Hi,
Did a quick test, exporting a fbx with one animated transform/shape per piece works fine, kinda wondering if it would be more efficient to export only object and bind the pieces to bones (one bone per piece), since transform could then be done on the gpu ?
Do you think it’s worth the trouble ? (export wouldn’t take long to script though)
Cheers,
Vincent
Skeletal deformation will definitely perform better than individual objects being transformed. Whether it’s worth the trouble depends on how complex your scene is. You will likely not see a difference if you don’t have too many objects, but you will if the scene gets complex enough.
Ok, thanks for the confirmation, will do some more tests and see how it goes!
Cheers
Hi Malcom,
Just read what you wrote in this topic
viewtopic.php?f=4&t=3786
and now I’m confused!
If COMP transforms are done on the GPU, then should be the same performance to have lots of objects (as in lots of geo COMPs) with animated transforms or only one object/geo comp rigidly skinned (meaning each point is bound to one transform maximum) to lots of transforms/bones no ?
I was about to ask if instancing an object with only one instance was more efficient than using the builtin transform parameters on a geo comp (which I thought were done on the cpu), but now I have my answer!
Still confused about your previous answer though.
Cheers
Vincent
There is a significant overhead involved with issueing a draw call, so the less SOPs you have rendering the better. Skinning is better for that reason, since youll have many transforms all affecting different parts of a single SOP.
Ok! Makes sense.
Then I was wondering if you could share the TouchDeform() code or if it’s already available somewhere ? Would be interested in trying to offset animation for instanced animated skinned geos (See my other topic, sorry for the double post)
Wiki (derivative.ca/wiki077/index. … _(Skinning) says
Thanks a lot!
Vincent