Hi Robert, i will do it as soon as possible. Unfortunately i am too snowed under with work. But i will work on, and i will post any progress.
is there a OSG meeting planed for this summer, may i don't work in august :-) /adegli 2007/5/18, Robert Osfield <[EMAIL PROTECTED]>:
Hi Adrian, I don't really know the ins and outs of driver rev's under Windows w.r.t tristripping vs triangles. It seems odd that there is a difference, especially if both are indexed. Could you create an test example that can create triangle mesh vs the same size tristripped meshes? Robert. On 5/18/07, Adrian Egli <[EMAIL PROTECTED]> wrote: > So just for your information. I tested same databases on Windows XP based > platforms with different GPU vendors (ATI, NVIDIA) and also different type > of graphic cards. Under OS/X i have just ATI X1600 Mobile graphic card > tested. But i can say that under windows XP the performance issue is much > bigger than under > OS/X. Is there a known problem by using Triangles Strips compared to > Triangles under Windows? Of course it's not the same graphic command and of > course the > strips needs different work steps on the card compared to just triangles. > but what i can not understand is the reason why os/x opengl is much faster > than windows xp opengl driver in processing tri strips. is there some known > bug in the windows core library ? if someone knows more about this issue > let me know, i made still a work around, so isn't quite importent to get a > fast answer. > > /adegli > > 2007/5/16, Adrian Egli <[EMAIL PROTECTED]>: > > Now i compared a database with 6M tri angles to a database with a 500K > primitve database. > > the second one has a lot small tri strips. (with about 10 tri per strip) > and for this i get > > the issue. the huge one, with 500k triangles as triangles rendered, not > strip > > is much faster > > > > i feel like the that the graphic pipeline with produce this problem > > > > /adegli > > > > > > > > > > > > 2007/5/16, Robert Osfield < [EMAIL PROTECTED]>: > > > Hi Adrian, > > > > > > Also make sure that you are compiling with a release build as this > > > makes a huge difference to performance. > > > > > > Also considering using osgViewer's support for DrawThreadPerContext, > > > as this can hide the cost of a long cull traversal very effectively on > > > dual core machines. > > > > > > Robert. > > > > > > On 5/16/07, Adrian Egli <[EMAIL PROTECTED]> wrote: > > > > Many thanks robert. > > > > > > > > so i will switch of > > > > > > > > osgUtil::GLObjectsVisitor: > > > > > :SWITCH_OFF_VERTEX_BUFFER_OBJECTS > > > > > > > > VBO > > > > > > > > :-) > > > > > > > > > > > > > > > > 2007/5/16, Robert Osfield < [EMAIL PROTECTED]>: > > > > > Hi Adrian, > > > > > > > > > > Performance optimization is a huge and complex topic. I'm not going > > > > > to go into it now, search the archives. > > > > > > > > > > W.r.t VBO's vs display lists, I find VBO's slower for big complex > > > > > models. Model with a small number of osg::Geometry which are all > > > > > large work well with VBO, while models with a > > > > > large number of osg::Geometry which all are small work best with > display > > > > lists. > > > > > > > > > > Robert. > > > > > > > > > > On 5/16/07, Adrian Egli <[EMAIL PROTECTED] > wrote: > > > > > > hi > > > > > > > > > > > > It got some probleme in opening of huge 3D database files (IVE > format). > > > > At > > > > > > the moment the file is first time displayed whe have to > > > > > > check for 2 different cases: > > > > > > > > > > > > (1) case would be: the whole file content (scene) is displayed: we > > > > rotate > > > > > > the camera, move the camera and so on, no probleme, the fps stay > > > > constant. > > > > > > on my carte 10fps, you see quite huge > > > > > > file. > > > > > > > > > > > > (2) the camera is in the center of the file, and only a small part > of > > > > the > > > > > > scene is display: we do same thing manipluating the camera, our > movement > > > > is > > > > > > not yet fluid. once we have displayed > > > > > > the whole file once, part by part the scene is greate working. is > there > > > > a > > > > > > problem in transfering datas to the GPU? buffered object, better > to use > > > > > > displaylist? > > > > > > > > > > > > > > > > > > i am using: > > > > > > > > > > > > > > > > > m_sceneView->setDefaults(mode|osgUtil::SceneView::COMPILE_GLOBJECTS_AT_INIT); > > > > > > > > > > > > > > > > > > > > > > > ((osgUtil::GLObjectsVisitor*)m_sceneView->getInitVisitor())->setMode( > > > > > > > > > > > > > > > > > ((osgUtil::GLObjectsVisitor*)m_sceneView->getInitVisitor())->getMode()| > > > > > > > > > > > > > > > > > osgUtil::GLObjectsVisitor::SWITCH_ON_VERTEX_BUFFER_OBJECTS > > > > > > > > > > ); > > > > > > > > > > > > > > > > > > regards > > > > > > adegli > > > > > > > > > > > > > > > > > > _______________________________________________ > > > > > > osg-users mailing list > > > > > > [email protected] > > > > > > > http://openscenegraph.net/mailman/listinfo/osg-users > > > > > > http://www.openscenegraph.org/ > > > > > > > > > > > _______________________________________________ > > > > > osg-users mailing list > > > > > [email protected] > > > > > > http://openscenegraph.net/mailman/listinfo/osg-users > > > > > http://www.openscenegraph.org/ > > > > > > > > > > > > > > > > > _______________________________________________ > > > > osg-users mailing list > > > > [email protected] > > > > http://openscenegraph.net/mailman/listinfo/osg-users > > > > http://www.openscenegraph.org/ > > > > > > > _______________________________________________ > > > osg-users mailing list > > > [email protected] > > > http://openscenegraph.net/mailman/listinfo/osg-users > > > http://www.openscenegraph.org/ > > > > > > > > > > _______________________________________________ > osg-users mailing list > [email protected] > http://openscenegraph.net/mailman/listinfo/osg-users > http://www.openscenegraph.org/ > _______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
_______________________________________________ osg-users mailing list [email protected] http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
