Hi Kai,
Disclaimer: I have not done anything like this. I did think about
implementing such technique, though. And I am deeply interested in the
results You could get. So I will try to provide few unproven suggestions in
hope you will share them later ;-)
1. The way you attach texture layers will not work. This way you will have
only one buffer active. Camera::atach in your usage scenario cause OSG to
set up FBO using
glFramebufferTextureLayer( GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT,
your_tex_array_gl_id, 0, layer ). Layer in this context does not define
FRAMEBUFFER layer but layer of texture 2D array. Effectively you call attach
3 times but two first calls are overriden with third so finally your three
calls:
shadowCamera->attach(osg::Camera::COLOR_BUFFER, ta, 0, 0); //
shadowCamera->attach(osg::Camera::COLOR_BUFFER, ta, 0, 1);
shadowCamera->attach(osg::Camera::COLOR_BUFFER, ta, 0, 2);
are effectively equal to last one:
shadowCamera->attach(osg::Camera::COLOR_BUFFER, ta, 0, 2);
I guess you may want to replace it with following three calls:
shadowCamera->attach(osg::Camera::COLOR_BUFFER0, ta, 0, 0);
shadowCamera->attach(osg::Camera::COLOR_BUFFER1, ta, 0, 1);
shadowCamera->attach(osg::Camera::COLOR_BUFFER2, ta, 0, 2);
this will set up multiple render targets with 3 independent color buffers.
However, I am not sure if this will propely work with gl_Layer produced in
Geometry shader. These OpenGL examples I saw usually set up layered
framebuffer rendering with one glFramebufferTexture call. For example see
this link (its sample code to render 6 faces of cubemap simultaneously)
http://pastie.org/796448: However, this brings another problem:
glFramebufferTexture functinality is not used by OSG yet.
2. Even assuming that you may set each layer as multiple render targets
using above suggestion I see another problem: Shadow maps are generated into
depth maps and hence you should rather set up DEPTH_BUFFERS insted of
COLOR_BUFFERS . I am not sure if support for multiple depth buffer targets
is already available for OpenGL (I know its present in recent Direct3D). You
will have to check it. But even if its available in OpenGL its not yet
supported by OSG.
3. On few forums I saw that people trying to set similar up rendering were
able to get it to work on COLOR_BUFFERS but even then only one DEPTH_BUFFER
layer was used. It looked like bugs in the drivers.
So the prospects are not too bright. I looks like you need to modify OSG FBO
code to get what your are after. I have no idea, how deep these
modifications should be....
In one of my concepts, instead of depth buffers I was considering using one
32 floating point color buffer (each channel would represent one depth
layer) with BlendEquation set to RGBA_MIN. This way one could effectively
replicate depth buffer behaviour with color buffer. Early Z -test is gone
though....
Dont forget to post your results ;-)
Cheers,
Wojtek Lewandowski
----- Original Message -----
From: "Kai Xia" <[email protected]>
To: <[email protected]>
Sent: Monday, March 08, 2010 9:01 AM
Subject: [osg-users] problems about layer rendering
Hi,
I have some problems while using geometry shader and TextureArray to
optimize PSSM shadow, the test code is like following:
C++ code:
osg::TextureArray *ta = new osg::TextureArray;
ta->setTextureSize(1024, 1024, 3); //3 layer texture
...
shadowCamera->attach(osg::Camera::COLOR_BUFFER, ta, 0, 0); //
shadowCamera->attach(osg::Camera::COLOR_BUFFER, ta, 0, 1);
shadowCamera->attach(osg::Camera::COLOR_BUFFER, ta, 0, 2);
Geometry shader code:
#extension GL_EXT_geometry_shader4 : require
varying vec4 color;
void main()
{
//Route to Layer 0
for (int i = 0; i < 3; i++)
{
// You will recieve 3 positions since we set the input type to
Triangles
gl_Position = 0.5*gl_PositionIn[i] - vec4(0.5, 0.0, 0.0, 1.0);
gl_FrontColor = vec4(1.0, 0.0, 0.0,1.0);
gl_Layer = 0;
EmitVertex();
}
EndPrimitive();
//Route to Layer 1
for (int i = 0; i < 3; i++)
{
gl_Position = 0.5*gl_PositionIn[i];
//Just to see a difference in Layer 1
gl_FrontColor = vec4(0.0, 1.0, 0.0,1.0);
gl_Layer = 1;
EmitVertex();
}
EndPrimitive();
//Route to Layer 2
for (int i = 0; i < 3; i++)
{
gl_Position = 0.5*gl_PositionIn[i] + vec4(0.5, 0.0, 0.0, 1.0);
//Just to see a difference in Layer 1
gl_FrontColor = vec4(0.0, 1.0, 0.0,1.0);
gl_Layer = 2;
EmitVertex();
}
EndPrimitive();
}
during the testing , only the 1st layer of texture have avaliable
context(scene with red color), other 2 layers are undefined. Has some one
sucessed in layer rendering? please give some advise, thanks a lot!
Cheers,
xiakai
------------------
Read this topic online here:
http://forum.openscenegraph.org/viewtopic.php?p=25314#25314
_______________________________________________
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
