HI Cedric, I can't add to the discussion on skinning specifically but general notes about bounding box and vertex attribute location as something I can chip in with.
First up, the bounding box issue is something I'd be lazy and conservative with - basically estimate the bounding box for the largest possible extents of the skinned geometry. You can use the osg::Drawable::setInitialBound() to provide the base. The actual code and API in Drawable w.r.t actually needs a bit of revist, but as it stands should be enough to get your going. W.r.t vertex attrib location & osg::Geometry, I've been doing a lot of work on osg::Geometry and vertex attributes this month. osg::Geometry itself hasn't had to have it's API modified, but the internal implementation has been change substantially w.r.t supporting both OpenGL and OpenGL ES 2.0, and now in conjunction with several supporting classes provides automatic conversion of the standard vertex arrays to vertex attrib equivalents. The arrays alias defaults to using the same attribute locations as set out in the original ARB_vertex_program extension, and these locations are bound to vertex shader attribute names by osg::Program's support for assigning names to attribute locations. Curiously to get it to work properly I've had to disable display lists, so it does look like the driver is using the osg::Program values to reassign the vertex attrib locations dynamically. Anyway, the work I've done shouldn't effect the skinning side too much, but I can say that it's perfectly doable to use vertex attributes in osg::Geometry as they stand right now (even before my changes) along with an appropriately set up osg::Program (via program->addBindAttribLocation(name, index)) as long as you disable display lists. There isn't any need to use a modified version of osg::Geometry - and should be possible with OSG-2.8. Robert. On Wed, Oct 21, 2009 at 1:37 PM, Cedric Pinson <[email protected]> wrote: > Hi Andrey, > > I did a proof of concept to put needed elements to osg. The main problem > is that currently there is no way to automatically get the location to > put the vertex attrib data. > The things i tried is to run the program once and dump the attrib > location and then edit the program to assign it. But it's hard coded and > not manageable for real use. The other stuff that would need some work > is the way to compute the bounding box. We can have different strategy: > - Define a specific bounding box, can be use when you just want to cull > without small detail so you setup a maximized bounding box. > - Define a low res skinning (software) and use the result to compute the > bounding box. > > Currently the software implementation compute the bounding box from the > original vertex transformed. It can consume too much time and if you > have a lot of vertex and you just want to approximate the bounding box. > > There is a thread that talk about it > Subject: [osg-users] Using vertex attributes > Date: 08/10/09 04:12:14 (Sun, 09 Aug 2009 20:12:14 -0600) > > > Anyway if you want to look to what is working, you can get my branch > at hg clone http://hg.plopbyte.net/osg-trunk/ > the code is disabled but can be enabled to try it. I wait Paul Martz > submit its work about a new osg::Geometry that would be able to get the > attrib location. > > Cheers, > Cedric > > -- > +33 659 598 614 Cedric Pinson mailto:[email protected] > http://www.plopbyte.net > > > On Wed, 2009-10-21 at 14:14 +0200, Andrey Krekhov wrote: >> Hi, >> >> Currently I'm working on my Bachelor thesis within a project using osg >> (and osgAnimation). One of my tasks will probably be to accelerate >> osgAnimation by using the GPU. Thus, probably i will have to implement >> Hardware Skinnig. Now my question: You mentioned that hardware skinnig >> is one of your short-term goals. What are your current status concerning >> this? Depending on that, I want to decide whether to wait for your >> release or implement it by my own (shader or Cuda based). >> >> Thanks in advance, >> >> Greetings, >> Andrey >> _______________________________________________ >> osg-users mailing list >> [email protected] >> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >> > > _______________________________________________ > osg-users mailing list > [email protected] > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org > > _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
