Hi Adrian,
I bet you based your current PSSM on osgShadow::ShadowMap code.
osgShadow::StandardShadowMap is functional equivalent of osgShadow::ShadowMap.
Its associated osgShadow::StandardShadowMap::ViewData object cull method
performs similar steps to osgShadow::ShadowMap::cull. The difference is that I
have dividied steps in the cull method into separate virtual functions.
Please have a look at these steps in StandardShadowMap::ViewData::cull()
method. They correspond to osgShadow cull functionality so you may learn which
functions you will need to override. Additionally MinimalShadowMap::ViewData
introduces frameShadowCastingCamera method which is used by all
MinimalXXXShadowMap & LispSM classes to fine tune ShadowMap projection matrix.
This tunning means that projection matrix is optimized to contain shadow volume
with minimal overhead. Actual LispSM matrix application is done during
frameShadowCastingCamera.
LIspSM is made using decorator design pattern. Actual class implementation is
derived from ProjectedShadowMap template class based on one of
MinimalXXXShadowMap classes and LispSM algorithm class which is used for light
space perspective projection.
If I was to create hybrid PSSM/LispSM I would also derive from
ProjectedShadowMap but would create special LispSM/PSSM algorithm. But
additionally one would need to override StandardShadowMap::ViewData methods to
fork cull traversal into a number of render stages and generate texgens needed
for each PSSM shadow map.
As I write this explanation I realize that whole topic might be actually very
complex. I wish the code was bit simpler but I wanted to create a grid of
solutions that could cooperate or be used alternatively so the framework has a
lot of internal crossdependencies which could make it into difficult labirynth.
I tried to make it clean and elegant but I am afraid I had to sacrifice
simplicity for multithreaded/multimonitor use in the same time providing number
of methods of shadow volume optimizations. I hope you will manage to pass
through this maze somehow ;-).
Cheers,
Wojtek
----- Original Message -----
From: Adrian Egli OpenSceneGraph (3D)
To: OpenSceneGraph Users
Sent: Monday, October 06, 2008 1:51 PM
Subject: Re: [osg-users] Lispsm / PSSM merge ?
Hi Wojtek,
the idea of PSSM is to split the view frustum into n-parts. then for each of
such sub volume (frustum) we have to render one shadow map and apply the shadow
for such a volume. or in other words, we like to get one texture (shadow map)
for one sub-frustum starting at near (plane : split i) and end at far (plane :
split i + 1). in this sub frustum the shadow map [i,i+1] is active, and this
map is used to render the shadow. so if we can define in the Lispsm a range
[i,i+1] /// [near,far] then we have one split for the pssm. next we need many
such splits, and finally we get the pssm working (many for n splits from
near_frustum = near_0, far_frustum = far_n
[near_0, far_0][near_1=far_0,far_1] [far_1,...] [far_n-1,far_n = far_frustum]
would this be easy to do? multiple shadow technic nodes in one scene to get
such maps.
/adrian
2008/10/6 Wojciech Lewandowski <[EMAIL PROTECTED]>
Adrian,
I am afraid I don't understand what you want to do. Could you provide bit
more explicit explanation ? At higher level of abstraction using LispSM with
PSSM should be possible but I am not that familiar with PSSM implementations to
judge whether it is easy or difficult task to merge two approaches.
If you closely look at LightPerspectiveShadowMap class you will find that
it basicaly overrides one method of ProjectionShadowMap template class. Rest of
heavy work is done by underlying MinimalXXXShadowMaps. These classes attempt to
compute minimal shadowed volume (as ConvexPolyhedron - polytope like object).
This volume could be used by modified PSSM instead of view frustum. This
computed volume is passed as one of the input arguments to the method that
LispSM overrrides.
Cheers,
Wojtek
----- Original Message -----
From: Adrian Egli OpenSceneGraph (3D)
To: OpenSceneGraph Users
Sent: Monday, October 06, 2008 11:40 AM
Subject: [osg-users] Lispsm / PSSM merge ?
Hi all
The osgShadow lib becomes a really nice shadow toolkit, thanks all to
work out on different shadow technics. As we all understand is that the PSSM
would be one of the most
robust shadowing technic for large terrain, now we should be able to
merge the lispsm with the pssm idea:
would it be possible to add one feature to lispsm:
start shadow at near plane and stop shadow at far plane, then we will
need a combiner of the lispsm shadow, then we gets pssm idea back. would this
be possible to do, or do we need to rewrite the whole pssm stuff?
would it be possible to add some filter into the GLSL shader code for
shadows: pcf ... ?
adrian
--
********************************************
Adrian Egli
--------------------------------------------------------------------------
_______________________________________________
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
--
********************************************
Adrian Egli
------------------------------------------------------------------------------
_______________________________________________
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
