From owner-nwchem-users@emsl.pnl.gov Wed Sep 12 09:00:17 2007 Received: from odyssey.emsl.pnl.gov (localhost [127.0.0.1]) by odyssey.emsl.pnl.gov (8.13.8/8.13.8) with ESMTP id l8CG0GOl011577 for ; Wed, 12 Sep 2007 09:00:17 -0700 (PDT) Received: (from majordom@localhost) by odyssey.emsl.pnl.gov (8.13.8/8.13.8/Submit) id l8CG0GGH011576 for nwchem-users-outgoing-0915; Wed, 12 Sep 2007 09:00:16 -0700 (PDT) X-Authentication-Warning: odyssey.emsl.pnl.gov: majordom set sender to owner-nwchem-users@emsl.pnl.gov using -f X-IronPort-AV: E=Sophos;i="4.20,245,1186383600"; d="scan'208";a="47784135" Message-ID: <46E80D0D.9030802@pnl.gov> Date: Wed, 12 Sep 2007 09:00:13 -0700 From: Dunyou Wang User-Agent: Thunderbird 2.0.0.6 (X11/20070728) MIME-Version: 1.0 To: Giacomo Mulas CC: NWChem users mailing list Subject: Re: [NWCHEM] recommended compilers for NWChem? References: In-Reply-To: Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit X-OriginalArrivalTime: 12 Sep 2007 16:00:12.0937 (UTC) FILETIME=[00990790:01C7F556] Sender: owner-nwchem-users@emsl.pnl.gov Precedence: bulk Hi Giacomo, Pathscale is recommended for AMD chips. NERSC (http://www.nersc.gov/nusers/systems/jacquard/) has a cluster exactly like you mentioned here. NWChem runs well on that machine. Hope this helps Dunyou Giacomo Mulas wrote: > Hello fellow NWChem users. > > I am a somewhat long time user of NWChem, but I would like to have some > first-hand reports about the performance gains/losses of using different > compilers to compine NWChem for use on a cluster composed of dual proc > dual > core Opterons, Infiniband interconnect. We have some (not a lot) money to > spend for compilers, and since NWChem is going to be one of the most used > packages I would like to hear some recommendations on what to buy and > what > to avoid. We are considering: > - Pathscale > - Portland Group > - Intel > - NAGware > Having some experience in getting software to compile on different > UNIX-like > platforms with different compilers I expect to be able, with some > effort, to use any of the above suits to compile NWChem, but I would > like to purchase > the one(s) which ought to get me the fastest NWChem executables. On a >60 > nodes cluster, a 10% difference amounts to 6 more (or less) nodes, which > translated in money is quite nontrivial... > > Thanks for any help > Giacomo >