<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#ffffff">
On 13/04/2011 10:01 PM, Mark Abraham wrote:
<blockquote cite="mid:4DA5908E.60505@anu.edu.au" type="cite">
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
On 13/04/2011 7:02 PM, delara aghaie wrote:
<blockquote
cite="mid:938030.73608.qm@web130104.mail.mud.yahoo.com"
type="cite">
<table border="0" cellpadding="0" cellspacing="0">
<tbody>
<tr>
<td style="font: inherit;" valign="top">
<div>Dear gromacs users</div>
<div>I have Dppc monolayer on Tip4p2005 water layer.</div>
<div>I use gromacs 4.0.7, tcoupl=v-rescale</div>
<div>command line in .ll file:</div>
<div>mpiexec mdrun -v -s topol.tpr</div>
<div>(I have made topol.tpr with the following command:</div>
<div><strong>grompp -c .gro -f .mdp -n .ndx -p .top -o
.tpr</strong></div>
<div><strong></strong> </div>
<div><strong>now submitting the .ll file with qsub, the
run has finished but I see this mesaage in the .e.ll
file;</strong></div>
<div><strong></strong> </div>
<div><font color="#00007f">NOTE: Turning on dynamic load
balancing</font></div>
<div><font color="#00007f">vol 0.69! imb F 149%<br>
Writing final coordinates.<br>
step 100, remaining runtime: 0 s</font></div>
<div><font color="#00007f"> Average load imbalance:
184.7 %<br>
Part of the total run time spent waiting due to
load imbalance: 45.5 %<br>
Steps where the load balancing was limited by -rdd,
-rcon and/or -dds: Z 5 %</font></div>
<div><font color="#00007f">NOTE: <font color="#6000bf">45.5
% performance was lost due to load imbalance<br>
in the domain decomposition</font>.</font></div>
<div> </div>
<div><font color="#00007f">Is there a way to fix this
load imbalance to get the better performance?</font></div>
</td>
</tr>
</tbody>
</table>
</blockquote>
<br>
Start using dynamic load balancing from the beginning of the run
with -dlb yes.<br>
<br>
<blockquote
cite="mid:938030.73608.qm@web130104.mail.mud.yahoo.com"
type="cite">
<table border="0" cellpadding="0" cellspacing="0">
<tbody>
<tr>
<td style="font: inherit;" valign="top">
<div>Does it mean that because of load imbalance, I have
lost almost 45 percentage of performance?</div>
</td>
</tr>
</tbody>
</table>
</blockquote>
<br>
Yes, but you simulated only 100 steps, so the time taken will be
dominated by I/O and setup costs.<br>
<br>
Some systems are intrinsically hard to treat in parallel. There's
a lot of diagnostic data before step 0, some of which might help
you trouble-shoot, once you demonstrate that load imbalance over a
longer run doesn't become less significant.<br>
</blockquote>
<br>
And non-load-balanced particle decomposition will do about as well
as dynamically load-balanced domain decomposition on moderate
numbers of processors for fairly short simulations where particles
cannot diffuse far. You definitely have no reason at this stage to
suppose that DD is worse than PD for your system.<br>
<br>
It may be that for a membrane system, customizing the domain
decomposition to suit the inhomogeneity of the particle and
interaction density will be necessary to get best performance early
in the run. This is why GROMACS 3.x had grompp -sort -shuffle.
GROMACS 4.x will eventually achieve a better load-balancing result
if the simulation runs for long enough.<br>
<br>
Mark<br>
</body>
</html>