Vertex Cache Optimisation

[Kojack]

It was something to play with during my vacation. I haven't had time to go back and fix it up. I'm not even sure where the source code is now.

Have a look at OgreTootle: http://www.ogre3d.org/forums/viewtopic.php?f=11&t=29760
and Sinbad's Tootle additions to mesh magick: http://www.ogre3d.org/forums/viewtopic.php?f=1&t=55742
Both use the ATI Tootle library which does vertex cache optimisation as well as overdraw optimisation.

[AshMcConnell]

Hi Folks,

This topic reminded me to try tootle (using Mesh Magick), but I got odd results (at least when I was using the default options)

I benchmarked before and after the meshes were "tootled"

Before: ~325
After: ~299

I reverted to the original meshes and the FPS returned to 325. I know it's not a huge difference, but I was hoping for an increase. Perhaps it is to do with me using an abnormal up vector (+Z)?

Any ideas what I could be doing wrong? Perhaps it just isn't suited by my model (an F1-style racing car)

All the best,
Ash

[Kojack]

Without knowning the exact details of the vertex cache (size, algorithm, etc, which are going to vary over different gpus), the optimisation is just a best guess. Typically it should make bad meshes better, but if the mesh is already in the best case order (perhaps your modelling tool or exporter already did an even more efficient optimisation), these optimisations can make it worse.

[bpaberg]

Yep, thanks Kojack,

I tried Tootle via MeshMagic, it seemed to give some improvement compared to MeshMagic optimize on a couple of big meshes I tried with.

And I don't really expect that Toms method would give much different result after reading this article : http://gameangst.com/?p=9

Cheers,
/PÅ