Particle cache and XSIBatchserver 2005-06-23 - By kim aldis
Back
they're kind of odd, caches. I always found they'd rebuild cache at
rendertime regardless of the state and if I write protected the files it'd
just fall over. I had bloody awful problems when I was messing with this but
I've a feeling it was something I was doing that was messing the cache
rebuild. Using scripted events or doing something in the scripted events.
Doing something recently and rendering not using batchserve and not using
scripted events, I had problems with the cache sometimes, for no apparent
reason, deciding it needed rebuilding. If that happens then wrong results
will happen. Mostly, I think, because you've got multiple machines all
fighting for the same files. Corruption occurs, no doubt about it. I found
myself doing a lot of re-rendering on odd passes, even when I'd vedry
carefully precached and tested everything.
There's two ways to go:
*
let all the machines see the same cache that's built prior to
rendertime. Like I said above, there could - and probably will - be
problems.
*
Alternatively set the cache path to something like c:\temp\foo.
C:/temp needs to exist on all the slaves. This way you'll have the cache
recalculated on all the slaves at rendertime but they will be correct. In
terms of the time it takes to render the frames the cache calculation time
isn't often that great. Probably safest to clean the cache out before
re-rendering, just in case.
The other thing you need to be aware of is that it's possible to have
particle caches named the same in any given scene and that it's highly
likely there'll be identical caches across scenes in the same project. They
fixed the thing - I think - where all particle cache files on all particle
clouds were created with identical names in a scene but now they get named
with sequential names; particle1, particle2, etc. Manging particle cache
filenames can rapidly turn ugly fast.
Pretty much most of this applies to hair caching too.
__ __
From: owner-xsi@(protected) [mailto:owner-xsi@(protected)] On Behalf Of
Marc-Andre Carbonneau
Sent: 23 June 2005 12:56
To: XSI@(protected)
Subject: RE: Particle cache and XSIBatchserver
Once cached, make you ptp files read-only
I believe it works that way. Otherwise, you'll need to render everything on
one machine or several machines with different segment of your timeline.(use
a cloud seed so they are all the same)
__ __
From: owner-xsi@(protected) [mailto:owner-xsi@(protected)] On Behalf Of
kim aldis
Sent: Thursday, June 23, 2005 5:51 AM
To: XSI@(protected)
Subject: RE: Particle cache and XSIBatchserver
As far as I'm aware there's no reason why it shouldn't work. Can you be sure
that the path to the cache files is valid on all your slaves?
__ __
From: owner-xsi@(protected) [mailto:owner-xsi@(protected)] On Behalf Of
guillaume laforge
Sent: 23 June 2005 10:42
To: XSI@(protected)
Subject: Particle cache and XSIBatchserver
Hi list,
I've seen this topic several times on the list but this time we got the pb
in the studio I work actually.
We can't render particules simulation with standard caching. Only standard
no caching works with batchserve.
I'm not a particle specialist and I don't understand why ?
Is it a limitation ?
Cheers,
Guillaume Laforge
|
|
