[osg-users] qt + osgviewer::viewer
Hi, I have created a widget qosgv that inherits from QWidget and osgViewer::Viewer. When I use this widget in my application, My scene appears normally. But, When I use the method setVisible(bool) and that I put true and false, My scene has completly disappeared from the qosgv widget (I only have the default background color of the osgviewer::viewer). On the other side, the scene graph always contains the scene. Why this happens ? Do I have forgotten to activ something when using setVisible ? Thank you! -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17637#17637 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Rendering Optimizations Not Working
HI Brian, You don't really provide enough information on what the complex app is doing to make any judgement on what is going on. The best I can suggest is that it sounds like you are breaking starte sorting and lazy state updating in some way, and in doing so are causing the draw dispatch and likely the draw down on the GPU to be heavily overloaded. The fix... don't break state sorting and lazy state updating. Robert. On Thu, Sep 24, 2009 at 10:12 PM, Brian R Hill bhil...@csc.com wrote: Folks, I'm rendering a scene with 3 different objects instanced 500 times each. When I run it in osgviewer is achieves 1000 hz, when I run it in a simple osgViewer::Viewer based app it runs at 1000 hz, when I run it in a more complex osgViewer::Viewer based app it runs at 30 hz. All the additional time is in the draw process. I used bugle to capture the opengl output (bugle is fantastic!) from the 3 apps and the first two outputs were practically identical - the 3 objects graphics state are setup once, followed by 500 glLoadMatrix and glCallList calls. 3 x (graphics setup, 500 x (matrix, calllist)) 3 graphics state changes The more complex app is completely different, the objects are drawn in arbitrary order with their graphics state updated every time. 3 x (500 x (graphics, matrix, calllist)) 1500 graphics state changes All apps call osgUtil::Optimizer::optimize on the scene. I'm digging through the complex app to see if there's a disable_rendering_optimizations call. Does anyone know what would cause this type of behavior? Thanks, Brian This is a PRIVATE message. If you are not the intended recipient, please delete without copying and kindly advise us by e-mail of the mistake in delivery. NOTE: Regardless of content, this e-mail shall not operate to bind CSC to any order or other contract unless pursuant to explicit written agreement or government initiative expressly permitting the use of e-mail for such purpose.. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] keyboard differences between OSG 2.8.2 vs 2.8.1
Hi Jim, On Fri, Sep 25, 2009 at 2:24 AM, Jim Brooks jimbl...@gmail.com wrote: On OSG 2.8.2, square brackets [] and parentheses() return the same values. OSG 2.8.1 returns different values as expected. Confirmed using my program and osgkeyboard. I'm using a USA keyboard on Debian 5 Lenny. I believe a bug fix in GraphicsWindowX11 in 2.8.2 revealled this bug. It's already fixed in svn/trunk and the OSG-2.8 branch. Robert. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Paged LOD for osg terrain
Hi Brett, On Fri, Sep 25, 2009 at 5:53 AM, Brett Thomas Lee brettle...@hotmail.com wrote: My requirements are terrain rendering on gpu and texturing should be compatible. All but software rendering is done on the GPU so it's a bit confusing still what you are actually curious about. Is that you want CLOD implemented down on the GPU? Are you looking for static geometry/LOD? Basically all the terrain rendering that OSG does or that users combines rendering of geometry for terrain and texturing for draping imagery and overlays on top. But this is what all graphics toolkits will do anyway so there really isn't much to explain, until become a bit more specific about what you are after. Robert. ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] [osgPlugins] Loading animation in OSG
Is it possible to try the FBX importer? Kind regards. Alessandro On 9/24/09, Michael Platings mplati...@gmail.com wrote: Hi Pradeep I have written an importer for the FBX format which supports animation. I've sent it to the submissions list so it's waiting review by Robert before it gets into the trunk, but you can get it and add it to your own version yourself. The email is Re: [osg-submissions] FBX plugin from the 21st August. -Michael ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] osgOcean collision detection
Hi Jan,
My apologies for not replying earlier, I haven't been checking
osg-users as much as I possibly should have. It appears that I've
missed quite a few osgOcean posts.
BTW, is there a reason why is the size of the screen hardwired to
1024x768? Do the shaders expect this? Or why is it so? It took me a while to
figure out why it doesn't draw correctly on our wall.
J-S pretty much answered this one- the screen resolution is only
required for setting up the FBOs involved in the fullscreen effects.
However, this can be changed using:
/** Sets the current screen size, needed to initialise the God Ray and
DOF frame buffers.
* Default is 1024x768.
*/
inline void OceanScene::setScreenDims( osg::Vec2s size ){
_screenDims = size;
_isDirty = true;
}
Setting that will cause the FBO textures to be resized and rebound at
the desired resolution on the next pass of the update visitor. You
could easily bind this function to a callback which calls it when the
screen resolution changes.
Back to the original problem with collision detection. As J-S said,
direct access to the vertices is not currently supported. However you
should be able to get access to the geometry by adding some simple
accessors. To be honest I haven't thought about collision detection
much as it wasn't an original requirement for me, but I can see that
it would be a valuable addition and i'll try and give you some
pointers.
The vertices and normals involved in the current frame are stored
within the Vec3Arrays called _activeVertices and _activeNormals in
FFTOceanSurface. Whenever the animation frame or LOD level changes the
entire vertex and normal arrays are updated from precomputed LOD and
animation frames stored in a 2D vector called _oceanTiles - where
they're stored in the format _oceanTiles[frame][lodlevel] .
These two arrays are then assigned to a series of custom osg::Geometry
objects called MipmapGeometry which are basically just osg::Geometry
objects with some extra member variables added.
The MipmapGeometry class stores the current LOD level of the patch and
the primitives that define the geometry for each of the ocean patches.
These objects are then stored within a 2D std::vector called
_oceanGeom. There is in fact a private inline function that will
returns a raw pointer to the desired patch:
/**
* Convenience method for retrieving mipmap geometry from _oceanGeom.
*/
inline FFTOceanSurface::MipmapGeometry* getTile( unsigned int x,
unsigned int y ){
return _oceanGeom.at(y).at(x).get();
}
Like I say I haven't thought about collision detection much, but my
initial thoughts are that you could use the bounding boxes of the
mipmap geometries as a quick way to find the patch in which the
collision occurs and the test the triangles within that patch. The
animation, of course complicates things somewhat. If you're planning
on doing accurate collision detection you should be able to get access
all the vertices involved for the frames at the current LOD from the
_mipmapData container.
I hope that helps a bit, if you have any further questions don't
hesitate to ask and I'll try and be more diligent with my replies.
Regards,
Kim.
2009/9/25 Jean-Sébastien Guay jean-sebastien.g...@cm-labs.com:
Hi Jsn,
I will have a look at it tomorrow.
Excellent, let me know if you have any questions though Kim C. Bale is more
familiar with the actual algorithms so for some questions I might have to
defer to him.
BTW, is there a reason why is the size of the screen hardwired to
1024x768? Do the shaders expect this? Or why is it so? It took me a while to
figure out why it doesn't draw correctly on our wall.
I think it's two things:
a) preference on Kim's part (the example was coded like that when he first
posted the code)
b) all the post-render cameras for the RTT effects (DOF, glare, distortion)
need to know the resolution of the screen, and I don't recall if/how they
get this info, but I think he may have hard-coded it to 1024x768 to avoid
having to pass the resolution all over the place. This can be changed.
Remember that oceanExample is just that - an example. The API in the
osgOcean lib should allow you to use it in any way, not just in the way
oceanExample uses it.
Bear in mind that osgOcean came from his own project, so the API was made
for the way he used it. Then I came in and I had some other requirements for
the API, so I made it evolve that way. It may well be that our two sets of
requirements still don't cover all possible uses (in fact the opposite would
be surprising) so if you need the API to evolve in some other way, it won't
be a problem! :-)
J-S
--
__
Jean-Sebastien Guay jean-sebastien.g...@cm-labs.com
http://www.cm-labs.com/
http://whitestar02.webhop.org/
___
osg-users mailing list
osg-users@lists.openscenegraph.org
Re: [osg-users] osgOcean collision detection
Hi Kim (and Jean-Sebastien),
Kim Bale kcb...@googlemail.com wrote:
Hi Jan,
My apologies for not replying earlier, I haven't been checking
osg-users as much as I possibly should have. It appears that I've
missed quite a few osgOcean posts.
No problem :)
BTW, is there a reason why is the size of the screen hardwired to
1024x768? Do the shaders expect this? Or why is it so? It took me a
while to figure out why it doesn't draw correctly on our wall.
J-S pretty much answered this one- the screen resolution is only
required for setting up the FBOs involved in the fullscreen effects.
However, this can be changed using:
/** Sets the current screen size, needed to initialise the God Ray and
DOF frame buffers.
* Default is 1024x768.
*/
inline void OceanScene::setScreenDims( osg::Vec2s size ){
_screenDims = size;
_isDirty = true;
}
Ah, OK - I have missed this function. I saw that the variables are initialized
in the constructors but that there wasn't a way to pass new screen dimension
to them (would be a good addition to the constructor, IMO).
Setting that will cause the FBO textures to be resized and rebound at
the desired resolution on the next pass of the update visitor. You
could easily bind this function to a callback which calls it when the
screen resolution changes.
Good, that will help. I more-or-less need to call this only on setup, but at
least it saves me hacking in an extra constructor.
Back to the original problem with collision detection. As J-S said,
direct access to the vertices is not currently supported. However you
should be able to get access to the geometry by adding some simple
accessors. To be honest I haven't thought about collision detection
much as it wasn't an original requirement for me, but I can see that
it would be a valuable addition and i'll try and give you some
pointers.
What I need is rather simple - I need an object to bob on the surface of the
waves, so basically I just need the current height at a given x,y position.
...
Like I say I haven't thought about collision detection much, but my
initial thoughts are that you could use the bounding boxes of the
mipmap geometries as a quick way to find the patch in which the
collision occurs and the test the triangles within that patch. The
animation, of course complicates things somewhat. If you're planning
on doing accurate collision detection you should be able to get access
all the vertices involved for the frames at the current LOD from the
_mipmapData container.
I think that is even an overkill for me - I just need the height to be able to
attach an object to the surface of the wave. But thanks for the pointers, I am
going to have a closer look at it today.
Regards,
Jan
signature.asc
Description: This is a digitally signed message part.
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Is the osgOcean Linux bug still present?
Hi Pierre, Anthousis, Thanks for further narrowing this bug down. Unfortunately I haven't used (or even come into contact) with an ATI card for quite some time, so I won't be able to test this. However, I do remember a while back that Robert tested osgOcean with an ATI 4670 and that worked with his linux distro although I'm unsure whether that was 32 or 64bit. I'm going to take the plunge and install Ubuntu on my 32bit laptop this weekend and see what happens. Although, having never used linux actually getting it to compile at all might be a challenge in itself. :) So yes if anyone with a linux box and an ATI card would like to raise their hand on this one that would be great. Regards, Kim. 2009/9/20 Pierre BOURDIN bour...@imerir.com: Hi Anthousis Andreadis , I had already sent a report going the same way as yours, using Debian Ubuntu distribution... So it confirms it is a 32bit only bug. The next step would be to test with non-nvidia hardware to know if it is related to nvidia graphic card only ? I don't have any recent/decent ATI graphic card to conduct some tests... If someone with a 32bit linux box and an ati card could make a test it would be a great help. Cheers, Pierre. 2009/9/17 Anthousis Andreadis anthou...@gmail.com Hi Kim, Update on the linux bug issue with the squary waves. Two days ago i compiled the osgOcean v1.0 (not the repository) and i found out that i had the bug that has been described in previous mails. My system is RHEL 4 and the GC a Quadro 5500 with the latest drivers and OpenSceneGraph 2.8.0 and for FFT's i used fftss lib. The same problem exists with the svn version of osgOcean. After a lot of tries i found out that 32bit compilation causes the problem. On the same system with 64bit compilation the problem is solved. If the other guys with the same problem can test both 32bit and 64bit we could get closer to the source of the problem. Regards Anthousis Andreadis ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] qt + osgviewer::viewer
I have discovered that the aspectRatio from the frustum of the perspective projection matrice was NAN. So I have looked at the width and height of my qosgv widget and the width is at zero (but the height is ok). I dont't know yet why the width of my qosgv widget has changed and not the height but I think I am on the good way. -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17646#17646 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] [3rdparty] osgOcean collision detection
Hi Kim Bale, your hints are valuable. As I've seen so far I need to know the height on a specific point on sea surface. I wrote a function in FFTOceanSurface passing the x,y of a point. That function finds the corresponding ocean tile and retrives the closest vertice using the getVertex of OceanTile function. That way I can program my objects on the sea surface to maintain that height from the sea zero level on every frame. I haven't done yet (actually it doesn't work so far) but I believe I am going to the right direction. I' ll make a post when I have good news. Thank you! Cheers, Dimitrios -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17647#17647 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] [3rdparty] osgOcean collision detection
Hi Dimitrios, I wrote a function in FFTOceanSurface passing the x,y of a point. That function finds the corresponding ocean tile and retrives the closest vertice using the getVertex of OceanTile function. That's good. We could add another function that would return a weighted average of the closest 4 points, which would be more accurate but also slower. Finally we could also add a function that returns a 2D array of heights given an axis-aligned bounding box which is the (x1, y1, x2, y2) of where on the grid your object is. This would allow the collision detection to do what it wants with the points. Those 3 functions would flesh out the API for one-way interaction (water - object). As I told Jan, I'm not working on this right now, but will in the near future. Let me know if you get something working, we can share the workload that way :-) J-S -- __ Jean-Sebastien Guayjean-sebastien.g...@cm-labs.com http://www.cm-labs.com/ http://whitestar02.webhop.org/ ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
[osg-users] Clarification on use of Manipulators in CAD-Style app
Hi there, I am developing a CAD-Style application and am interested in using osgManipulator to provide picking, moving, rotating and scaling of objects. I just have a couple of questions with regard to the use of manipulators, and hope some of you can take the time to answer them for me. Firstly my application currently has a relatively flat scenegraph, where each Object in the CAD app is arranged as follows: Code: PositionAttitudeTransform | Geode | Geometry Then I have N Objects arranged as above in the scene that are a child of the root scene node (so a flat scene graph). From my understanding from reading the osgManipulator example, to manipulate an object I can use the following hierarchy: [Image: http://i137.photobucket.com/albums/q217/andyb1979/manpulatorhierachy.png ] In this diagram Geode is the object I want to manipulate, so I would replace that with the PAT from my 3D Object implementation, Selection wraps the object to manipulate, Dragger can go anywhere in the scene and the CommandManager links the Dragger to the Selection. So -- some questions on how I may use this in my app: * Firstly, to select an object, I pick it. At this point should I create a Selection and place in-between my 3D object and SceneRoot? Or can selection be anywhere and just place the object as a child of that? * Secondly can the dragger be anywhere in the scene hierachy or must it be a parent of the selection? * Similarly, does the Selection have to be specifically put between my 3D object and the scene root (In a CAD-Style app I will be selecting/deselecting objects quite regularly) * Say my object has a transform applied to it by its PositionAttitudeTransform. I need to somehow get that transform, apply to the dragger, then let the user manipulate it and apply the resultant transform to the PAT, removing the dragger/selection from the scene. Any ideas how this could work? * Is there a way to get an event or call-back out for each translate, rotate, scale operation as I need to push these into an undo-redo stack * Finally, could the above be modified to work with multiple objects in the same selection and how would I apply the Dragger transform to all of them before/after the event? Thanks for your time, any insights would be gratefully received. Andrew[/list] -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17650#17650 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] noob question about texture3d
Hi Otto, Texture3D is no different to Texture1D, 2D etc, all just accept an osg::Image which packs the raw byte data. There are also standard Texture methods for setting the filtering and wrapping, for examples of these it best to look at the osgtexture2D example. As for creating a 3D osg::Image, again this is much the same as a 2D osg::Image, the only difference is the r dimension having a value that isn't 1. The osgvolume example has code that packs an osg::Image so have a look at this. Robert. On Fri, Sep 25, 2009 at 12:48 PM, Otto Cologne schur...@gmx.de wrote: Hi, ok I give up. After a lot of trying and searching i still fail to create a simple 3d texture from a raw byte array for example. Could anyone help me with how to put something like Code: glPixelStorei(GL_UNPACK_ALIGNMENT, 1); glGenTextures(1, m_GLuint); glBindTexture(GL_TEXTURE_3D, m_GLuint); glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP); glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP); glTexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP); glTexImage3D(GL_TEXTURE_3D, 0, GL_RGBA, m_columns, m_rows, m_frames, 0, GL_LUMINANCE, GL_UNSIGNED_BYTE,m_data); into osg code or point me towards some example code. The texture3d example here doesn't really help. Any help would be greatly appreciated. ... Thank you! Cheers, rs -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17648#17648 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Clarification on use of Manipulators in CAD-Style app
Hi Andrew, In the svn/trunk version of the OSG I have refactored osgManipulator a bit to make it more flexible and easier to use. I would strongly suggest having a look at it as the extra complexity of CommandManager is no longer needed, also you no longer need to implement your own event handling, as osgManipulator can now automatically pick up on events. There is also a new callback support that allows you to move objects around in the scene to follow manipulators. Robert. On Fri, Sep 25, 2009 at 1:58 PM, Andrew Thompson andyb1...@yahoo.co.uk wrote: Hi there, I am developing a CAD-Style application and am interested in using osgManipulator to provide picking, moving, rotating and scaling of objects. I just have a couple of questions with regard to the use of manipulators, and hope some of you can take the time to answer them for me. Firstly my application currently has a relatively flat scenegraph, where each Object in the CAD app is arranged as follows: Code: PositionAttitudeTransform | Geode | Geometry Then I have N Objects arranged as above in the scene that are a child of the root scene node (so a flat scene graph). From my understanding from reading the osgManipulator example, to manipulate an object I can use the following hierarchy: [Image: http://i137.photobucket.com/albums/q217/andyb1979/manpulatorhierachy.png ] In this diagram Geode is the object I want to manipulate, so I would replace that with the PAT from my 3D Object implementation, Selection wraps the object to manipulate, Dragger can go anywhere in the scene and the CommandManager links the Dragger to the Selection. So -- some questions on how I may use this in my app: * Firstly, to select an object, I pick it. At this point should I create a Selection and place in-between my 3D object and SceneRoot? Or can selection be anywhere and just place the object as a child of that? * Secondly can the dragger be anywhere in the scene hierachy or must it be a parent of the selection? * Similarly, does the Selection have to be specifically put between my 3D object and the scene root (In a CAD-Style app I will be selecting/deselecting objects quite regularly) * Say my object has a transform applied to it by its PositionAttitudeTransform. I need to somehow get that transform, apply to the dragger, then let the user manipulate it and apply the resultant transform to the PAT, removing the dragger/selection from the scene. Any ideas how this could work? * Is there a way to get an event or call-back out for each translate, rotate, scale operation as I need to push these into an undo-redo stack * Finally, could the above be modified to work with multiple objects in the same selection and how would I apply the Dragger transform to all of them before/after the event? Thanks for your time, any insights would be gratefully received. Andrew[/list] -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17650#17650 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] [osgPPU] Using osgPPU
Hi Art! Thank you very much for the correct version! I am sorry I didn't give my whole code, I was afraid it would have confused everyone. I also apologize to bother you with huge mistakes. Everything works fine now thanks to you ! -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17654#17654 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Changing CopyOp
This change wouldn't be a clone() method, it'd be a shareASingleClone() method which is a totally different thing and in the general case would not be safe or appropriate a replacement for clone(). Maybe I explained it incorrectly, the idea is simply that calling clone should give you an exact clone of the argument, with all the relationships between its child nodes intact. But I accept that it's a big change, I'm glad I checked with you before I started relying on it :) As it stands, osgAnimation is compatible with cloning as it is: relationships between nodes are inferred from their names. Issues with COLLADA using non-unique IDs prompted me to suggest setting pointers to identify related objects, but I can work around it another way. Thanks for your quick response -Michael ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.or g __ This email has been scanned by the MessageLabs Email Security System. For more information please visit http://www.messagelabs.com/email __ __ This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the system manager. This email has been scanned by the MessageLabs Email Security System. For more information please visit http://www.messagelabs.com/email __ ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
Re: [osg-users] Changing CopyOp
Hi Robert
I ll try to explain the situation about the cloning problem.
Cloning works as long as the animation manager relink target, this
operation is done automatically but can be done manually.
The problem Michael has encountered is that using collada reader it
seems to not possible to setup callback with name suitable for the
linker work. It means that without the linker the cloned update callback
are not relinked and then are not updated with the result of animation.
In order it works the linking operation (manually or automatically) must
be executed after a deep copy of a graph.
The current operation of the clone operation is to duplicate the update
callback like other deep copy of objects.
Using a shallow copy for a deep copy will have side effect, like animate
two node instead of the copied node.
What is the linker and why ?
The linker is a nodevisitor, and is run from the animation manager (it's
an update callback) this visitor will traverse child and link what is
needed to be linked. It will do this operation if it's tagged 'need
relink', this state are set when a deep copy occur.
The linker will create the link between animation manager and Target in
update callback own by a node or other object. Those target can be
shared between different update callback (eg you can animate an osgText
or a color with the same target).
Once the linker has done its work, node or objects are updated correctly
corresponding to the animation and ready to be updated.
The linker to make its job needs a kind of convention about the update
callback name to be able to link with something eg:
If an animation contains a channel named diffuse, the update callback
that are 'compatible' with this channel will do something like that:
bool UpdateMaterial::link(osgAnimation::Channel* channel)
{
if (channel-getName().find(diffuse) != std::string::npos)
return channel-setTarget(_diffuse.get());
return false;
}
The same principle is used for other type. But if you want to skip this
behaviour and do your link by hand you will have to fix your link by
hand when you do a deep copy to maintain the shared target inside update
callback ...
That's the case of Michael, because the collada reader seems not able to
use the linker.
I hope it will helps to understand better, but if it's not enough clear
dont hesitate to ask. I will try to keep the doc of osgAnimation with
this kind of post in order people can understand how it works.
Cheers,
Cedric
-
+33 659 598 614 Cedric Pinson mailto:cedric.pin...@plopbyte.net
http://www.plopbyte.net
On Fri, 2009-09-25 at 16:37 +0100, Robert Osfield wrote:
HI Michael,
On Fri, Sep 25, 2009 at 4:02 PM, Michael Platings
mplati...@pixelpower.com wrote:
While working on osgAnimation I've encountered the following problem: if an
object is referenced multiple times in a scene graph, and the scene graph is
deep cloned, the scene graph clone will contain a different copy of that
object for each reference. I think so far this hasn't been a problem because
having multiple copies of most objects doesn't matter other than taking up
extra memory, but in osgAnimation meshes won't animate correctly if they're
not linked to the right bones so this has become an issue.
A solution to this could be to add a map of old objects to new objects into
CopyOp, and in operator() only call clone() on an object if it hasn't
already been cloned. I've implemented this and it works fine, but before I
start to rely on this, would such a change be acceptable to you?
This change wouldn't be a clone() method, it'd be a
shareASingleClone() method which is a totally different thing and in
the general case would not be safe or appropriate a replacement for
clone().
It sounds to me like osgAnimation's support for cloning is suspect, as
one should be able to deep copy a subgraph with an animation and have
it still work. If it's not possible then a shallow copy should be
enforced. I'm not familiar with the internals of osgAnimation so l'll
have to leave this to Cedric to consider.
Cheers,
Robert.
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
signature.asc
Description: This is a digitally signed message part
___
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
[osg-users] [osgCompute] PostProcessGL
First of all, great work on bring CUDA to the OSG! I'm just getting schooled up on it right now with the main goal being to duplicate what the PostProcessGL example in the sdk is doing. According to their comments, this requires the following: 1 - render the scene to the framebuffer 2 - copy the image to a PBO (pixel buffer object) 3 - map this PBO so that its memory is accessible from CUDA 4 - run CUDA to process the image, writing to memory mapped from a second PBO 6 - copy from result PBO to a texture 7 - display the texture I've mainly struggling with steps 1,2, and 3. What's the best way to do this? Michael Guerrero -- Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=17662#17662 ___ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
