Thx for sharing your mind on the subject Ciaran. It's added to my to do
list.
- Take a look at the cache on file node more in depth (blending, append
caches...etc)
Has the opposite I let down the cache on file node when I encountered
crashes and corrupted scene. When the cache on file node was sets to
Read, Save scene/close and reopen. XSI wasn't able to open the scene, it
was a crash at each try.
On 8/15/2012 8:21 AM, Ciaran Moloney wrote:
Well I guess the mixer op isn't as bad as I make out - I guess it's
fine for most situations. It's just a bit of a PITA. Mostly I've
encountered problems with some renderers, like Dominik has and in the
early days I found it to be much less reliable when reading in caches
e.g. missing attributes. So I got burned and didn't go back.
Mostly I find that mixer workflow quite limiting. I can have a lot
more control using cache nodes and have just become used to it. e.g. I
can read in multiple caches and do interesting things with the data
between each file being read, I can move the cache op into any
construction region, can do cool tricks with string attributes etc.
But mostly it's just a sense that I know what's going on compared to
the mixer.
On Wed, Aug 15, 2012 at 1:11 PM, Dominik Kirouac <[email protected]
<mailto:[email protected]>> wrote:
Hey Ciaran,
So you don't use the cache manager to load you Sim cache? You use
the cache on file node instead ?
Does it means that you can't take advantage of the cachelist when
you want to load caches for multiple pointclouds ?
On 8/15/2012 8:03 AM, Ciaran Moloney wrote:
Don't use the mixer op (ever).
On Wed, Aug 15, 2012 at 12:54 PM, Dominik Kirouac <[email protected]
<mailto:[email protected]>> wrote:
Hey Chris,
Thanks to bear interest to my problem. I know the way to
cache pointcloud and add attributes. I did it several time. I
think it's more a problem related to voodoo magic :)
I tried to create a repro scn with a simple setup(emit from
geo, set instance shape, simulate particle) using the same
instances that I use in my project, but it don't recreate
the problem.
It seems like I do something in my simulation that break the
link with the instances between the simulation scene and the
render scene. We are 2 here doing simulation on this project
and the other guy encounter the same problem with a different
simulation/scene... but the setup is similar.
I though it could be caused by the use of a post simulation
tree for cropping by camera that maybe mixed up the pointID
or the shape index, but I deleted the post process tree and
the problem was still there.
Forcing the shape by adding another icetree over the MixerOp
on the cached pointcloud works well... in mental ray, but not
in Arnold. So I'm stuck with a icecache bug that can be
solved with a workaround not working in arnold :).
Maybe I can send you my scenes(one simulation and one
render), but I'm using Exocortex Slipstream, do you have a
license to be able to simulate the Icetree?
--
--
- Ciaran
--
