From david.link at aei.mpg.de Thu Nov 6 03:41:32 2008 From: david.link at aei.mpg.de (David Link) Date: Thu, 6 Nov 2008 10:41:32 +0100 Subject: [Users] Reminder: AEI Subversion servers down tomorrow Message-ID: <47E6529F-48AD-4604-B9B9-8EA751CF793A@aei.mpg.de> Dear users, this is a reminder that the Subversion servers at AEI (svn.aei.mpg.de and svn.whiskycode.org) will be down tomorrow due to the migration to a new virtual environment. The servers are expected to be up again the following day. Cheers, David Link -- ____________________________________________________________ David Link Max-Planck-Institut fuer Gravitationsphysik Albert-Einstein-Institut Am Muehlenberg 1, 14476 Golm, Germany Office: +49(0)331-567-7185 AIM: star1scraper PGP: http://numrel.aei.mpg.de/People/pgp/david_link.pgpkey ____________________________________________________________ From bentivegna at cct.lsu.edu Tue Nov 11 09:52:40 2008 From: bentivegna at cct.lsu.edu (Eloisa Bentivegna) Date: Tue, 11 Nov 2008 09:52:40 -0600 Subject: [Users] Debugging tools. Message-ID: <1226418760.3588.11.camel@numrel10.cct.lsu.edu> Dear all, as a part of ALPACA, over the next weeks I'll be making a census of the current debugging tools and best practices that people use to develop codes with Cactus. The purpose is to add a debugging How-To guide to the documentation on cactuscode.org. If you have any knowledge you'd like to share, please feel free to. I leave it up to people's preference whether this exchange should happen on or off the list. Thanks! Eloisa From baiotti at ea.c.u-tokyo.ac.jp Mon Nov 17 01:01:26 2008 From: baiotti at ea.c.u-tokyo.ac.jp (Baiotti Luca) Date: Mon, 17 Nov 2008 16:01:26 +0900 Subject: [Users] Cray XT4 Message-ID: <492116C6.5080503@ea.c.u-tokyo.ac.jp> Hello, has someone successfully run a production simulation involving cactus/carpet on a Cray XT4? I am having problems, because the compute nodes don't have local hard discs and so no swap memory and at some point there always seem to be need for it (and so of course the job crashes). Luca From jshalf at lbl.gov Mon Nov 17 10:52:21 2008 From: jshalf at lbl.gov (John Shalf) Date: Mon, 17 Nov 2008 10:52:21 -0600 Subject: [Users] Cray XT4 In-Reply-To: <492116C6.5080503@ea.c.u-tokyo.ac.jp> References: <492116C6.5080503@ea.c.u-tokyo.ac.jp> Message-ID: Most batch systems will kill a job if it starts to swap as a policy. So if you run out of memory, this is likely going to be a problem on many different systems. Which XT4 are you using? Some of them are configured with very little memory per processor relative to typical clusters. -john On Nov 17, 2008, at 1:01 AM, Baiotti Luca wrote: > Hello, > has someone successfully run a production simulation involving > cactus/carpet on a Cray XT4? > > I am having problems, because the compute nodes don't have local hard > discs and so no swap memory and at some point there always seem to be > need for it (and so of course the job crashes). > > Luca > _______________________________________________ > Users mailing list > Users at cactuscode.org > http://www.cactuscode.org/mailman/listinfo/users From schnetter at cct.lsu.edu Mon Nov 17 15:05:19 2008 From: schnetter at cct.lsu.edu (Erik Schnetter) Date: Mon, 17 Nov 2008 15:05:19 -0600 Subject: [Users] Cray XT4 In-Reply-To: <492116C6.5080503@ea.c.u-tokyo.ac.jp> References: <492116C6.5080503@ea.c.u-tokyo.ac.jp> Message-ID: <1124CF5F-515B-44E2-B5BB-7F0266DEFEBA@cct.lsu.edu> On Nov 17, 2008, at 01:01:26, Baiotti Luca wrote: > Hello, > > has someone successfully run a production simulation involving > cactus/carpet on a Cray XT4? > > I am having problems, because the compute nodes don't have local hard > discs and so no swap memory and at some point there always seem to be > need for it (and so of course the job crashes). I have run one of the XiRel benchmarks (vacuum, 9 levels AMR, standard problem size) on up to 1024 processors on Kraken. Using more than 1024 processors leads to problems, presumably due to memory problems. Frank Herrmann and Thomas Radke also tried running on Kraken and also reported similar problems. I want to emphasise that I don't know yet what causes these problems, and while I assume that memory has something to do with it I don't know this yet. I assume that a regular simulation setup uses a certain amount of memory for I/O buffers, as assigned by the operating system. Without swap space, it may be that only significantly less than 1 GB per core can be used by the simulation. For hydrodynamics simulations I suggest trying with the XiRel hydro benchmark first. You can then reduce the problem size until to know how much memory can be used on every core. Do you know for sure that you are having problems with memory consumption? If so, why? Do you see any particular error messages? Have you tried disabling certain thorns to reduce memory consumption? Have you monitored memory usage of your simulation? -erik -- Erik Schnetter http://www.cct.lsu.edu/~eschnett/ My email is as private as my paper mail. I therefore support encrypting and signing email messages. Get my PGP key from www.keyserver.net. -- Erik Schnetter http://www.cct.lsu.edu/~eschnett/ My email is as private as my paper mail. I therefore support encrypting and signing email messages. Get my PGP key from www.keyserver.net. -------------- next part -------------- A non-text attachment was scrubbed... Name: PGP.sig Type: application/pgp-signature Size: 194 bytes Desc: This is a digitally signed message part Url : http://www.cactuscode.org/pipermail/users/attachments/20081117/318dc1f9/attachment.bin