Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!uwm.edu!cs.utexas.edu!wuarchive!udel!princeton!idacrd!desj
From: desj@idacrd.UUCP (David desJardins)
Newsgroups: comp.unix.cray
Subject: Re: Cray Autotasking
Message-ID: <509@idacrd.UUCP>
Date: 29 Nov 89 02:28:39 GMT
References: <MCCALPIN.89Nov27080806@masig3.ocean.fsu.edu>
Distribution: usa
Organization: idacrd, princeton, nj
Lines: 13

From article <MCCALPIN.89Nov27080806@masig3.ocean.fsu.edu>, by mccalpin@masig3.ocean.fsu.edu (John D. McCalpin):
> Does anyone know offhand how the Cray autotasking splits loops
> between processors?  I plan to sort the segments by length to help
> with load balancing, and need to know which segments go to which cpu
> so that I can decide exactly how to sort them.

   The tasks are dynamically allocated, in the order that they occur
in the loop.  For load balancing, my strong intuition is that you will
do pretty well if you simply sort the tasks in descending order of
work.  (This is particularly true if you aren't sure how many
processors you will have or what else will be running.)

   -- David desJardins