From jtao at cct.lsu.edu Mon Mar 3 15:12:06 2008 From: jtao at cct.lsu.edu (Jian Tao) Date: Mon, 03 Mar 2008 15:12:06 -0600 Subject: [Users] Cactus newsletter Message-ID: <47CC69A6.2030201@cct.lsu.edu> _______________________________________________________________________ ___| | \ | | _` | __| __| | | __| \ | _ \ \ \ / __| | ( | ( | | |\__ \ |\ | __/\ \ \ /\__ \ \____|\__,_|\___|\__|\__,_|____/ _| \_|\___| \_/\_/ ____/ _______________________________________________________________________ Summary: ======== 1. Cactus Showcased at SC07 2. Cactus Run on First Loongson 2F Chinese Cluster 3. Cactus Benchmarks Run on Ranger 4. First XiRel Workshop (Baton Rouge, Jan 23 - 25, 2008) 5. Significant Improvements of Carpet 6. Upcoming Event: Cactus Tutorial at LSU HPC Workshop 7. Opening Positions * The next newsletter will be posted in early June. If you have any Cactus related events or announcements that you want to add to the next newsletter please send them to cactusmaint at cactuscode.org. Contents: ========= 1. Cactus Showcased at SC07 The Cactus Computational Toolkit has been showcased again by the Center for Computation & Technology(CCT) researchers at the annual Supercomputing Conference 2007 (SC07) from Reno, Nevada. SC07 is an international conference on high performance computing, networking, storage and analysis. (http://www.cactuscode.org/News/CactusSC07) 2. Cactus Run on First Loongson 2F Chinese Cluster On December 26, 2007, China revealed its first supercomputer of 1 TFlop/s utilizing the domestic Chinese CPU Loongson 2F at Hefei, designated as KD-50-I. This supercomputer was designed by a joint team led by Dr. Chen Guoliang, professor of the Computer Science and Technology Department of the University of Science and Technology of China (the primary contractor, with the Research Institute of Computational Technology of Chinese Academy of Sciences as the secondary contractor). With the help of the system administrator Huimin Li, we compiled and benchmarked Cactus on the cluster. (http://kd50.ustc.edu.cn/) (http://www.cactuscode.org/News/CactusLoongson) 3. Cactus Benchmarks Run on Ranger Cactus tested on Ranger, the new 60,000 core supercomputer at TACC The Cactus group at LSU successfully ported the Cactus framework to Ranger, the new 60,000 core supercomputer at TACC, and ran a series of numerical relativity benchmarks there. We are happy to report that these benchmarks scale up to 4096 cores, the maximum job size possible while Ranger is still in "friendly user" mode. Ranger is one of the largest computational resources in the world, consisting of 3,936 16-way SMP compute-nodes with 123 TByte of memory, and with a theoretical peak performance of 504 TFlop/s. (http://www.carpetcode.org/) (http://www.cactuscode.org/News/Ranger) 4. First XiRel Workshop (Baton Rouge, Jan 23 - 25, 2008) The Center for Computation & Technology held the first XiRel workshop that brought together researchers from LSU, Rochester Institute of Technology, Pennsylvania State University, and the Albert-Einstein Institute from Germany. (http://www.cct.lsu.edu/XiRel/) (http://www.cactuscode.org/News/XiRelWorkshop08) 5. Significant Improvements of Carpet Significant improvements were contributed to the development version of Carpet on March 1,2008. Thanks to more efficient data structures and algorithms, Carpet is now able to scale to more than 8,000 cores. Special thanks go to Luca Baiotti, Denis Pollney, Christian Reisswig, Jian Tao, and also to the CCT numerical relativity group, the AEI numerical relativity group, and the XiRel collaboration. More information about the improvements can be found at http://www.carpetcode.org/ 6. Upcoming Event: Cactus Tutorial at LSU HPC Workshop HPC @ LSU will hold a High Performance Computing Workshop on March 11 & 12 in Frey 307 at LSU in Baton Rouge. This workshop will also be held on the Access Grid so that TeraGrid and LONI users on remote sites may attend. Erik Schnetter is scheduled to give a tutorial on Cactus for potential Cactus users starting at 3:00PM on March 12. (http://www.hpc.lsu.edu/training/20080311/index.php) 7. Opening Positions There are several Cactus related open positions for postdocs and graduate students at LSU through the XiRel, Alpaca, CyberTools and other projects. Please contact Gabrielle Allen at gallen at cct.lsu.edu for more information. _______________________________________________________________________ ___| | \ | | _` | __| __| | | __| \ | _ \ \ \ / __| | ( | ( | | |\__ \ |\ | __/\ \ \ /\__ \ \____|\__,_|\___|\__|\__,_|____/ _| \_|\___| \_/\_/ ____/ _______________________________________________________________________ From wongst at ihpc.a-star.edu.sg Wed Mar 5 04:06:16 2008 From: wongst at ihpc.a-star.edu.sg (Wong Sum Thai) Date: Wed, 05 Mar 2008 18:06:16 +0800 Subject: [Users] PETSc in Cactus Message-ID: <000701c87ea8$8cbc8630$6916a8c0@IHPC> Hi, I have just started working with Cactus with the intention of using PETSc as the underlying solver. Although I have yet to set up PETSc with Cactus, I would like to know if I configure Cactus to use MPI, should I still compile PETSc to use MPI as well? In this case, since Cactus is already distributed among multiple processors, then PETSc will follow even if it is not compile to use MPI. Is this correct? Thanks. -------------------------------------------------------- This email is confidential and may be privileged. If you are not the intended recipient, please delete it and notify us immediately. Please do not copy or use it for any purpose, or disclose its contents to any other person. Thank you. -------------------------------------------------------- From schnetter at cct.lsu.edu Wed Mar 5 09:02:28 2008 From: schnetter at cct.lsu.edu (Erik Schnetter) Date: Wed, 5 Mar 2008 09:02:28 -0600 Subject: [Users] PETSc in Cactus In-Reply-To: <000701c87ea8$8cbc8630$6916a8c0@IHPC> References: <000701c87ea8$8cbc8630$6916a8c0@IHPC> Message-ID: <58302D60-4D34-4533-B5B2-B72BBB8E8D6D@cct.lsu.edu> On Mar 5, 2008, at 04:06:16, Wong Sum Thai wrote: > Hi, > > I have just started working with Cactus with the intention of using > PETSc as > the underlying solver. Although I have yet to set up PETSc with > Cactus, I > would like to know if I configure Cactus to use MPI, should I still > compile > PETSc to use MPI as well? In this case, since Cactus is already > distributed > among multiple processors, then PETSc will follow even if it is not > compile > to use MPI. Hi, Cactus requires that PETSc is already installed independently of Cactus. It can use only those features of PETSc which are already built in -- while PETSc solves a system, it has complete control for a short time to achieve maximum efficiency. If you want to solve a system in parallel, both Cactus and PETSc need to be built with MPI. -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. -------------- 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/20080305/ac4e8295/attachment.bin From wongst at ihpc.a-star.edu.sg Wed Mar 5 19:57:37 2008 From: wongst at ihpc.a-star.edu.sg (Wong Sum Thai) Date: Thu, 06 Mar 2008 09:57:37 +0800 Subject: [Users] PETSc in Cactus In-Reply-To: <58302D60-4D34-4533-B5B2-B72BBB8E8D6D@cct.lsu.edu> References: <000701c87ea8$8cbc8630$6916a8c0@IHPC> <58302D60-4D34-4533-B5B2-B72BBB8E8D6D@cct.lsu.edu> Message-ID: <000601c87f2d$740d3e60$6916a8c0@IHPC> Hi Erik, Thanks for the clarification. I get the idea now. -------------------------------------------------------- This email is confidential and may be privileged. If you are not the intended recipient, please delete it and notify us immediately. Please do not copy or use it for any purpose, or disclose its contents to any other person. Thank you. -------------------------------------------------------- -----Original Message----- From: users-bounces at cactuscode.org [mailto:users-bounces at cactuscode.org] On Behalf Of Erik Schnetter Sent: Wednesday, March 05, 2008 11:02 PM To: Cactus users mailing list Subject: Re: [Users] PETSc in Cactus On Mar 5, 2008, at 04:06:16, Wong Sum Thai wrote: > Hi, > > I have just started working with Cactus with the intention of using > PETSc as > the underlying solver. Although I have yet to set up PETSc with > Cactus, I > would like to know if I configure Cactus to use MPI, should I still > compile > PETSc to use MPI as well? In this case, since Cactus is already > distributed > among multiple processors, then PETSc will follow even if it is not > compile > to use MPI. Hi, Cactus requires that PETSc is already installed independently of Cactus. It can use only those features of PETSc which are already built in -- while PETSc solves a system, it has complete control for a short time to achieve maximum efficiency. If you want to solve a system in parallel, both Cactus and PETSc need to be built with MPI. -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. From stokes at marshall.edu Sun Mar 9 13:01:53 2008 From: stokes at marshall.edu (Paul Stokes) Date: Sun, 09 Mar 2008 15:01:53 -0400 Subject: [Users] compiler In-Reply-To: References: <7165605.1203701865131.JavaMail.pduser@MULM3PDWWW01> <1BAC73F2-EAC7-4CE6-BAD7-1225BA6A7A3D@cct.lsu.edu> <47C73C6E.6030307@marshall.edu> Message-ID: <47D43421.5050005@marshall.edu> Erik Schnetter wrote: > On Feb 28, 2008, at 16:57:50, Paul Stokes wrote: > >> I tried the >> make waveDemo SILENT=no >> but it did the same thing. > > Yes, but it produced more output. In fact, it lists all the commands > which it is going to execute. Can you show us the last command that > Cactus was trying to run? It should begin with "c++". > >> Here is the file you asked for > >> CC = gcc >> CXX = c++ >> F90 = ifort >> F77 = ifort > > This indicates that you are mixing gcc's C and C++ compilers with > Intel's Fortran compilers. That is not bad in itself, but you have to > be aware that you are not using Intel's C and C++ compilers. > >> LDFLAGS = -Qy /opt/intel/fc/10.1.012/lib/icrt.link > > This is where Cactus picks up the problematic option. Does this file > exist? If not, where is your Intel Fortran compiler installed? You > find out by issuing > > which ifort > > The remainder looks fine. > >> -- No virus found in this incoming message. Checked by AVG Free >> Edition.Version: 7.5.516 / Virus Database: 269.21.1/1302 - Release Date: >> 2/27/2008 4:34 PM > > > I'm happy to know that. > > -erik > > ------------------------------------------------------------------------ > > _______________________________________________ > Users mailing list > Users at cactuscode.org > http://www.cactuscode.org/mailman/listinfo/users > > ------------------------------------------------------------------------ > > I can't find the icrt.link anywhere. I have ifort installed in: opt/intel/fc/10.1.012/bin From schnetter at cct.lsu.edu Sun Mar 9 18:28:21 2008 From: schnetter at cct.lsu.edu (Erik Schnetter) Date: Sun, 9 Mar 2008 19:28:21 -0500 Subject: [Users] compiler In-Reply-To: <47D43421.5050005@marshall.edu> References: <7165605.1203701865131.JavaMail.pduser@MULM3PDWWW01> <1BAC73F2-EAC7-4CE6-BAD7-1225BA6A7A3D@cct.lsu.edu> <47C73C6E.6030307@marshall.edu> <47D43421.5050005@marshall.edu> Message-ID: <66BBB2BF-15A5-4B2E-B18F-C1F9AB916B5E@cct.lsu.edu> On Mar 9, 2008, at 14:01:53, Paul Stokes wrote: > Erik Schnetter wrote: >> On Feb 28, 2008, at 16:57:50, Paul Stokes wrote: >> >>> I tried the >>> make waveDemo SILENT=no >>> but it did the same thing. >> >> Yes, but it produced more output. In fact, it lists all the commands >> which it is going to execute. Can you show us the last command that >> Cactus was trying to run? It should begin with "c++". >> >>> Here is the file you asked for >> >>> CC = gcc >>> CXX = c++ >>> F90 = ifort >>> F77 = ifort >> >> This indicates that you are mixing gcc's C and C++ compilers with >> Intel's Fortran compilers. That is not bad in itself, but you have >> to >> be aware that you are not using Intel's C and C++ compilers. >> >>> LDFLAGS = -Qy /opt/intel/fc/10.1.012/lib/icrt.link >> >> This is where Cactus picks up the problematic option. Does this file >> exist? If not, where is your Intel Fortran compiler installed? You >> find out by issuing >> >> which ifort >> >> The remainder looks fine. >> >>> -- No virus found in this incoming message. Checked by AVG Free >>> Edition.Version: 7.5.516 / Virus Database: 269.21.1/1302 - Release >>> Date: >>> 2/27/2008 4:34 PM >> >> >> I'm happy to know that. >> >> -erik >> > I can't find the icrt.link anywhere. I have ifort installed in: > > opt/intel/fc/10.1.012/bin Can you also answer my other question? That is, you run Cactus with SILENT=no and tell us what the last command was that Cactus was executing. This tells us what stage of the compilation went wrong, especially what Cactus is trying to do with icrt.link. -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. -------------- 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/20080309/2a55f651/attachment.bin From hosseinf at guilan.ac.ir Mon Mar 10 02:31:11 2008 From: hosseinf at guilan.ac.ir (hosseinf at guilan.ac.ir) Date: Mon, 10 Mar 2008 12:01:11 +0330 Subject: [Users] help to run ADM Message-ID: Dear Sir I would like to run ADM after I already done some simple examples like wavetoy. I have problem with ThornList. Would you please send us the proper stable Thorn for this application. Thanks Hossein From stokes at marshall.edu Mon Mar 10 08:27:07 2008 From: stokes at marshall.edu (Paul Edward Stokes) Date: Mon, 10 Mar 2008 10:27:07 -0400 (EDT) Subject: [Users] [Compiler] Message-ID: <2627160.1205159227796.JavaMail.pduser@MULM3PDWWW03> A non-text attachment was scrubbed... Name: make1_dev Type: application/octet-stream Size: 1977 bytes Desc: not available Url : http://www.cactuscode.org/pipermail/users/attachments/20080310/b1ddfa12/attachment-0004.obj -------------- next part -------------- A non-text attachment was scrubbed... Name: make1_stable Type: application/octet-stream Size: 1848 bytes Desc: not available Url : http://www.cactuscode.org/pipermail/users/attachments/20080310/b1ddfa12/attachment-0005.obj -------------- next part -------------- A non-text attachment was scrubbed... Name: make2_dev Type: application/octet-stream Size: 2195 bytes Desc: not available Url : http://www.cactuscode.org/pipermail/users/attachments/20080310/b1ddfa12/attachment-0006.obj -------------- next part -------------- A non-text attachment was scrubbed... Name: make2_stable Type: application/octet-stream Size: 1986 bytes Desc: not available Url : http://www.cactuscode.org/pipermail/users/attachments/20080310/b1ddfa12/attachment-0007.obj From tradke at aei.mpg.de Tue Mar 11 06:58:09 2008 From: tradke at aei.mpg.de (Thomas Radke) Date: Tue, 11 Mar 2008 13:58:09 +0100 Subject: [Users] [Cluster-users] problems recovering job from checkpoint In-Reply-To: References: Message-ID: <47D681E1.7020600@aei.mpg.de> moesta wrote: > Hi, > > I have a problem recovering a job from checkpoint. It recovers but then > is extremly slow and the constraint violations nearly vanish which i > don't really understand. Have tried various parameter settings, but they > don't seem to make any difference. Ideas what can be the reason for this? Hi Philipp, I guess we need some more information to find out what's wrong. What machine are you trying to recover (same as where you checkpointed) ? Do you think the recovery itself was successful, only then run is progressing much slower afterwards ? Which version of Carpet are you using ? Did you change/update your source code in the meantime ? -- Cheers, Thomas. From schnetter at cct.lsu.edu Thu Mar 13 13:57:49 2008 From: schnetter at cct.lsu.edu (Erik Schnetter) Date: Thu, 13 Mar 2008 14:57:49 -0500 Subject: [Users] Warnings when compiling on OS X.5.2 In-Reply-To: <342C3DD7-820B-4A6A-A5F5-A02BB43CFD89@rit.edu> References: <342C3DD7-820B-4A6A-A5F5-A02BB43CFD89@rit.edu> Message-ID: <9EDD62A6-2142-4C19-A92A-0AAF17033FD8@cct.lsu.edu> On Mar 13, 2008, at 12:21:16, Joshua Faber wrote: > Hi Erik, > I noticed you committed the most recent update to the Carpet > architectures listing. I have the following warnings appearing when > I compile our code, and figured you might be able to help. When I > compile on a machine running Leopard, Dual quad-core Xeons, OS > 10.5.2, with standard fink installations of gcc42, hdf5, and > openmpi, I get the following errors > > ld: warning alignment lost in merging tentative definition > _staticconformalrest_ > ld: warning alignment lost in merging tentative definition > _admbaserest_ > ld: warning alignment lost in merging tentative definition > _spacemaskrest_ > ld: warning alignment lost in merging tentative definition > _admmacrosrest_ > > > Ever seen this before? I'll be checking soon to find out if this > obviously affects the output of the code or not. [Reposted with permission.] Joshua, this seems to be a problem in gfortran. There is a posting on one gcc mailing list (I don't remember the details) stating that gfortran requests too much alignment for certain data structures, especially more than gcc requests. These data structures are used in Cactus to communicate the parameter values from C to Fortran, meaning they are defined by C and used from Fortran. I believe that these warning messages are harmless, and I expect them to go away in future versions of gfortran. -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. -------------- 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/20080313/4bcc13ae/attachment.bin From heeilkim at gmail.com Mon Mar 31 02:38:57 2008 From: heeilkim at gmail.com (Hee Il Kim) Date: Mon, 31 Mar 2008 17:38:57 +0900 Subject: [Users] How to generate a profiling file using mpiP Message-ID: Hi, I'm trying to profile Cactus run using mpiP. I could do it for a simple test and could view the ouput using mpipview. But it failed to make an ouput file for Cactus run. To save building time (actually it seems hang for a sophisticate configuration), I took the minimal configuration for profiling. Do I need more options? Thanks in advance. Hee Il ==== .cactus/config ========== ... CC=icc CXX=icpc F77=ifort F90=ifort CFLAGS = -shared-intel -g CXXFLAGS = -shared-intel -g F77FLAGS = -shared-intel -g F90FLAGS = -shared-intel -g LDFLAGS=-lutil LIBDIRS=/usr/local/fftw3/lib /opt/intel/fce/10.1.011/lib /opt/intel/cce/10.1.011/lib /usr/lib64 /usr/lib/gcc/x86_64-redhat-linux/3.4.6 LIBS= ifcore fftw3 imf g2c PERL=perl FPP=$${PERL} $${CCTK_HOME}/lib/sbin/cpp.pl DEBUG = yes CPP_DEBUG_FLAGS = -DCARPET_DEBUG FPP_DEBUG_FLAGS = -DCARPET_DEBUG C_DEBUG_FLAGS = CXX_DEBUG_FLAGS = F77_DEBUG_FLAGS = F90_DEBUG_FLAGS = PROFILE = yes CPP_PROFILE_FLAGS = FPP_PROFILE_FLAGS = C_PROFILE_FLAGS = -p CXX_PROFILE_FLAGS = -p F77_PROFILE_FLAGS = -p F90_PROFILE_FLAGS = -p MPI=MPICH MPICH_DIR=/opt/mpich2/intel MPI_LIBS=mpiP bfd iberty mpich MPI_LIB_DIRS=/home/khi/local/mpiP/lib /opt/mpich2/intel/lib MPI_INC_DIRS=/opt/mpich2/intel/include ... -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cactuscode.org/pipermail/users/attachments/20080331/4698c997/attachment.html From tradke at aei.mpg.de Mon Mar 31 03:30:49 2008 From: tradke at aei.mpg.de (Thomas Radke) Date: Mon, 31 Mar 2008 11:30:49 +0200 Subject: [Users] How to generate a profiling file using mpiP In-Reply-To: References: Message-ID: <47F0AF49.3080409@aei.mpg.de> Hee Il Kim wrote: > Hi, > > I'm trying to profile Cactus run using mpiP. I could do it for a simple test > and could view the ouput using mpipview. But it failed to make an ouput file > for Cactus run. To save building time (actually it seems hang for a > sophisticate configuration), I took the minimal configuration for profiling. > Do I need more options? Thanks in advance. There are two webpages describing how to profile Cactus simulations with mpiP and IPM on one of our clusters at AEI: http://numrel.aei.mpg.de/Research/Peyote/Docs/mpiP.html http://numrel.aei.mpg.de/Research/Peyote/Docs/IPM.html The information there is about 2 years old but might still work. What's the actual problem with creating an mpiP output file for your Cactus simulation ? -- Cheers, Thomas. From heeilkim at gmail.com Mon Mar 31 13:20:19 2008 From: heeilkim at gmail.com (Hee Il Kim) Date: Tue, 1 Apr 2008 04:20:19 +0900 Subject: [Users] How to generate a profiling file using mpiP In-Reply-To: <47F0AF49.3080409@aei.mpg.de> References: <47F0AF49.3080409@aei.mpg.de> Message-ID: Thanks Thomas, I referred the site you introduced, http://numrel.aei.mpg.de/Research/Peyote/Docs/mpiP.html. I didn't try IPM yet. I recently made a cluster of 10 computing nodes with dual quad core Xeon (Harpertown E5420). The nodes are connected with a usual giagabit switch. Unfortunately, when I tested the system performance using Cactus benchmarks, BSSN_PUGH and BSSN_Carpet_Whisky, I got a bad scalability (~700sec for 1CPU, 1500sec for 64CPU). It's not a code problem. I got similar reports from my colleague that it required unexpectedly large communication time for his GADGET-2 test. I did ping-pong test for the network and heard that it is relatively good as a gigabit system. So, I'm trying to get profiling information. Actually, I don't know what I can do next with it and whether it could be improved by a user-level prescription. Anyway I think the problem looks serious. I don't think it's soley a problem of the large latency network because the benchmarks will not require large communications. Now I'm going to apply Open-MX to reduce latency time. I haven't heard any benchmarks for this Harpertown cluster with gigabit network. I appreciate any comments and suggestions. Wishes, Hee Il 2008/3/31, Thomas Radke : > > Hee Il Kim wrote: > > Hi, > > > > I'm trying to profile Cactus run using mpiP. I could do it for a simple > test > > and could view the ouput using mpipview. But it failed to make an ouput > file > > for Cactus run. To save building time (actually it seems hang for a > > sophisticate configuration), I took the minimal configuration for > profiling. > > Do I need more options? Thanks in advance. > > > There are two webpages describing how to profile Cactus simulations with > mpiP and IPM on one of our clusters at AEI: > > http://numrel.aei.mpg.de/Research/Peyote/Docs/mpiP.html > http://numrel.aei.mpg.de/Research/Peyote/Docs/IPM.html > > The information there is about 2 years old but might still work. > > What's the actual problem with creating an mpiP output file for your > Cactus simulation ? > > > -- > Cheers, Thomas. > _______________________________________________ > Users mailing list > Users at cactuscode.org > http://www.cactuscode.org/mailman/listinfo/users > -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.cactuscode.org/pipermail/users/attachments/20080401/b4ba4c47/attachment.html From schnetter at cct.lsu.edu Mon Mar 31 16:58:50 2008 From: schnetter at cct.lsu.edu (Erik Schnetter) Date: Mon, 31 Mar 2008 17:58:50 -0500 Subject: [Users] How to generate a profiling file using mpiP In-Reply-To: References: <47F0AF49.3080409@aei.mpg.de> Message-ID: <4952A3CC-C372-467A-B735-F1BFD3D1A5BD@cct.lsu.edu> On Mar 31, 2008, at 14:20:19, Hee Il Kim wrote: > Thanks Thomas, > > I referred the site you introduced, http://numrel.aei.mpg.de/Research/Peyote/Docs/mpiP.html > . I didn't try IPM yet. > > I recently made a cluster of 10 computing nodes with dual quad core > Xeon (Harpertown E5420). The nodes are connected with a usual > giagabit switch. Unfortunately, when I tested the system performance > using Cactus benchmarks, BSSN_PUGH and BSSN_Carpet_Whisky, I got a > bad scalability (~700sec for 1CPU, 1500sec for 64CPU). Hee Il, how large is the cluster? A factor of 2 is not really bad, since you still get a factor of 32 speedup compared to running on a single CPU. What is the slowdown when you go from using 1 to using 2 full nodes? -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. -------------- 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/20080331/2a6f3d78/attachment.bin