Sorry, didn't want to give an
oversimplified answer,
but I also found this.. (from Mathaeus) using Multithreaded ICE :)
http://www.si-community.com/community/viewtopic.php?f=41&t=1199
a small tut about transferring
attributes, even [if] vertex IDs doesn't match, but some parts
of geometries still are on 'identical' positions.
It's usable when ever you want to work on only a part of
object - various facial rigs, for example.
It's similar method with GATOR, but it can transfer any type
of deformation. From my small experience, it's faster than
GATOR.
[...]
On 10/27/14 13:50, Eric Thivierge wrote:
Thanks
Cesar,
The merge itself is relatively fast. Transferring the attributes
is what is slow. UVs are super quick but skinning is the culprit.
I don't have a low res cage as I need to maintain the weighting
that is there pretty much 1 to 1. What is appearing to be the
problem is that there are too many inputs to the transfer op.
Lower numbers of inputs seems to be quicker (I need to do some
solid tests on this). My strategy at this point is to partition
the group into sub groups and merge in batches building up to the
final fully merged set of meshes. We'll see how this goes.
It is really simple honestly. Create 436 objects of varying point
counts that add up to 161,074 points with 172 deformers. Try to
merge all of them at once then use the UI to selectively transfer
the UVs and Envelopes. Envelopes will take forever.
Eric T.
On Monday, October 27, 2014 1:43:52 PM, Cesar Saez wrote:
Hi Eric,
A couple of things to keep in mind (I'm sure you already went
over
these, but just in case):
- Merge is expensive, make sure to merge your groups at once
instead
of add part by part through multiple merge operations.
- Transfer weights from a single cage mesh works way better than
from
small/partial parts, if you don't have a cage try merging
everything
together once and use that as source for your weight transfer.
- FreezeM all the things! looks like you're dealing with
expensive
operations (merge and gatoring) + dense meshes, FreezeM between
operations makes a huge difference!
Maybe you can share a pseudo-code/outline of how you're
approaching
this, that way would be easier to spot possible sources of
slowness.
Cheers!
|