Hi Wojtek,
I think if it gives a verifiable gain (less video memory used for
depth component RTT) then you can submit it.
I saw direct improvement on GF 7800. On newer boards its smaller unless
the OSG application is GPU memory bound. Which happens quite often in
our practice.
But even if the application is not video memory bound, using up less
video memory means it will be less likely to become video memory bound
in the future... I meant verifiable in terms of using a tool that shows
how much video memory is used, before and after applying your changes.
An app should use as much video memory as it can if it improves
performance, but in this case I really don't think it improves anything
to use more memory.
Indeed, at least on cards with 512MB video memory you will be more
likely to be able to use 4096x4096 shadow map size...
Code attached. Changes are quite simple. You may as well test them.
Considering Robert tight schedules I doubt he will be eager to play with
it.
You're right, it's a simple change. I wonder if we could get some other
people to test it with some of their apps and see if it causes problems.
Looking at the code, it seems like the same thing happens with the depth
buffer, i.e. a depth attachment will be created if none is present.
IMHO, a depth buffer is not mandatory - you could want to do an RTT pass
on a single fullscreen quad, in which case you don't need depth and can
disable depth test. In that case, wouldn't it make sense to only attach
a color buffer? An added depth buffer will only add to the video memory
requirements.
In other words, can't we trust the user to have set up their FBO
attachments correctly? In these two cases, the code checks if there's a
given attachment (depth or color) and if not, unconditionally adds one
whether the user needs it or not. The code can't know exactly what the
user needs, but isn't it better to be conservative instead of creating
everything in case the user needs it but doesn't know it?
I hope I'm making sense... Anyways, just for the shadow map, I think it
would be interesting to at least see if this change affects something else.
Thanks,
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
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
