[ODE] ODE] Convex Volume Geom (Shipped games using GJK)
Jovani Cassini
jovanicassini at yahoo.com
Thu Dec 22 17:41:49 MST 2005
Hello I am new to the ODE list, but I had been working on the video game industry for a
few years implementing physics. The pass few years has been integrating commercial
physics engines into game engines and I had some experience using Havok.
Quote:
"Lots of commercial shipped games succesfully used GJK:
1) David Wu from Pseudointeractive just mailed me that Full Auto and Cell
Damage both use GJK.
2) On glalgorithms-list Jay Stelly from Valve talks about his use of GJK in
Half Life 2.
3) Check out the games in this long list, for example Halo 2 and Painkiller:
http://www.havok.com/content/blogcategory/29/73/
Happy Holidays!
Erwin Coumans"
I do not think this is totally accurate.
First of all I think that all the games mentioned in the list are using that havok 2.x. So if that is the case then every title cannot count for a different game since they are all using the same implementation.
Second of all havok also has an ill implementation of the GJK, it is funny because they even renamed to GHK, which stand for the Havoc variant of GJK. What they do is that the have and alternate collision solution. The core will call for a normal GJK and if is fail they will recover from the failure by calling the SAP. They did not solved the instability problems that are inherent to GJK they just went around it.
For those of you using havok, it is easy to verify the failure rate, all you need to do is to setup the error handler.
The error code for the GHK error is undocumented, but there is a very easy way to check it out by just tracing all of the messages. You will see that it has a failure rate higher than 1 in 1000 calls.
It is even much higher than Ginos implementation.
---------------------------------
Yahoo! Photos
Ring in the New Year with Photo Calendars. Add photos, events, holidays, whatever.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://q12.org/pipermail/ode/attachments/20051222/e331b340/attachment.htm
More information about the ODE
mailing list