On Wed, 2008-04-16 at 15:43 +0100, Robert Osfield wrote:
> On Wed, Apr 16, 2008 at 3:30 PM, Jeremy Moles <[EMAIL PROTECTED]> wrote:
> > My question, then, is twofold:
> >
> > 1. Given the above (and the fact that I am able to modify the
> > osg::Array objects within the Widget directly and see immediate
> > results), is it safe to say I am forcing the drawing implementation to
> > use the slowest, "immediate" mode for rendering? I've always assumed
> > this, and have been comfortable with that during development.
>
> Immediate mode can still use vertex arrays, this will typically be
> slower than using display lists. As long as you don't use per
> primitive normals/colours and vertex indices you'll still be using
> OpenGL fast paths (vertex arrays) rather than slow paths
> (glBegin/glVertex/glEnd) so it needn't be a major issue.
>
> Also one needs a bit of perspective, you are unlikely to have
> thousands of GUI elements all on screen at one time - so cull and draw
> dispatch are very unlikely to be bottlenecks - fill rate is much more
> likely a limit with GUI's with the overdraw associated with building
> up layers, even then its still not likely to impact performance too
> much on modern machines.
You're right about this, of course. Thanks. :)
> > 2. If the above is true, consider the following situation: a user
> > creates his interface element, positions, resizes, colors it, etc. When
> > he's done with this he knows that, for the most part, this particular
> > element will remain static and require no further updates. Thus, there
> > could be some benefit then in asking this object to "reconfigure" itself
> > to use display lists or VBOs. Is this possible? I'd call this method
> > "locking" or similar, seeing as conceptually it would prevent you from
> > making any direction geometry modifications as long as the object was
> > static. However, I'm not entirely sure I have a full understanding of
> > the process. Is it as easy as calling:
> >
> > setDataVariance()
> > set{UseVertextBufferObjects/UseDisplayList}()
> >
> > ...or is there something more complicated required?
>
> Are you not overcomplicating things. If people use the code correctly
> it'll work correctly, if they abuse it then the results are undefined.
> Trying to add extra complexities to try and avoid all eventualities
> can lead to far more problems than it solves.
>
> > P. S. Is it lunch time yet? :)
>
> That's a third question. And yes, it's almost always lunch time
> somewhere in the world ;-)
>
_______________________________________________
osg-users mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
