From owner-nwchem-users@emsl.pnl.gov Mon May 19 14:03:12 2008 Received: from odyssey.emsl.pnl.gov (localhost.pnl.gov [127.0.0.1]) by odyssey.emsl.pnl.gov (8.14.1/8.14.1) with ESMTP id m4JL3BVb002006 for ; Mon, 19 May 2008 14:03:12 -0700 (PDT) Received: (from majordom@localhost) by odyssey.emsl.pnl.gov (8.14.1/8.14.1/Submit) id m4JL3Boj002005 for nwchem-users-outgoing-0915; Mon, 19 May 2008 14:03:11 -0700 (PDT) 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: 2.4 X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: AtMAAOOHMUhCxGRafGdsb2JhbACSMQEBCwUCBgcRA5wfAQ X-IronPort-AV: E=Sophos;i="4.27,511,1204531200"; d="scan'208";a="52928278" DomainKey-Signature: a=rsa-sha1; q=dns; c=nofws; s=s1024; d=yahoo.com; h=X-YMail-OSG:Received:X-Mailer:Date:From:Reply-To:Subject:To:In-Reply-To:MIME-Version:Content-Type:Content-Transfer-Encoding:Message-ID; b=u8cIOzAyiZ4X6LTyv8TJj3nq4aDjN/uYEmhDemOlLv8/Qrxkxj6aCp6+3xA08jsBbdVjKbiFe6DlmPXZ5iDIPNi2VvzIHDwC4J6ofdE8Hgy6mTeckoBJa2uIRfccGQIZTPgdlpa3EHMCxsHSRasxI0Oz+Dk2o0rNX4QVx23ncuE=; X-YMail-OSG: zaRYTO4VM1nLMa88akq.KoutL2i0lDV3hv41HsAqyFG6wRhiA3OxYMk9Fx8XRLzyl8N6BAR_rkbsX3zc0prai8J3ICgvg.JrimCxTAB6Ds3ylnauLhqaXaSJiOGv93wzHR8TAg-- X-Mailer: YahooMailWebService/0.7.185 Date: Mon, 19 May 2008 14:03:08 -0700 (PDT) From: Francesco Pietra Reply-To: chiendarret@yahoo.com Subject: Re: [NWCHEM] MKL libraries To: users nwchem , =?utf-8?B?SsO2cmcgU2HDn21hbm5zaGF1c2Vu?= In-Reply-To: <200805192040.30230.sassmannshausen@tugraz.at> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Message-ID: <484674.76648.qm@web57608.mail.re1.yahoo.com> Sender: owner-nwchem-users@emsl.pnl.gov Precedence: bulk Hi Joerg: --- On Mon, 5/19/08, J=C3=B6rg Sa=C3=9Fmannshausen wrote: > From: J=C3=B6rg Sa=C3=9Fmannshausen > Subject: Re: [NWCHEM] MKL libraries > To: "users nwchem" > Date: Monday, May 19, 2008, 11:40 AM > Hi Francesco, > as Jeff pointed out, it is always good to use an highly > optimised BLAS library=20 > for your program. I noticed a few days ago that the > difference between=20 > mkl-10.0.1.014 and ATLAS3.7.11 is only marginal (for a > different program=20 > though). So that might be an alternative to mkl.=20 For me (non commercial use) MKL is free. Perhaps another alternative with = the AMD math libraries, which are also free and may be especially tailored = for the opterons. >Also, some > people are saying=20 > that GotoBLAS works well with their system. You can test > out that fairly=20 > quickly by building NWChem and then link it manually to the > libraries you=20 > want to test. As a chemist that has turned from experiments to the virtual only recently,= my experience is very limited but I can try. My first effort with NWChem/D= ebian has produced F. Pietra =E2=80=9CWhy colchicine does not show mutarot= ation. With M05=E2=80=912X density functional in the realm of tricky natura= l products. J. Phys. Org. Chem. 2007, 20, 1102=E2=80=911107. Now I am heavi= ly involved with docking and classical molecular dynamics in explicit media= , though I have left many problems of organic chemistry to submit to NWChem= and I am self stimulated to come back to quantum mechanics. >=20 > As I am using Debian as well (currently etch) I am a bit > curious about your=20 > experiences with lenny. Also, what I would like to know is: > you need libnuma=20 > for the OpenMPI, or? Currently I am using MPICH2 but I am > hearing people=20 > saying that OpenMPI is better, what are your experiences? Can't compare. My experience is limited to openmpi. What I can say is that = that extremely involved compilation flows smoothly thanks to a perfect conf= iguration file provided by the developers. I must add that also Debian - = as a genuine linux - helps a lot. Finally, the big seems also come at the o= penmpi table. Including IMB! And Infiniband is providing some or their syst= ems on openmpi. >=20 > I think it would be benificial if you could share you > experiences with=20 > different BLAS versions here so other people could gain > from that as well.=20 Now I am helping the developers (an myself) of the MD application to fix a = bug and those of the docking application to speed up the procedures. Once f= inished, I'll try to compile NWChem with MKL. As I am always dealing with b= ig molecules, I expect a benefit. > Let me know how you getting on with the compilation of > NWChem, as I have done=20 > it recently I might be able to help if you are stuck. Thanks a lot. >=20 > Best wishes from a rainy Graz! Same in Lucca, where I live. Raining cats and dogs. Graz, beautiful. My wif= e, Gudrun, is from Villach. francesco >=20 > J=C3=B6rg >=20 > Am Montag 19 Mai 2008 18:44 schrieb Francesco Pietra: > > nwchem 5.1 was installed from binary > > nwchem-5.1-LINUX64-RHELWS4-ifort-AMD64.tar.bz2 on a > NUMA-type machine with > > two dual-opteron, runtime provided by intel version 9. > OS debian Linux > > amd64 etch. > > > > The two raid1 HHDDs have been moved to a new 4-socket > dual-opteron > > NUMA-type machine; nwchem worked fine. > > > > Now, on the latter machine, I upgraded the OS to > "lenny", intel compilers > > to vers 10.0.1.014 with MKL libraries 10.0.1.014 > added, openmpi to vers > > 10.1 with libnuma vers 1.0.2-1. > > > > As I have noticed a tremendous speed up by the mkl > libraries of the QMMM > > part of a molecular dynamics package, I wonder whether > using mkl libraries > > for nwchem in my type of system might be beneficial. > If so, should I build > > nwchem from source with intel compilers on openmpi? Is > it a list of flags > > to that scope, or the binary you provide is anyway > expected to be faster? > > > > Thanks > > francesco pietra > --=20 > ************************************************************* > J=C3=B6rg Sa=C3=9Fmannshausen > Institut f=C3=BCr Chemische Technologie von Materialien > TU-Graz > Stremayrgasse 16 > 8010 Graz > Austria >=20 > phone: +43 (0)316 873 8954 > fax: +43 (0)316 873 4959 > homepage: http://sassy.formativ.net/ >=20 > Please avoid sending me Word or PowerPoint attachments. > See http://www.gnu.org/philosophy/no-word-attachments.html=0A=0A=0A