Hi Markus,

Thanks for the benchmark results.  Its nice to see CVS outperforming
1.0.. at least with the old plugin :-)  Now the job is to pick up on
what is causing the problem with the new data.

Could you run the old and the new flight plugins through osgconv with :

 export OSG_NOTIFY_LEVEL=INFO
 osgconv myfile.flt new.ive &> new.txt

And then post the new.txt and old.txt.  You can just send this to me
as this will be rather large.   In the debug output should be the
values of number groups,transforms etc.

Robert.

On 7/18/06, Markus Trenkwalder <[EMAIL PROTECTED]> wrote:
Robert Osfield wrote:
> On 7/17/06, Markus Trenkwalder <[EMAIL PROTECTED]> wrote:
>> Oh, sorry.  It seems that I did not make myself clear enough: I used the
>> .ive database in both cases, but the one converted with the 1.1-rc2
>> version of osgconv is 1/3 slower than the one converted with the 1.0
>> version of osgconv.
>
> You you try selecting the old .flt plugin and redoing the tests:
>
> osgcov myfile.flt new.ive
>
> export OSG_OPEN_FLIGHT_PLUGIN=old
> osgconv myfile.flt old.ive
>
> And to revert to the new plugin
>
>  export OSG_OPEN_FLIGHT_PLUGIN=new
>  osgconv myfile.flt old.ive
>
> Note export changes to set or setenv depending upon your platform.

I've done this now and here are the results:

Position 1:
OSG version | fps | Drawables | Vertices
------------+-----+-----------+----------
1.0         | 30  | 5321      | 226508
cvs-old     | 33  | 5321      | 226508
cvs-new     | 23  | 5320      | 226503

Position 2:
OSG version | fps | Drawables | Vertices
------------+-----+-----------+----------
1.0         | 50  | 3037      | 179605
cvs-old     | 51  | 3037      | 179605
cvs-new     | 40  | 3033      | 179579

Platform is win32 (WinXP), OSG was compiled with VS7.1.  No extra
environment settings for OSG are done.  The viewer was in any test the
same and was built with osg-cvs.  The databases converted with 1.0 and
osg-old are of exactly the same size (275369100 bytes), the one
converted with osg-new is smaller (269215442 bytes).

> The thing is to isolate what is different in the new.ive vs old.ive to
> determine what needs to be optimized.  For instance does you database
> contain lots of MatrixTransform or Billboards?

There are lots of Transforms because the .flt-database consists of about
350 flight files collected in one master file.  We do not use billboards
but for example the first scene has lots of trees with alpha textures
which leads to a high depth complexity.

I can not provide you with the ive-files because these are production
databases which I'm not allowed to hand over but I could provide you
with the differences of the two .osg files if this would be of any help.

Markus
--
Dipl.-Ing. Markus Trenkwalder
Software Architecture & Development
  _________
(___  ____)        www.s-p.cc
  ___) S&P Simmulationstechnik
   Phone:  +43 (0)512 393255
   Fax: +43 (0)512 393255 33

_______________________________________________
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/

Reply via email to