Mathias wrote:
Dear NG,
can somebody tell me how the work packages are scheduled to the workers?
From the code it seems to me like a a static distribution, ie each
worker gets the same amount of work, not taking into account if a faster
worker already finished all work packages.
Thanks,
Mathias
PS: Is there a better NG for scipy related questions, espescially
scipy.cow?
I don't know the general newsgroup name, but:
gmane.comp.python.scientific.user
or gmane.comp.python.scientific.devel
on news.gmane.org are the places I watch for scipy stuff.
As to the worker problem, a standard line of attack is a queue of
work-to-do tasklets, with each worker going back for more when he
finishes. The slowest running worker should then only be finished
after the fastest worker by at most the longest-running tasklet's time.
--Scott David Daniels
[EMAIL PROTECTED]
--
http://mail.python.org/mailman/listinfo/python-list
