Hi,<div><br></div><div>take a look at GMX_DLB_FLOP and GMX_DD_LOAD environment variables defined in domdec.c. They might help with what you are trying to do.</div><div><br></div><div>Roland<br><br><div class="gmail_quote">
---------- Forwarded message ----------<br>From: <b class="gmail_sendername">Bogdan Costescu</b> <span dir="ltr"><<a href="mailto:bcostescu@gmail.com">bcostescu@gmail.com</a>></span><br>Date: Thu, Jul 21, 2011 at 7:15 AM<br>
Subject: [gmx-developers] Reproducible runs with DLB<br>To: "<a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a>" <<a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a>><br>
<br><br>Dear GROMACS developers,<br>
<br>
I need to be able to restart from an earlier point in a simulation and<br>
exactly reproduce the original simulation while running in parallel<br>
with DD. Although I save the state of the simulation in a checkpoint<br>
file (using mdrun -cpnum), upon restart with the same number of ranks,<br>
there are differences, small at the beginning but which become larger<br>
later, which seem to appear due to the different DD cell sizes as they<br>
are modified by the dynamic load balancing (DLB). Turning DLB off<br>
(mdrun -dlb no) or running in reproducible mode (mdrun -reprod) makes<br>
the restart exactly reproduce the original (at least based on the<br>
criteria I'm interested in), however the run is significantly slower -<br>
the molecular system is not homogeneous, so DLB helps a lot in<br>
redistributing the calculations.<br>
<br>
If my understanding of the issue is correct, saving the state of the<br>
DD together with the checkpoint data and loading it upon restart would<br>
allow me to keep DLB enabled and exactly reproduce the original run.<br>
Is this so ? What are the difficulties in doing it ? If this is<br>
doable, is someone with a good understanding of DD willing to guide me<br>
in implementing it ? Of course, if someone with a good understanding<br>
of DD would be willing to implement it, I'd be more than glad to test<br>
it :-)<br>
<br>
Thanks in advance!<br>
Bogdan<br>
<font color="#888888">--<br>
gmx-developers mailing list<br>
<a href="mailto:gmx-developers@gromacs.org">gmx-developers@gromacs.org</a><br>
<a href="http://lists.gromacs.org/mailman/listinfo/gmx-developers" target="_blank">http://lists.gromacs.org/mailman/listinfo/gmx-developers</a><br>
Please don't post (un)subscribe requests to the list. Use the<br>
www interface or send it to <a href="mailto:gmx-developers-request@gromacs.org">gmx-developers-request@gromacs.org</a>.<br>
<br>
<br>
</font></div><br><br clear="all"><br>-- <br>ORNL/UT Center for Molecular Biophysics <a href="http://cmb.ornl.gov">cmb.ornl.gov</a><br>865-241-1537, ORNL PO BOX 2008 MS6309<br>
</div>