From owner-nwchem-users@emsl.pnl.gov Wed Jul 5 07:45:28 2006 Received: from odyssey.emsl.pnl.gov (localhost [127.0.0.1]) by odyssey.emsl.pnl.gov (8.13.6/8.13.6) with ESMTP id k65EjSlO024008 for ; Wed, 5 Jul 2006 07:45:28 -0700 (PDT) Received: (from majordom@localhost) by odyssey.emsl.pnl.gov (8.13.6/8.13.6/Submit) id k65EjSqr024007 for nwchem-users-outgoing-0915; Wed, 5 Jul 2006 07:45:28 -0700 (PDT) Date: Wed, 05 Jul 2006 15:44:45 +0100 From: Herbert Fruchtl Subject: Re: [NWCHEM] task tce energy - memory allocation error In-reply-to: <20060705095734.AKO79215@punts5.cc.uga.edu> X-Originating-IP: 129.215.221.142 To: Shan-ho Tsai Cc: nwchem Message-id: <1152110685.44abd05de219c@webmail.st-andrews.ac.uk> MIME-version: 1.0 Content-type: text/plain; charset=ISO-8859-1 Content-transfer-encoding: 8bit User-Agent: Internet Messaging Program (IMP) 3.2.8 X-StAndrews-MailScanner-Information: Please contact the ISP for more information X-StAndrews-MailScanner: No virus detected X-StAndrews-MailScanner-SpamCheck: not spam, SpamAssassin (score=-4.9, required 5, autolearn=not spam, BAYES_00 -4.90) X-StAndrews-MailScanner-From: herbert.fruchtl@st-andrews.ac.uk References: <20060705095734.AKO79215@punts5.cc.uga.edu> Sender: owner-nwchem-users@emsl.pnl.gov Precedence: bulk A few guesses: You are specifying 5GB in the input. Try giving it more (since you have 16GB of physical memory). TCE is a memory hog! Check if the amount of memory you can allocate is limited. Try the command "limit". If it shows a too small number, try "limit datasize unlimited" or whatever the equivalent is in AIX. There might still be a system-wide limit. The sequential version should not allocate GAs in shared memory, but you say it's compiled with MPI. Could it be the SHMMAX problem? It is described at http://www.emsl.pnl.gov/docs/global/support.html, but I don't know if this also applies to AIX. Hope this helps, Herbert Quoting Shan-ho Tsai : > > Hello all, > > We have NWCHEM 4.7 installed on an IBM power4 cluster, > running AIX5.2. It is a 64bit compilation (NWCHEM_TARGET > was IBM64), built with MPI. A serial run with the input > file > > ------------------begin input file---------------------- > echo > memory global 5000 mb > charge 0 > geometry > C -0.91962526 0.21517163 0.00014901 > C 0.48177378 0.21512985 0.00179164 > C 1.18251107 1.42875620 0.00124980 > C 0.48184939 2.64242427 -0.00098280 > C -0.91954965 2.64246604 -0.00262833 > C -1.62028699 1.42883974 -0.00204574 > H -1.45465374 -0.71145902 0.00056633 > H 1.01674454 -0.71153267 0.00348240 > H 2.25251034 1.42872430 0.00250279 > H -1.45452042 3.56912857 -0.00431178 > H -2.69028625 1.42887163 -0.00330245 > C 1.25189043 3.97607954 -0.00162549 > C 1.85254760 5.01631613 -0.00212608 > end > > basis spherical > * library 6-31g* > end > > scf > doublet > uhf > end > > tce > scf > ccsd(t) > print low > end > > task tce energy > ----------------------end input file-------------------- > > aborts with the following error message: > > > ********************************************************** > available GA memory 588164016 bytes > createfile: failed ga_create size=488150548 > ------------------------------------------------------------------------ > current input line : > 35: task tce energy > ------------------------------------------------------------------------ > ------------------------------------------------------------------------ > You need to assign at least 488150548 bytes more global memory > ------------------------------------------------------------------------ > For more information see the NWChem manual at > http://www.emsl.pnl.gov/docs/nwchem/nwchem.html > > > For further details see manual section: NWChem -> Top Level > Directives -> Memory > > > > 0:0:createfile: failed ga_create size=:: 488150548 > > ******************************************************** > > The compute node has 16GB of RAM plus 8GB of swap space > (almost all unused), so it is unlikely that the node > actually ran out of memory. Also the ulimit was set to > unlimited (except for stack memory, whose limit was 4GB). > > I get this same error message even if I increase the > amount of memory requested in the input parameter file > (tried up to "memory global 8000 mb". Also tried specifying > more stack and heap memory. Tried without specifying any > memory requests as well). In all cases the error says > that I need to assign at least 488150548 bytes more > global memory. > > Any suggestions of what might be causing the errors > and how I can fix the problem will be very much > appreciated. > > Thank you very much! > > Shan-Ho > > -------- > Shan-Ho Tsai > University of Georgia, Athens, GA > > -- Herbert Fruchtl EaStCHEM Fellow School of Chemistry University of St Andrews ------------------------------------------------------------------ University of St Andrews Webmail: https://webmail.st-andrews.ac.uk