Hi Mattihas,
Please have a look at the OSG header for include/osg/Geometry, in
particular the section on setVertexIndices() :
/** deprecated - forces OpenGL slow path, just kept for
backwards compatibility.*/
void setVertexIndices(IndexArray* array);
See the forces OpenGL slow path... yep as you are using this you are
basically driving the OSG/OpenGL is least efficient way possible.
This method is only kept for backwards compatibility, but I very
strongly recommended NEVER EVER EVER using for any other application
;-)
Use DrawElements*() instead if you want to use vertex indices.
Robert.
2012/3/21 Matthias Thöny <[email protected]>:
> Hi,
>
> I have some troubles with an OSG integration and tracked down some stuff
> which I do not really get in conecpt. What I wanted to do in this short
> program is checking how much triangles can be rendered with OSG with indexed
> triangle strips in one geode on one side with OSG standard implementation and
> on the other hand with an overridden drawimplementation of an own geometry
> class. Why? because I need to know the borders of OSG.
>
> In the picture you see one Geode with one geometry object. This object is set
> up the following:
>
> Code:
>
> void create()
> {
> useOpenGL = true;
> sidelenght = 500;
> triangleStripPerColumn = sidelenght - 1;
> triangleCountperSide = sidelenght * 2;
> dist = 1.0f;
> vertexCount = sidelenght * sidelenght;
> indexCountperSide = triangleCountperSide + 2;
> indexCount = sidelenght * (triangleCountperSide + 2);
>
> vertices = new osg::Vec3Array(vertexCount);
> for(int i = 0; i < sidelenght; i++)
> {
> for(int j = 0; j < sidelenght; j++)
> {
> (*vertices)[j + i * sidelenght].set(dist * j, 0.0f, dist * i);
> //std::cout << "vertices: " << j + i * sidelenght << " " <<
> (*vertices)[j + i * sidelenght].x() << " " << (*vertices)[j + i *
> sidelenght].z() << " " << std::endl;
> }
> }
>
> indices = new osg::UIntArray(indexCount);
> int indexTriangleCount = 0;
> for(int i = 0; i < sidelenght; i++)
> {
> for(int j = 0; j < sidelenght; j++)
> {
> (*indices)[indexTriangleCount] = sidelenght * (i + 1) + j;
> indexTriangleCount++;
> (*indices)[indexTriangleCount] = (sidelenght * i) + j;
> indexTriangleCount++;
> }
> }
>
> this->setVertexArray(vertices.get());
> this->setVertexIndices(indices.get());
>
> for(int j = 0; j < triangleStripPerColumn; j++)
> {
> this->addPrimitiveSet(new osg::DrawArrays(GL_TRIANGLE_STRIP,j *
> sidelenght * 2, triangleCountperSide));
> }
> }
>
>
>
> So, it is mainly a square 500 vertices widht and height and there is one
> triangle strip (addPrimitiveSet) per row. The indices are set right, and it
> seems to be fine in the picture.
>
> But...
>
> why are there 500'000 vertices in the statics, when I created just 250'000.
> Do I have to delete something here to get this number?
> why are there almost 500'000 triangle strips? I have seen that you are doing
> a lot of complicated stuff, so how can this be done properly? In fact in this
> case, it should have 500 or less.
>
> I used a
>
> MacPro 3.1 with
> ATI Radeon HD 2600 XT 256 GB RAM
>
> for this test. This is a configuration which I have seen already causes some
> problems on OSG, but that's the point on testing in this case. :)
>
> Further on the very interesting thing is, that if you add the following Code
> piece:
>
> Code:
>
> this->setDataVariance(STATIC);
> this->setUseVertexBufferObjects(true);
> this->setUseDisplayList(false);
>
>
>
> The draw rate increases to 20 ms instead of 0.2 ms. This is Debug version,
> but compiled in Release it is the same factor 100 or so, which I think is
> really weared, especially, because vertex buffer object should be faster in
> this case (if erverything is in native OpenGL implemented without OSG). In
> this case the OpenGL version is 2.1.
>
> To the draw implementation. If i override the draw Implementation with an own
> function call just straight forward like:
>
>
> Code:
>
> void drawImplementation(osg::RenderInfo& renderInfo) const
> {
> osg::State& state = *renderInfo.getState();
> bool checkForGLErrors =
> state.getCheckForGLErrors()==osg::State::ONCE_PER_ATTRIBUTE;
> if (checkForGLErrors) state.checkGLErrors("start of
> Geometry::drawImplementation()");
>
> glEnableClientState( GL_VERTEX_ARRAY );
> glBindBuffer( GL_ARRAY_BUFFER,0);
>
> glBindBuffer( GL_ARRAY_BUFFER, vboVertexObject);
> glVertexPointer(3,GL_FLOAT,0,(char*) NULL);
>
> // glBindBuffer( GL_ARRAY_BUFFER, vboIndexObject);
> // glIndexPointer(GL_UNSIGNED_INT,0,(char*) NULL);
>
> glBindBuffer( GL_ARRAY_BUFFER, vboNormalObject);
> glNormalPointer(GL_FLOAT,0,(char*) NULL);
>
> for(unsigned int primitiveSetNum = 0; primitiveSetNum !=
> _primitives.size(); ++primitiveSetNum)
> {
> glDrawArrays(GL_TRIANGLE_STRIP, triangleCountperSide *
> primitiveSetNum, triangleCountperSide);
> }
>
> // unbind the VBO's if any are used.
> state.unbindVertexBufferObject();
> state.unbindElementBufferObject();
>
> if (checkForGLErrors) state.checkGLErrors("end of
> Geometry::drawImplementation().");
>
> //end_draw_time = timer.tick();
> }
>
>
>
> this is in fact working, but with some sideeffect (crash while statics blend
> in, etc.etc...), but I am really confused about is, why this method is not
> always called while the draw process, because if my xcode debugger does not
> lie to me (because it does that a lot of times in version 4.2), this method
> is reached once. Are you cloneing the object anyhow, to traverse in the draw
> method (which cannot be overriden), or how does this mechanism work?
>
> There is also the compileGLObjects method which implements the buffer usage
> and I have seen that, there are no implementations for Vertex Array Objects
> (OpenGL 3.0). Is there any implementation for that? or is this not supported?
>
> Sorry for the long post and the detailled questions.
> Thank you!
>
> Cheers,
> Matthias
>
> ------------------
> Read this topic online here:
> http://forum.openscenegraph.org/viewtopic.php?p=46492#46492
>
>
>
>
> _______________________________________________
> 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
