005 - 10:24 AM

Quote:
Original post by Conner McCloud
MS bends over backwards to make game development on their platforms as easy as possible...do you honestly believe they're going to release an algorithm for use in video games and then not let anybody use it?


Good point, well made!






Old topic!
Guest, the last post of this topic is over 60 days old and at this point you may not reply in this topic. If you wish to continue this conversation start a new topic.



PARTNERS