Hi Sukender, On Tue, Apr 20, 2010 at 3:31 PM, Sukender <[email protected]> wrote: > I'm not aware of compilers having such problems. I guess it should be okay, > but I'm not a cross-compiler wizard ;)
Previously under Solaris/SunOS compilers the STL containers don't support templated method on them. I don't know if this has been addressed yet. We've had record various parts of the OSG to avoid this issue in the past - using the std algorithm templated functions rather than methods. > About my problem, well I: > 1. Use osgModeling to generate normals from double-precision vertices > (modified version of osgModeling. Pending, because I wait for this submission > to be accepeted) > 2. Have some custom code using TriangleFunctor to apply on double-precision > vertices. > I may have problems using LineSegmentIntersector later, but not at the moment. So I guess even with the new flexibility you code adds you'll still need two functors and calls on the drawable to compute the normals. For testing purposes it'd be good to have a tested bed for whatever functionality we come up, would the osgUtil::SmoothingVisitor or LineSegmentIntersector a good match? We need to generate some double vertex data to test against - we could add this to OpenSceneGraph-Data, or have a spefic osgexample that convertes from Vec3Array to Vec3dArray and it's own testing code. > I often have problems handling both simple and double precision vertices > and/or vertices arrays... Maybe there is something we can do at a very low > level? Some wrappers to read "Vec[234]d?Array"? I was wondering whether we could have a local cache that converts Vec3d's to Vec3 to enable an easier mapping, efficiency is a real issue though. The TriangleIndexFunctor might be a used template class to extend, perhaps even TriangleFunctor itself could utilise it. Right now TriangleIndexFunctor doesn't record the vertex array details, but this would be an easy thing to extend to record a pointer to the vertex array. I think the next step should be to isolate a good example in the existing OSG code base to get working better with doubles and see how cleanly we can get it to work. Robert. _______________________________________________ osg-submissions mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
