<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 TRANSITIONAL//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV="Content-Type" CONTENT="text/html; CHARSET=UTF-8">
<META NAME="GENERATOR" CONTENT="GtkHTML/3.16.1">
</HEAD>
<BODY>
Hi Carsten and Berk,<BR>
<BR>
Your fixes did the trick. Man, there is some great scaling on the CVS version now! Thanks a lot for the help!<BR>
<BR>
/ Erik<BR>
<BR>
ons 2008-01-23 klockan 09:20 +0100 skrev Berk Hess:
<BLOCKQUOTE TYPE=CITE>
<PRE>
<FONT COLOR="#000000">----------------------------------------</FONT>
<FONT COLOR="#000000">> Date: Tue, 22 Jan 2008 20:15:03 +0100</FONT>
<FONT COLOR="#000000">> From: <A HREF="mailto:ckutzne@gwdg.de">ckutzne@gwdg.de</A></FONT>
<FONT COLOR="#000000">> To: <A HREF="mailto:gmx-users@gromacs.org">gmx-users@gromacs.org</A></FONT>
<FONT COLOR="#000000">> Subject: [Fwd: Re: [gmx-users] mdrun CVS version crashes instantly when run across nodes in parallel]</FONT>
<FONT COLOR="#000000">> </FONT>
<FONT COLOR="#000000">> Hi Erik,</FONT>
<FONT COLOR="#000000">> </FONT>
<FONT COLOR="#000000">> I have made a test with today's CVS version and I also run into the</FONT>
<FONT COLOR="#000000">> problem you described. It happens as soon as one uses more than one node</FONT>
<FONT COLOR="#000000">> and at the same time more than one process per node.</FONT>
<FONT COLOR="#000000">> </FONT>
<FONT COLOR="#000000">> The problem seems to be in gmx_sumd where in the two-step summation in a</FONT>
<FONT COLOR="#000000">> call to MPI_Allreduce the variable cr->nc.comm_inter happens to be a</FONT>
<FONT COLOR="#000000">> NULL pointer, which should clearly not be.</FONT>
<FONT COLOR="#000000">> </FONT>
<FONT COLOR="#000000">> The inter-node communicator is freed in gmx_setup_nodecomm (network.c,</FONT>
<FONT COLOR="#000000">> line 393) if an intra-node communicator is present - I do not understand</FONT>
<FONT COLOR="#000000">> why the communicator is freed here.</FONT>
<FONT COLOR="#000000">> </FONT>
<FONT COLOR="#000000">> Maybe Berk can help us on that? If I comment out the MPI_Comm_free the</FONT>
<FONT COLOR="#000000">> code runs happily - haven't checked the results, though.</FONT>
<FONT COLOR="#000000">> </FONT>
<FONT COLOR="#000000">> Carsten</FONT>
<FONT COLOR="#000000">Ah, there is a "typo" in the conditional, it should be rank_intra iso comm_intra.</FONT>
<FONT COLOR="#000000">I have committed the fix.</FONT>
<FONT COLOR="#000000">Thanks,</FONT>
<FONT COLOR="#000000">Berk.</FONT>
<FONT COLOR="#000000">_________________________________________________________________</FONT>
<FONT COLOR="#000000">Express yourself instantly with MSN Messenger! Download today it's FREE!</FONT>
<FONT COLOR="#000000"><A HREF="http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/">http://messenger.msn.click-url.com/go/onm00200471ave/direct/01/</A></FONT>
</PRE>
</BLOCKQUOTE>
<TABLE CELLSPACING="0" CELLPADDING="0" WIDTH="100%">
<TR>
<TD>
-- <BR>
Erik Brandt <<A HREF="mailto:erikb@theophys.kth.se">erikb@theophys.kth.se</A>><BR>
KTH
</TD>
</TR>
</TABLE>
</BODY>
</HTML>