Hello Wang,
Thank you for your submission. Yesterday I've compiled it successfully
against osg 3.1.2.
However performance of the vdsm with blurring was quite depressing
(~4fps with the standard shadow scene, two shadow maps with 1024x1024)
on a quadro NVS.
Also I had some nasty artifacts when moving the scene via trackball. The
shadow map starts to flicker like it is a frame behind.
Anyways, what I'm really interested in, is the new framework, as I want
to use it in my deferred pipeline.
Could you briefly explain how to bind the shadow pass to a certain
subgraph, define the main camera and set up the textures in this case?
How do I access the shadow matrices afterwards (I really want to do the
mapping part myself)
cheers
Sebastian
Hi Robert,
I'd like to submit my changes to ViewDependShadowMap, which will
mainly support Parallel-Split shadow map (PSSM), and support exact
customizing of the entire shadow mapping process. The osgshadow
example is also modified to support all new features, with a simple
Variance shadow map (VSM) implementation, and new debugging
functionalities for displaying HUD shadow textures and light space
frustum. Some old interface are also replaced by new ones.
Developers can find in the CustomShadowMap.cpp about how to
use ShadowSettings callbacks to customize their own shadow techniques,
including defining shadow camera/textures, rewriting caster/receiver
shaders, computing own split distances, and adding extra post
processors (for fullscreen shadow blurring). It is even possible now
to redirect the shadowing result to a deferred pipeline.
For a first glare, you may run osgshadow with the following arguments:
Shadowed scene with PSSM, compatible with both fixed pipeline and shaders
./osgshadow --vdsm --num-sm 3
Custom VSM implementation, compatible with PSSM
./osgshadow --vdsm --custom --num-sm 3
Custom VSM with debugging information
./osgshadow --vdsm --custom --debugHUD --debug-frustum
Future plans include efficiency improvement, SDSM implementation (much
better than PSSM) and other optimization we may think out.
I believe most features and their combinations are proved to be usable
in my own tests and client work, but still with finger crossed to see
if it could be tested and merged, and would be helpful to the whole
community. And feedback and suggestions are always welcome. :-)
Cheers,
Wang Rui
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
_______________________________________________
osg-submissions mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
