Hi All,

Daniel Wickeroth wrote:
> Hi,
> 
> I ran another test last night. This time with Sync to Vblank turned
> off. I also changed the performance  settings of the graphics card to
> full performance, instead of quality.
> 
> The results can be found here : http://vis.uni-koeln.de/FrameTests/Test3.htm
> 
> This time I also used OpenSG 1.8 to load obj files. Like Eric, I also
> found, that the obj files are more performant than the wrl files.

I guess that's because they are single-index, while the WRL files keep the VRML 
multi-index unless explicitly told not to.

> Compared to OpenSG 1.6 the obj files need much less time to load, but
> it still takes extremely long compared to OSG (up to 3,5 hours).

I looked into that and kludged it a little bit, the tie486.obj now loads in ~10 
minutes on my box. Much faster, but still too slow for the model.

The main problem is that the half edge structure that is used for striping is 
not designed for geometries that only use a very small piece of the position 
property (it uses the vertex index as an index into a vector of structures, and 
ends up spending 99% of the time initializing the elements of this vector that 
are never used).

I don't want to switch it to a map, as that would penalize the 99% of models 
that use the positions as expected. The solution would be to find the lowest 
index actually used and use that as an offset for the vector, but that's too 
big 
of a change for me, I'll leave that to Johannes.

Johannes, do you think you can take a look at that in the next few days? If 
not, 
I would leave this until after 1.8. Given that this hasn't been noted so far, I 
assume not many people use very large obj files.

> I also tried to load the obj files eric posted into the applications
> that I posted earlier. Both files loaded and were displayed correctly
> in OpenSG 1.8 as well as OSG 1.2. When loading sphere-UVs.obj into
> OpenSG 1.8 I received some Warnings like:
> WARNING: Unknown obj data elem 0.0000000000
> 
> Apart from that there were no warnings or error messages.
> 
> I will rerun the tests with other models in the next days and post more 
> results.

I added support for the "d" option in material files, that was not used so far.

Yours

        Dirk

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Opensg-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensg-users

Reply via email to