Hi Mike, yes, the question confuses when one starts to work with pure shader environments.
the approach is to have a custom format and custom materials which are not so far implemented in core of osg. I found one osg compatible approach for materials see https://code.google.com/p/hogbox/ while the code has nice materials class https://code.google.com/p/hogbox/source/browse/trunk/src/hogbox/HogBoxMaterial.cpp the code provided lacks loader, another example format for shader-only based osg is internal format for http://osgjs.org/ there is though one catch - it is possible to save to this format from osg, but load only in osgjs, but it might shed light on how to proceed with own custom format https://github.com/cedricpinson/osg/tree/master/src/osgPlugins/osgjs also there is a split plugin which is needed if there is hardware limitation on vertex count https://github.com/cedricpinson/osg/tree/master/src/osgPlugins/split Regards Sergey On Thu, Jun 26, 2014 at 12:24 AM, Mike Metcalf <[email protected]> wrote: > We are putting together an mapping application targeted for the imx6 > hardware. The drivers support GLES 2.0, with no fallback/support for > previous versions. > > We have models and other resources in osgb files, and load them at > run-time, attaching them to our scene graph as necessary. On our dev > platforms (win7 and linux on x86) things render without issue, but when > cross-compiled and deployed to our imx6 hardware, these osgb resources did > not show. > > We traced it to bad shaders, and found on these forums that the reason is > that ShaderGen.cpp was generating (fallback?) shaders which are not > compatible with GLES 2.0. > > My first naive step was to modify the vertex and fragment shaders in > ShaderGen to enable them to compile in our GLES GPU. We had to remove > references to the fixed-function pipeline, and that included references to > gl_Color (et al), which was how each vertex got it's starting material > color. So... we could see our geometry on-screen, but the material colors > as stored in the osgb file were unavailable. > > I then dug around GLSL 2.0 enough to tentatively conclude that the new > method of supplying those vertex colors to the shader code is to declare > vertex attributes in the vertex shader and then bind an array of attributes. > > But, and this is my question, how do we use the generalized approach of > reading in osg::nodes from osgb files, attach them to the scenegraph and > still bind attributes to them? > > It seems that we would either need to create a unique osg::program per > osg::node we load and then somehow glean the vertex color information from > the osg::node, load it into an array and bind that array to the shader > program. > > Can anyone shed any light on this? I thing I'm missing something fairly > fundamental. > > Many thanks in advance! > > Mike > > ------------------ > Read this topic online here: > http://forum.openscenegraph.org/viewtopic.php?p=59905#59905 > > > > > > _______________________________________________ > 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
