From owner-nwchem-users@emsl.pnl.gov Wed Nov 28 09:18:43 2007 Received: from odyssey.emsl.pnl.gov (localhost [127.0.0.1]) by odyssey.emsl.pnl.gov (8.14.1/8.14.1) with ESMTP id lASHIglW013221 for ; Wed, 28 Nov 2007 09:18:43 -0800 (PST) Received: (from majordom@localhost) by odyssey.emsl.pnl.gov (8.14.1/8.14.1/Submit) id lASHIgad013220 for nwchem-users-outgoing-0915; Wed, 28 Nov 2007 09:18:42 -0800 (PST) X-Authentication-Warning: odyssey.emsl.pnl.gov: majordom set sender to owner-nwchem-users@emsl.pnl.gov using -f X-Ironport-SG: OK_Domains X-Ironport-SBRS: 5.9 X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: AgAAANMxTUegWwRun2dsb2JhbACPPgEBAQEHBAYJCBhQ X-IronPort-AV: E=Sophos;i="4.23,225,1194249600"; d="scan'208";a="58038032" Date: Wed, 28 Nov 2007 12:18:37 -0500 From: "de Almeida, Valmor F." Subject: RE: [NWCHEM] gfortran (gcc-4.1.2) system error message: Bad filedescriptor In-reply-to: <474D9FCB.309@pnl.gov> To: Dunyou Wang Cc: nwchem-users@emsl.pnl.gov Message-id: MIME-version: 1.0 X-MIMEOLE: Produced By Microsoft Exchange V6.5 Content-type: text/plain; charset=iso-8859-1 Content-transfer-encoding: quoted-printable Thread-Topic: [NWCHEM] gfortran (gcc-4.1.2) system error message: Bad filedescriptor Thread-Index: Acgx4OLrX53B6LtjRgiet2pFw/esYAAAUTiA Content-class: urn:content-classes:message X-MS-Has-Attach: X-MS-TNEF-Correlator: References: <20071128170929.13313ev0gr463o4k@webmail.tugraz.at> <474D9FCB.309@pnl.gov> Sender: owner-nwchem-users@emsl.pnl.gov Precedence: bulk Thanks for the suggestion. I was hoping TCGMSG would work with gfortran. = This is how I had it in the past on the cluster. It did work but I used = the ifort compiler. -- Valmor > -----Original Message----- > From: Dunyou Wang [mailto:dunyou.wang@pnl.gov] > Sent: Wednesday, November 28, 2007 12:05 PM > To: de Almeida, Valmor F. > Cc: nwchem-users@emsl.pnl.gov > Subject: Re: [NWCHEM] gfortran (gcc-4.1.2) system error message: Bad > filedescriptor >=20 > Valmor, >=20 > You could download MPICH and have you MPI built over gigabit Ethernet, > instead of MPI over Myrinet MX. In this way, you still could build = your > NWChem using MPI on your cluster. >=20 > Hope this helps > Dunyou >=20 > de Almeida, Valmor F. wrote: > > Hi J=F6rg, > > > > I compiled NWChem5.0 on a linux-2.6.22.9/gcc-4.1.2 system (4 CPUS) = with > OpenMPI-1.2.4. This works fine; just do > > > > mpirun -n 4 nwchem nwchem.input > > > > Then I recompiled from scratch without MPI, so I could test the = TCGMSG > parallel version. The compilation is fine but the executable = ("parallel") > is broken if more than one 1 processor is used. Say > > > > parallel nwchem input.data > > > > will start the parallel job, as described at the bottom of this = e-mail, > but will crash at the DFT calculation level when there is attempt for = data > exchange between processors. > > > > The reason I need the TCGMSG version is that I have a cluster with = the > Myrinet MX library, which does not support one-sided communication = yet, > therefore I can't build NWChem on the cluster since OpenMPI is built = on > top of MX. > > > > Thanks for the reply. > > > > -- > > Valmor > > > > > >> -----Original Message----- > >> From: owner-nwchem-users@emsl.pnl.gov [mailto:owner-nwchem- > >> users@emsl.pnl.gov] On Behalf Of J=F6rg Sa=DFmannshausen > >> Sent: Wednesday, November 28, 2007 11:09 AM > >> To: nwchem-users@emsl.pnl.gov > >> Subject: RE: [NWCHEM] gfortran (gcc-4.1.2) system error message: = Bad > >> filedescriptor > >> > >> Hi Valmor, > >> > >> I compiled NWChem5.0 on an old PIII with gfortran-4.1.2 and MPICH2. > >> So this is working, at least for 32bit. Also, I got it running = using > >> the Intel Compiler which makes more sense on Intel CPUs (IMHO). > >> > >> How did you start the job? As you are using MPI(CH), you don't = really > >> need the TCGMSG as as far as I understand it, this is only needed = if > >> you are not using MPI. > >> With MPICH, you have to use > >> > >> mpiexec -n /path/to/nwchem /path/to/input/file > >> > >> I am currently away, so please don't contact me directly as I won't = be > >> able to reply straight away but others on the email list might be = able > >> to help. > >> > >> Best wishes > >> > >> J=F6rg > >> > >> > >> > >>> Hello, > >>> > >>> A bit more info on this issue. If I compile with MPI enabled, the > >>> resultant code works. Therefore it seems that there is a problem = with > >>> gfortran (gcc-4.1.2) and TCGMSG. > >>> > >>> Has anyone been able to compile TCGMSG with gfortran 4.1.2 and = obtain > a > >>> functional "parallel" executable? > >>> > >>> Thanks, > >>> > >>> -- > >>> Valmor de Almeida > >>> > >>> > >>>> -----Original Message----- > >>>> From: owner-nwchem-users@emsl.pnl.gov [mailto:owner-nwchem- > >>>> users@emsl.pnl.gov] On Behalf Of de Almeida, Valmor F. > >>>> Sent: Saturday, November 24, 2007 3:46 PM > >>>> To: nwchem-users@emsl.pnl.gov > >>>> Subject: [NWCHEM] system error message: Bad file descriptor > >>>> > >>>> > >>>> Hello, > >>>> > >>>> After an OS system upgrade, I decided to build nwchem-5.0 with > >>>> > >>> gfortran > >>> > >>>> (gcc-4.1.2) on an intel linux box. Before the upgrade, I had = nwchem-5 > >>>> working on the system but it was compiled with the intel ifort > >>>> > >>> compiler > >>> > >>>> (intel-9.1). > >>>> > >>>> Here is what happens now: > >>>> > >>>> parallel nwchem input.nw > >>>> > >>>> works fine as long as the nwchem.p file calls for 1 processor. = More > >>>> > >>> than > >>> > >>>> 1 processor gives the error below. > >>>> > >>>> Any inputs appreciated. > >>>> > >>>> Thanks, > >>>> > >>>> -- > >>>> Valmor > >>>> > >>>> > >>>> [snip...] > >>>> > >>>> Summary of "ao basis" -> "" (cartesian) > >>>> > >>>> > >>>> > >>> = ---------------------------------------------------------------------- > -- > >>> > >>>> ------ > >>>> Tag Description Shells = Functions > >>>> > >>> and > >>> > >>>> Types > >>>> ---------------- ------------------------------ ------ > >>>> --------------------- > >>>> h 6-31g 2 2 = 2s > >>>> o 6-31g 5 9 > 3s2p > >>>> > >>>> NWChem DFT Module > >>>> ----------------- > >>>> > >>>> Water in 6-31g basis set DFT optimization xc = b3lyp > >>>> > >>>> sock=3D0, pid=3D1, nsent=3D-1, len=3D28 > >>>> 1: snd_remote: writing message to socket 199999999 (0xbebc1ff). > >>>> 1: snd_remote: writing message to socket 199999999 (0xbebc1ff). > >>>> system error message: Bad file descriptor > >>>> 0:Child process terminated prematurely, status=3D: 256 > >>>> 0:Child process terminated prematurely, status=3D: 256 > >>>> Last System Error Message from Task 0:: Inappropriate ioctl for > device > >>>> 0: ARMCI aborting 256 (0x100). > >>>> 0: ARMCI aborting 256 (0x100). > >>>> system error message: Inappropriate ioctl for device > >>>> 2: interrupt(1) > >>>> WaitAll: No children or error in wait? > >>>> > > > >