Transform packed fragments at render time

[michaelb-01]

I am working on a big destruction job and I’m trying to optimise our workflow as much as possible.

my usual workflow:

- cache a single frame of my fractured geo as packed fragments

- import rbd sim and create point per rbd piece (dop import)

then lighting would import these two caches and use transform pieces. This works well since we only cache a single frame of the fractured geo, and the cache for the rbd points is very small.

however the geo is still written into the ifd file every frame, which can be quite large. 

 

So, is it possible to specify the geo needed for the whole sequence and then have mantra transform the geo at render time (the same as transform pieces). This would make the ifd’s tiny

 

an alternative method is to write each fractured piece out as a separate file then copy empty packed disk primitves to the rbd points then use the unexpandfilename intrinsic to expand it at render time.

This makes tiny and fast ifd files which is great but seems quite slow to render - probably because it has to pull so many files from disk (1000s of pieces)..

is it possible to do the render time transform pieces approach or does anybody have a better method? (The two I’ve mentioned are fine I’m just trying to optimise it!)

[Luijee]

Have the same question, interested to know if you ever found a solution?

We've tried the seperate file per piece approach and it proved easier to manage and give you small ifds, but mantra start up times & input/output seemed to be a bottleneck. 

Edited August 24, 2019 by Luijee

[michaelb-01]

I never did unfortunately.. Let me know if you do!