Hi all,
I had a chance to take in scipy2013.
Here are the 
tutorials<http://conference.scipy.org/scipy2013/tutorials_schedule.php?utm_source=Python+Weekly+Newsletter&utm_campaign=4253cb6a56-Python_Weekly_Issue_94_July_4_2013&utm_medium=email&utm_term=0_9e26887fc5-4253cb6a56-312680293>and
the radio
documentary clips <http://sagewrydesign.com/GoneAppy/> that I made
(includes a chat with Travis Oliphant - creator of numpy).

One of the most interesting talks was on dmtcp.
http://dmtcp.sourceforge.net/demo.html
I was sitting behind Fernando Perez - author of ipython notebook.
He was shaking his head in disbelief at its awesome potential.

The presenter showed how dmtcp could be used to checkpoint all running
programs.  Re-start, continuing on before the crash.  He claims less than a
1% performance hit.
I thought i'd mention it here as it might be useful to Blender for
crash-recovery or checkpointing long-running renders.  Also, in discussions
with the presenters, they seem to have a team of phd's looking for
challenging problems to test their metal on.  They expressed interest in
having one of their members investigate the utility/potential of applying
dmtcp to blender.  If there is interest, you can contact [email protected]
(note, dmtcp only supports linux at the moment but they plan to extend that
and have a partial port to android).  I presume one of the potential
applications would be to checkpoint long-running renders.  If crash then
recover with minimal loss.

kesten



-- 

Kesten Broughton
President and Technology Director,
Solar Mobile Trailers
[email protected]
www.sunfarmkitchens.ca <http://www.sunfarmkitchens.ca>
512 701 4209
_______________________________________________
Bf-committers mailing list
[email protected]
http://lists.blender.org/mailman/listinfo/bf-committers

Reply via email to