W dniu 2009-11-12 11:52, Chris Marshall pisze:
> 2009/11/12 piotrek marczak <orig...@gmail.com <javascript:_e({}, 'cvml', 'orig...@gmail.com');>>
W dniu 2009-11-14 10:51, Frantisek Hradecky pisze:
Exactly the same approach here. ;)
http://www.czarekkwasny.com/?id=casestudy
(I still need my friend to get the translation corrected)
The nail'n'hammer approach with rendering ten frames per sequence then
having the whole process restarted. It was fun watching stairs-like
memory graph and seeing if it hits the roof or restarts before it
happens. ;) Especially considering that preprocessing (limited to just
one core) took often more than rendering, running 2-4 instances on one
machine was like... regular.
And for disappearing ICE instances. It's quite common experience... But
I was given some really nice feedback regarding this issue on the list. :)
Dan Yargici wrote:
> Don't ask me how but we managed to do this project (all the flocking
> birds) using nothing but offset ShapeInstanceTime trickery.
>
> http://www.glassworks.co.uk/index.php?link=description&job=295
> <http://www.glassworks.co.uk/index.php?link=description&job=295>
> <http://www.glassworks.co.uk/index.php?link=description&job=295>
> I think we were rendering in one frame packets though. So perhaps we
> just got lucky...
>
> DAN
>
>
> On Sat, Nov 14, 2009 at 6:24 PM, Bradley Gabe <with...@gmail.com
> <mailto:with...@gmail.com>> wrote:
>
> Hi Folks. Sorry you are having problems rendering your instances.
>
> There is a known bug with the current production versions of MR that
> pertains to animated instances. During render preprocessing, as MR
> is attempting to sort through all the different timings of instance
> animation and build an array of standins to meet those timing
> offsets ("parsing assembly..."), it builds up in memory and fails to
> flush away enough RAM before hitting the ceiling. You can watch this
> happen if you keep your system resources open while you attempt to
> render.
>
> The specific attribute in ICE that triggers this process and will
> lead to crashes is ShapeInstanceTime. The following factory compound
> nodes are currently NOT reliable for consistent rendering as they
> manipulate ShapeInstanceTime:
>
> * Set Instance Geometry
> * Control Displacement Instance Animation
> * Control Instance Animation
> * Set Particle Instance Animation Time
>
> The following node is completely safe and will lead to very stable
> instance rendering:
>
> * Instance Shape