Re: tee to bilboard

New Message Reply Date view Thread view Subject view Author view

Angus Dorbie (dorbie++at++multipass)
Fri, 21 Nov 1997 14:02:38 -0800


I have some code which improves on the billboard performance for many
trees.

Consider the typical billboard in Performer, you have several geosets
which billboard but each geoset has at least one glBegin glEnd to
render.

This is undesirable if you just want to draw a bunch of quads for
example when billboarding trees so a better approach for this case
is to have a geoset with PFGS_QUADS and billboard the vertex
attributes. This gives much better graphics performance.

Additional optimizations include sparse updates of billboarded
geometry since they don't move very much frame to frame, group
vector billboarding in the distance and sparse group updates
in the distance since these move less w.r.t the viewer frame
to frame.

I can get .5 Mtrees/sec on an Octane MXI, each tree being a quad,
I'm sure if you had a coniferous forrest you could improve on this.

Aditionally you probably want to billboard to a vehicle, not a
channel, so the code does this, although multiple vehicles would
require some channel masking and more compute.

You also don't need multibuffer billboarding, like I said, trees
don't rotate much, so I save some attribute memory also.

I'm back now so if anyone wants the code I can post it,
next week (I'm in email purgatory right now).

I'll get the requested showcase presentation on large
coordinate management to everyone who asked, sorry for the
delay, please don't request again or I'll send it twice
(and it's big).

Cheers,Angus.

On Nov 17, 2:06pm, Steve Baker wrote:
> Subject: RE: tee to bilboard
> On Mon, 17 Nov 1997, Andy Walker wrote:
>
> > Garrett
> >
> > I have found crossed trees to be faster. Billboard trees stress the CPU
> > more than crossed trees. You are really only adding one polygon times
> > the number of trees when using crossed trees. It really depends upon
> > your training requirements and our interoperability range with the
> > trees. If your application is a tank simulator for example you may want
> > to spend more of your hardware and polygon budget on rendering
> > billboard since they will add visual realism.
>
> Yep - that's been my finding. Also:
>
> 1) Don't use billboards with stereo viewing (they rotate differently
> for left and right eyes - and give me a headache).
> 2) Don't use billboards with wide field-of-view displays because
> it's too easy to spot the trees rotating as you drive past them.
>
> > I have played with using billboards for trees for my high level of
> > detail and having them swith to crossed trees for my lower resolution.
> > This seems to cause a few more state changes than I would like to
> > impress upon the system though.
>
> Yep - we tend to add more polygons to the crossed trees at high level
> of detail to make them gradually become more solid - but that's a bad
> thing if you can really get close to them because the pixel fill rates
> can get a bit steep.
>
> Trees and clouds are the hardest things to render well - they are soft,
> fuzzy and volumetric. All the things that polygons arent!
>
>
> Steve Baker 817-619-8776 (Vox/Vox-Mail)
> Hughes Training Inc. 817-619-4028 (Fax)
> 2200 Arlington Downs Road SBaker++at++link.com (eMail)
> Arlington, Texas. TX 76005-6171 SJBaker1++at++airmail.net (Personal eMail)
> http://www.hti.com http://web2.airmail.net/sjbaker1 (personal)
>
> ** Beware of Geeks bearing GIF's. **
>
>
> =======================================================================
> List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
> Submissions: info-performer++at++sgi.com
> Admin. requests: info-performer-request++at++sgi.com
>-- End of excerpt from Steve Baker

=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
            Submissions: info-performer++at++sgi.com
        Admin. requests: info-performer-request++at++sgi.com


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:56:14 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.