Hi Robert,
yes indeed, but the OSG tri stripper always outputs tristrip and not
just plain good old triangles list.
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
Igor.
Robert Osfield wrote:
FYI, The tri stripper the OSG uses does has support for tweaking the
vertex cache it builds for. It currently defaults to 16. The fast
tristripping performance in theory could allow you to run it are start
up with the ideal cache size for the hardware, however, I've not yet
gone this route.
On 9/5/06, *Igor Kravtchenko* <[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>> wrote:
Hi,
yes I was aware of that thread (I'm on GDA too).
I'm not really interested indeed to have a tristrip but rather a
geometry
that is just vertex cache friendly.
What pleased me with NVTriStrip is that it can leave plan triangles
list if required.
D3DX has two functionalities, one to optimize vertex cache being
device
independent, one another method being device dependent.
The device independent seems to use a 12 cache entries.
I suppose the device dependent is more optimal though.
I've also studied the Tom Forsyth's method he explained and
that they used in Granny. Hoppe's method seems good but
more complex to implement.
Maybe I could merely use the D3DX device independent method
as an offline preprocess and resave our meshes.
Have to find something.
Thanks all,
Igor.
Smeenk, R.J.M. (Roland) wrote:
> There's currently a discussion going on on the gdalgorithms list
about
> the usefulness of tristripping nowadays.
> http://sourceforge.net/mailarchive/message.php?msg_id=36714138
<http://sourceforge.net/mailarchive/message.php?msg_id=36714138>
>
> The tendency is to don't bother with tristripping at all and purely
> try to optimize for vertex cache usage. Preferably the vertex cache
> optimization should be done independent of the target hardware
vertex
> cache size.
>
> Here's a number of interesting links about optimizing for vertex
cache.
>
> Vertex cache optimization, NVidia, Summer Camp 1999
> http://developer.nvidia.com/object/vertex_cache_opt.html
>
> Universal Rendering Sequences for Transparent Vertex Caching of
> Progressive Meshes
> http://www.cs.technion.ac.il/~gotsman/caching/
<http://www.cs.technion.ac.il/%7Egotsman/caching/>
> <http://www.cs.technion.ac.il/%7Egotsman/caching/>]
> Vertex cache optimisation, TomForsyth's blog Saturday,
December 24, 2005
> http://tomsdxfaq.blogspot.com
>
> Optimization of mesh locality for transparent vertex caching,
> HuguesHoppe, ACM SIGGRAPH 1999, 269-276
> http://research.microsoft.com/%7Ehoppe/
> Triangle Order Optimization for Graphics Hardware Computation
Culling
> http://pedrosander.com/publications/
> kind regards,
>
> Roland Smeenk
>
>
>
------------------------------------------------------------------------
> *From:* [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
> [mailto:[EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>] *On Behalf Of *Robert
> Osfield
> *Sent:* maandag 4 september 2006 20:33
> *To:* osg users
> *Subject:* Re: [osg-users] Geometry Stripifier
>
> Hi Igor,
>
> I moved across to Tanguy's code because it was faster, and more
robust
> and appeared at the time to provide just as efficient tristripping.
>
> It is worth noting that when I originally moved the code across
I was
> getting good performance improvements with tri stripping, doing so
> more recently, say last year/six months the results have been
rather
> mixed, sometimes faster, sometimes slower. I haven't had a
chance to
> investigate. It might be that I've just tested it on a wider
range of
> code, it could be that in fixing a bug we've inhiteched something
> else, or it could simply be that OpenGL drivers and hardware have
> changed and have a different set of senstivities w.r.t performance.
>
> Robert.
>
> On 9/4/06, *Igor Kravtchenko* < [EMAIL PROTECTED]
<mailto:[EMAIL PROTECTED]>
> <mailto:[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>>> wrote:
>
> Hi!
>
> I had a look at the TriStrip visitor of osgUtil based on the
Tanguy
> Fautre's code and
> I just wondered what kind of benefit it has over the
NVTriStrip's
> (NVidia) code.
> Supposed to be more performant, faster?
>
> Stop me if I'm wrong, but the Tanguy's code seems to generated
> only tristrip
> whereas the NVidia's code can stay in trilist if required.
>
> Any remark is welcome,
>
> Igor.
>
> _______________________________________________
> osg-users mailing list
> [email protected]
<mailto:[email protected]>
<mailto:[email protected]
<mailto:[email protected]>>
> http://openscenegraph.net/mailman/listinfo/osg-users
> <http://openscenegraph.net/mailman/listinfo/osg-users
<http://openscenegraph.net/mailman/listinfo/osg-users>>
> http://www.openscenegraph.org/
>
>
>
> This e-mail and its contents are subject to the DISCLAIMER at
> http://www.tno.nl/disclaimer/email.html
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>osg-users mailing list
>[email protected] <mailto:[email protected]>
>http://openscenegraph.net/mailman/listinfo/osg-users
> http://www.openscenegraph.org/
>
_______________________________________________
osg-users mailing list
[email protected] <mailto:[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/
_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/
