Re: [Algorithms] Fast picking against triangle soup
Brought to you by:
vexxed72
From: Luis V. <lui...@mi...> - 2007-01-31 17:56:54
|
There are different ways to speed up the construction process, but you need= to understand first how it is being constructed now. Better heuristics do= not necessarily mean higher construction times. There were some papers pr= esented on RT06 on this specific topic (i.e. using Adaptive Error Bounded H= euristics) As somebody else mentioned on another post, if the construction process end= s up being too expensive to be done at load time you can probably do it at = a different stage of the pipeline (i.e. asset import time) Building some kind of stat reporting into the Kd-tree traversal code would = help you better understand the problem: tree depth, node count, average tri= angles per-node, ray/triangle intersections performed for a ray cast, etc. Luis Villegas -----Original Message----- From: gda...@li... [mailto:gdalgorithms-= lis...@li...] On Behalf Of Robert Blum Sent: Wednesday, January 31, 2007 9:09 AM To: Game Development Algorithms Subject: Re: [Algorithms] Fast picking against triangle soup > How are you building your kd-tree and how often? How big is the > data set you are working with? Once, at load time. (And they get squirreled away in a cache, since they're not cheap to build). Can't tell you how we're building them - just inherited the code yesterday. Still digging through it. Data set size is somewhere from ~100 tri's to 100,000 tris - depending on the object. > > The symptoms you are describing are not typical for kd-trees, so > your construction heuristics could be better or your traversal code > is doing something is not supposed to. I'm blaming the traversal code. With a semi-reproducible case (i.e. I keep picking till it's bad), if you break execution, you're about 100 calls or so deep in the callstack. Im just not sure if I want to fix that, or if there's a better structure around. The loose octree sounds tempting, and it's not going to take much time to rig up a test case for it. > Investing on good construction heuristics can improve performance > dramatically. Yes, but aren't they usually negatively impacting construction time? > Take a look at the real time ray tracing notes for SIGGRAPH 05 and > 06, and RT06: Thanks for the pointers - will dig through it. - Robert ------------------------------------------------------------------------- Take Surveys. Earn Cash. Influence the Future of IT Join SourceForge.net's Techsay panel and you'll get the chance to share you= r opinions on IT & business topics through brief surveys - and earn cash http://www.techsay.com/default.php?page=3Djoin.php&p=3Dsourceforge&CID=3DDE= VDEV _______________________________________________ GDAlgorithms-list mailing list GDA...@li... https://lists.sourceforge.net/lists/listinfo/gdalgorithms-list Archives: http://sourceforge.net/mailarchive/forum.php?forum_id=3D6188 |