[phenixbb] reduce hangs
docandreas at gmail.com
Thu Dec 5 01:57:56 PST 2013
I'm not sure how much of my conversation with Jeff Headd and the
comments from others that I received made it onto the mailing list. The
problem with reduce being stuck in an infinite loop and refinement never
completing has now a workaround, if no solution:
Mark van Raaij pointed out that running refinement repeatedly while
varying the number of cores would eventually avoid the infinite loop.
I didn't observe that, but while running refinement with a handful of
Xeons got stuck, running the same job with the same executable on the
same OS with a bucketful of Opterons instead succeeded.
As I said, there's no solution but a number of ways to get unstuck if
you find yourself in the same situation like Mark and me.
-------- Original Message --------
Subject: Re: [phenixbb] reduce hangs
Date: Wed, 4 Dec 2013 09:46:20 -0500
From: Jeff Headd <jjheadd at lbl.gov>
To: Andreas Förster <docandreas at gmail.com>
If the different machine is a slightly different architecture it's
possible that it may run, even under the same OS version. My guess is
that it's down to slight floating point differences in the coordinates
throughout the course of refinement, but it's so hard to pinpoint. If
you have access to other machines it is probably worth trying.
More information about the phenixbb