> The only gain from only using indexed triangle list could be from the
larger batching that you can achieve i.e you can use a single draw
elements call rather than multiple draw elements calls required for
seperate tri strips.
Actually, this seems not true.
If you generate a plain good old sphere, you will approximatively have
twice more triangles than vertices.
In theory, that could give you 1 triangle per 0.5 transformed vertices.
With a tristrip, on the opposite, you couldn't achieve better than 1
triangle per transformed vertices
since you always introduce a new "uncached" vertex for each triangle.
I suppose you already know the link, but just in case of:
http://tomsdxfaq.blogspot.com/2005_12_01_tomsdxfaq_archive.html#113546436585770597
Igor.
Robert Osfield wrote:
Hi Igor,
On 9/5/06, *Igor Kravtchenko* <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
yes indeed, but the OSG tri stripper always outputs tristrip and not
just plain good old triangles list.
It does generated index tri strip lists though, and as sets of tri
strips that are too short are placed into index triangle lists.
The only gain from only using indexed triangle list could be from the
larger batching that you can achieve i.e you can use a single draw
elements call rather than multiple draw elements calls required for
seperate tri strips.
Although not implemented, there is also the possibility of using
repeated indices to create zero area triangles, these could then act
to join together tri strips to allow them to be sent as a single tri
strip primitive. You should be able to use less indices than you
need be sending all triangles down so I wouldn have though tri strips
will still have a performance gain over index triangles, unless those
zero area triangles clog up the triangle setup part of the GPU.
About the cache size, unfortunately GL nor D3D doesn't provide this
information (in D3D it's just
some kind of super secret information you don't have access to).
Maybe free Linux drivers do (as we have access to the source source)?
Do not know here and would have to investigate this
The best guide is the docs from ATI and NVidia. There isn't any
runtime querry unfortunately. Perhaps you might be able to write a
test to see what the performance characteristics are like, but
personally I think a database of GPU to cache size might be the easist.
Robert.
------------------------------------------------------------------------
_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/
_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/
