Re: [SIESTA-L] Posible SPAM: elastic constant calculations
Thank you for your message. I think if we relax the other two lattice parameters at the same time, the obtained values should be Young's modulus, but not C_33. I have checked my calculations, the k points and cutoff energies have been carefully tested. I wonder if the error comes from the basis sets we used (DZP)? Is there anyone who used to perform similar calculations? Is that a common error in SIESTA? I urgently need your help, thank you! Best regards, Chun - Original Message - From: Vasilii Artyukhov To: SIESTA-L@listserv.uam.es Sent: Monday, March 19, 2007 4:41 PM Subject: Re: [SIESTA-L] Posible SPAM: elastic constant calculations I believe that if you're willing to apply strain in the c direction only, you might also want to relax the other two lattice parameters. Obviously this will decrease your values, maybe to something reasonable. 2007/3/19, Chun Li [EMAIL PROTECTED]: Dear SIESTA users, I used SIESTA to calculate the elastic constant C_33 of bulk ZnO. Both LDA and GGA are employed, respectively. But the obtained results are much larger than the experimental value (for LDA, I got 358 GPa; GGA, 282 GPa; but the experimental value is around 210 GPa). I wonder why such big error appears? What important factors might be responsible for that? Could anyone give me some ideas? In my calculations, first the cell is fully relaxed; after that I apply different strains in the c direction and evaluate their total energies, respectively. Then I fit the obtained energies to obtain the elastic constant. In both relaxation and energy calculations, double-zeta polarized numerical atomic orbitals basis sets for both Zn and O are used. Thank you and best regards, Chun Li
Re: [SIESTA-L] paralle problem
Hi, Kun, It seems I have also met the similar problem before. But I tried using mpich-gm to compile then the parallel siesta-2.0 can run efficiently on our cluster (also a AMD-Opteron cluster). So I think you can try pgi. The following is my arch.make file, hope that helps. Chun SIESTA_ARCH=pgf90-gm # FC=/share/data/software/mpich-gm/pgi/bin/mpif90 FC_ASIS=$(FC) # FFLAGS= -fastsse FFLAGS_DEBUG= -g -O0 RANLIB=echo COMP_LIBS=dc_lapack.a # #NETCDF_LIBS=/share/data/software/netcdf-3.6.1/lib/libnetcdf.a #NETCDF_INTERFACE=libnetcdf_f90.a #DEFS_CDF=-DCDF # MPI_ROOT_MYRINET=/share/data/software/mpich-gm/pgi MYRINET_LIBS=-L/opt/gm/lib64 -lgm MPI_LIBS_MYRINET= -L$(MPI_ROOT_MYRINET)/lib -lmpich $(MYRINET_LIBS) # MPI_ROOT_MPICH=//share/data/software/mpich-gm/pgi MPI_LIBS_MPICH= -L$(MPI_ROOT_MPICH)/lib -lmpich # MPI_ROOT=$(MPI_ROOT_MYRINET) MPI_LIBS=$(MPI_LIBS_MYRINET) # MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=$(MPI_ROOT)/include DEFS_MPI=-DMPI # LIBS= -L/share/data/software/acml270/acml2.7.0/pgi64/lib -lscalapack \ -lblacs -lblacsF77init -lblacsCinit -lblacsF77init -lblacs \ -llapack -lblas \ $(MPI_LIBS) $(NETCDF_LIBS) SYS=cpu_time DEFS= $(DEFS_CDF) $(DEFS_MPI) # # # Important (at least for V5.0-1 of the pgf90 compiler...) # Compile atom.f and electrostatic.f without optimization. # atom.o: $(FC) -c $(FFLAGS_DEBUG) atom.f # electrostatic.o: $(FC) -c $(FFLAGS_DEBUG) electrostatic.f # .F.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ .F90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ # - Original Message - From: Tao K [EMAIL PROTECTED] To: SIESTA-L@listserv.uam.es Sent: Wednesday, September 27, 2006 5:13 PM Subject: [SIESTA-L] paralle problem Dear Siesta user, I have compiled the parallel Siesta-2.0 on a AMD-Opteron cluster. When I tried to run it in parallel on 6 nodes, however, I found the head of the out file was : Siesta Version: siesta-2.0-release Architecture : x86_64-unknown-linux-gnu--Nag Compiler flags: mpif90 -axP -xW -ip -O3 SERIAL version * Running in serial mode ... . Does it mean that it runs in sequential not in parallel? But when I checked the nodes, I found the task was running on 6 nodes. I don't know whether these 6 nodes run the same job in parallel or each of them runs the job once time (so the same job was done 6 times by these 6 nodes). Attached below is my arch.make file. Any opinion would be appreciated! Thanks in advance! Sincerely Yours, Kun Tao # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia, J.Junquera, P.Ordejon, D.Sanchez-Portal # and J.M.Soler, 1996-2006. # # Use of this software constitutes agreement with the full conditions # given in the SIESTA license, as signed by all legitimate users. # .SUFFIXES: .SUFFIXES: .f .F .o .a .f90 .F90 SIESTA_ARCH=x86_64-unknown-linux-gnu--Nag FPP= FPP_OUTPUT= FC=mpif90 # FC=ifort RANLIB=ranlib SYS=nag SP_KIND=4 DP_KIND=8 KINDS=$(SP_KIND) $(DP_KIND) FFLAGS=-axP -xW -ip -O3 FPPFLAGS= -DFC_HAVE_FLUSH -DFC_HAVE_ABORT LDFLAGS= ARFLAGS_EXTRA= FCFLAGS_fixed_f= FCFLAGS_free_f90= FLAGS_fixed_F= FPPFLAGS_free_F90= BLAS_LIBS=/data/intel/cmkl/8.1/lib/64/libguide.a LAPACK_LIBS=/data/intel/cmkl/8.1/lib/64/libmkl_lapack.a BLACS_LIBS=/data/intel/cmkl/8.1/lib/64/libmkl_blacs.a SCALAPACK_LIBS=-L/data/intel/cmkl/8.1/lib/64/libmkl_scalapack.a /data/BLACS/LIB/blacsF77init_MPI-LINUX-0.a /data/BLACS/LIB/blacs_MPI-LINUX-0.a COMP_LIBS=linalg.a LIBS=$(SCALAPACK_LIBS) $(BLACS_LIBS) $(LAPACK_LIBS) $(BLAS_LIBS) #SIESTA needs an F90 interface to MPI #This will give you SIESTA's own implementation #If your compiler vendor offers an alternative, you may change #to it here. MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=/data/mpich_ifort/include DEFS_MPI=-DMPI DEFS= $(DEFS_MPI) #Dependency rules are created by autoconf according to whether #discrete preprocessing is necessary or not. .F.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(FPPFLAGS) $(FPPFLAGS_fixed_F) $ .F90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(FPPFLAGS) $(FPPFLAGS_free_F90) $ .f.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(FCFLAGS_fixed_f) $ .f90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(FCFLAGS_free_f90) $
Re: [SIESTA-L] [SIESTA-L: [ SIESTA-L] ( )Re: [SIESTA-L] Test error?
Hi, Dr. Kong, Thank you very much for your timely message! Following your advice, I have managed to run parallel SIESTA on our cluster :). Best regards. Chun - Original Message - From: Yong Kong To: SIESTA-L@listserv.uam.es Sent: Thursday, September 14, 2006 10:06 PM Subject: Re: [SIESTA-L] R e: [ SIESTA-L] ( )Re: [SIESTA-L] Test error? Hi Chun, I just came back to Stuttgart and saw your post in the list. Previously we met similar problem when running siesta-2.0 on our IBM p5 nodes. After trying all kinds of compiling options, we finally found the problem is because we use our old input fdf file directly from siesta-1.3. In that input file, we use AllocReportLevel = 3. However, that doesn't work for siesta-2.0. Simply delete it from your fdf file or set it to 0. It does work for us. --Yong On 9/9/06, Chun Li [EMAIL PROTECTED] wrote: Hi, I have tried that, but the same problem still exist. I wonder is there anyone who have never run parallel SIESTA successfully like me? I am really nervous about that. Chun - Original Message - From: Jyh-Shyong Ho To: SIESTA-L@listserv.uam.es Sent: Friday, September 08, 2006 5:25 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] ()Re: [SIESTA-L] Test error? Please try the PGI version, and see if it works on your computer. Ho Chun Li wrote: Hi, Thank you very much! I have modified your intel-mkl arch.make file and managed to build parallel siesta on our cluster. But the strange problem still exists! In the beginning: Siesta Version: siesta-2.0-release Architecture : x86_64-unknown-linux-gnu--Intel Compiler flags: /opt/mpich/intel/bin/mpif90 -no-ipo -g -w -mp -tpp7 -O3 -no-ipo PARALLEL version NetCDF-capable * Running on4 nodes in parallel Start of run: 8-SEP-2006 17:12:49 When the output reach the following lines, there is not any output then! siesta: Simulation parameters siesta: siesta: The following are some of the parameters of the simulation. siesta: A complete list of the parameters used, including default values, siesta: can be found in file out.fdf siesta: coor: Atomic-coordinates input format = Cartesian coordinates coor: (in Angstroms) I wonder what happened?? Any idea or suggestions are highly appreciated. Chun - Original Message - From: Jyh-Shyong Ho To: SIESTA-L@listserv.uam.es Sent: Friday, September 08, 2006 1:09 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] (鐟炴槦鎻愮ず-姝?閭欢鍙?鑳芥槸鍨冨溇閭欢)Re: [SIESTA-L] Test error? Hi, I have managed to build parallel SIESTA successfully on our dual-core dual Opteron cluster with the following combination: 1. Intel compiler 9 2. OpenMPI library built with Intel compiler 9 3. Intel CMKL 8.0 library for blas, lapack, BLACS and Scalapack library. 4. NETCDF built with Intel compiler 9 and 1. PGI 6.1 compiler 2. OpenMPI library built with PGI compiler 3. BLACS/SCALAPACK library built with OpenMPI/PGI compiler 4. NETCDF built with PGI compiler Perhaps you can try these combination. I have tried mpich and lam but failed to build the executable successfully, so I change to OpenMPI, I found that OpenMPI is much more robust than MPICH and LAM. Jyh-Shyong Ho, Ph.D. Research Scientist National Center for High Performance Computing Hsinchi, Taiwan, ROC Chun Li wrote: Hi, Nestor Correia, Thanks for your kind help. Unfortunately, I still cannot solve the problem. I have spent many days on this strange problem, tried to use mpich-pgi, mpich-intel, intel-mkl, mpich-gm. etc. and feel really.. I wonder if it is impossible to run parallel siesta on our cluster! Thank you anyway. Chun - Original Message - From: Nestor Correia [EMAIL PROTECTED] To: SIESTA-L@listserv.uam.es Sent: Thursday, September 07, 2006 10:23 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] Test error? Hi Chun Li, This one works fine. I followed the instructions from Sebastien Le Roux look in archives. I needed to recompile all libraries. [EMAIL PROTECTED] Src]$ more arch.make # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia, J.Junquera, P.Ordejon, D.Sanchez-Portal # and J.M.Soler, 1996-2006. # # Use of this software constitutes agreement with the full conditions # given in the SIESTA license, as signed by all legitimate users. # SIESTA_ARCH=pgf90-mpich
Re: [SIESTA-L] parallel runing problem
Hi, Marcos, Thank you for your suggestions and kind words. I wonder if it is the problem of ssh/rsh? Our cluster is ssh by default. But in the SIESTA guide, there is no statement about that? Chun - Original Message - From: Marcos Verissimo Alves [EMAIL PROTECTED] To: SIESTA-L@listserv.uam.es Sent: Sunday, September 10, 2006 2:30 AM Subject: Re: [SIESTA-L] Hi Chun, A while ago I posted a text file, to the list, with a detailed set of instructions for compiling siesta and many other libraries from scratch. It must be on the mailing list archive. It might help you. One word of caution: looking at your optimization options, I see you have set one of them to -O3. Be careful with -O3, sometimes, besides making your code slower, can give you incorrect results. Of course this will only be a problem when you get siesta to run, but then you'll have to worry about this kind of things for efficiency matters. If this is of any consolation, it took me a few months to correctly compile and run siesta in a cluster, from the first time. It was siesta 1.3f1p, but as far as I see, a good many people have a hard time compiling and running siesta in parallel right. Cheers, Marcos Hi, I have tried that, but the same problem still exist. I wonder is there anyone who have never run parallel SIESTA successfully like me? I am really nervous about that. Chun - Original Message - From: Jyh-Shyong Ho To: SIESTA-L@listserv.uam.es Sent: Friday, September 08, 2006 5:25 PM Subject: (ç'zæ~Yæ示-æ¤é,®ä»¶å¯èf½æ~¯åzfåo¾é,®ä»¶)Re: [SIESTA-L] ()Re: [SIESTA-L] Test error? Please try the PGI version, and see if it works on your computer. Ho Chun Li wrote: Hi, Thank you very much! I have modified your intel-mkl arch.make file and managed to build parallel siesta on our cluster. But the strange problem still exists! In the beginning: Siesta Version: siesta-2.0-release Architecture : x86_64-unknown-linux-gnu--Intel Compiler flags: /opt/mpich/intel/bin/mpif90 -no-ipo -g -w -mp -tpp7 -O3 -no-ipo PARALLEL version NetCDF-capable * Running on4 nodes in parallel Start of run: 8-SEP-2006 17:12:49 When the output reach the following lines, there is not any output then! siesta: Simulation parameters siesta: siesta: The following are some of the parameters of the simulation. siesta: A complete list of the parameters used, including default values, siesta: can be found in file out.fdf siesta: coor: Atomic-coordinates input format = Cartesian coordinates coor: (in Angstroms) I wonder what happened?? Any idea or suggestions are highly appreciated. Chun - Original Message - From: Jyh-Shyong Ho To: SIESTA-L@listserv.uam.es Sent: Friday, September 08, 2006 1:09 PM Subject: (ç'zæ~Yæ示-æ¤é,®ä»¶å¯èf½æ~¯åzfåo¾é,®ä»¶)Re: [SIESTA-L] (éYç,´æ§¦éZ»æ®ãs-å§?é-î?»æ¬¢éT?é'³èS¥æ§¸é¨å?¨æº?é-î?»æ¬¢)Re: [SIESTA-L] Test error? Hi, I have managed to build parallel SIESTA successfully on our dual-core dual Opteron cluster with the following combination: 1. Intel compiler 9 2. OpenMPI library built with Intel compiler 9 3. Intel CMKL 8.0 library for blas, lapack, BLACS and Scalapack library. 4. NETCDF built with Intel compiler 9 and 1. PGI 6.1 compiler 2. OpenMPI library built with PGI compiler 3. BLACS/SCALAPACK library built with OpenMPI/PGI compiler 4. NETCDF built with PGI compiler Perhaps you can try these combination. I have tried mpich and lam but failed to build the executable successfully, so I change to OpenMPI, I found that OpenMPI is much more robust than MPICH and LAM. Jyh-Shyong Ho, Ph.D. Research Scientist National Center for High Performance Computing Hsinchi, Taiwan, ROC Chun Li wrote: Hi, Nestor Correia, Thanks for your kind help. Unfortunately, I still cannot solve the problem. I have spent many days on this strange problem, tried to use mpich-pgi, mpich-intel, intel-mkl, mpich-gm. etc. and feel really.. I wonder if it is impossible to run parallel siesta on our cluster! Thank you anyway. Chun - Original Message - From: Nestor Correia [EMAIL PROTECTED] To: SIESTA-L@listserv.uam.es Sent: Thursday, September 07, 2006 10:23 PM Subject: (ç'zæ~Yæ示-æ¤é,®ä»¶å¯èf½æ~¯åzfåo¾é,®ä»¶)Re: [SIESTA-L] Test error? Hi Chun Li, This one works fine. I followed the instructions from Sebastien Le Roux look in archives. I needed to recompile all libraries. [EMAIL PROTECTED] Src]$ more arch.make # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia
Re: [SIESTA-L] (瑞星提示-此邮件可 能是垃圾邮件)Re: [SIESTA-L] ( )Re: [SIESTA-L] Test error?
Hi, I have tried that, but the same problem still exist. I wonder is there anyone who have never run parallel SIESTA successfully like me? I am really nervous about that. Chun - Original Message - From: Jyh-Shyong Ho To: SIESTA-L@listserv.uam.es Sent: Friday, September 08, 2006 5:25 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] ()Re: [SIESTA-L] Test error? Please try the PGI version, and see if it works on your computer. Ho Chun Li wrote: Hi, Thank you very much! I have modified your intel-mkl arch.make file and managed to build parallel siesta on our cluster. But the strange problem still exists! In the beginning: Siesta Version: siesta-2.0-release Architecture : x86_64-unknown-linux-gnu--Intel Compiler flags: /opt/mpich/intel/bin/mpif90 -no-ipo -g -w -mp -tpp7 -O3 -no-ipo PARALLEL version NetCDF-capable * Running on4 nodes in parallel Start of run: 8-SEP-2006 17:12:49 When the output reach the following lines, there is not any output then! siesta: Simulation parameters siesta: siesta: The following are some of the parameters of the simulation. siesta: A complete list of the parameters used, including default values, siesta: can be found in file out.fdf siesta: coor: Atomic-coordinates input format = Cartesian coordinates coor: (in Angstroms) I wonder what happened?? Any idea or suggestions are highly appreciated. Chun - Original Message - From: Jyh-Shyong Ho To: SIESTA-L@listserv.uam.es Sent: Friday, September 08, 2006 1:09 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] (鐟炴槦鎻愮ず-姝?閭欢鍙?鑳芥槸鍨冨溇閭欢)Re: [SIESTA-L] Test error? Hi, I have managed to build parallel SIESTA successfully on our dual-core dual Opteron cluster with the following combination: 1. Intel compiler 9 2. OpenMPI library built with Intel compiler 9 3. Intel CMKL 8.0 library for blas, lapack, BLACS and Scalapack library. 4. NETCDF built with Intel compiler 9 and 1. PGI 6.1 compiler 2. OpenMPI library built with PGI compiler 3. BLACS/SCALAPACK library built with OpenMPI/PGI compiler 4. NETCDF built with PGI compiler Perhaps you can try these combination. I have tried mpich and lam but failed to build the executable successfully, so I change to OpenMPI, I found that OpenMPI is much more robust than MPICH and LAM. Jyh-Shyong Ho, Ph.D. Research Scientist National Center for High Performance Computing Hsinchi, Taiwan, ROC Chun Li wrote: Hi, Nestor Correia, Thanks for your kind help. Unfortunately, I still cannot solve the problem. I have spent many days on this strange problem, tried to use mpich-pgi, mpich-intel, intel-mkl, mpich-gm. etc. and feel really.. I wonder if it is impossible to run parallel siesta on our cluster! Thank you anyway. Chun - Original Message - From: Nestor Correia [EMAIL PROTECTED] To: SIESTA-L@listserv.uam.es Sent: Thursday, September 07, 2006 10:23 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] Test error? Hi Chun Li, This one works fine. I followed the instructions from Sebastien Le Roux look in archives. I needed to recompile all libraries. [EMAIL PROTECTED] Src]$ more arch.make # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia, J.Junquera, P.Ordejon, D.Sanchez-Portal # and J.M.Soler, 1996-2006. # # Use of this software constitutes agreement with the full conditions # given in the SIESTA license, as signed by all legitimate users. # SIESTA_ARCH=pgf90-mpich # #FC=pgf90 FC=mpif90 FC_ASIS=$(FC) # FFLAGS= -fast FFLAGS_DEBUG= -g -O0 RANLIB=echo COMP_LIBS=dc_lapack.a # NETCDF_LIBS= # /usr/local/netcdf-3.5/lib/pgi/libnetcdf.a NETCDF_INTERFACE=# libnetcdf_f90.a DEFS_CDF=# -DCDF # MPI_INTERFACE=libmpi_f90.a #MPI_INCLUDE=/usr/local/include MPI_INCLUDE=/usr/local/ibg2/mpi/pgi/mvapich-0.9.5-mlx2.0.1/include DEFS_MPI=-DMPI # # There are (were?) some problems with command-line processing compatibility # that forced the extraction of pgi.aux and pgiarg as independent # libraries (details unfortunately lost) # #LIBS= -L/usr/local/lib/pgi \ # -lscalapack -l1upblas -l1utools -l.pgi.aux -lredist \ # -lfblacs -llapack -lblas \ # -l1umpich -lpgiarg $(NETCDF_LIBS) HOME_LIB=/home/ranescor/lib LIBS= -L/home/ranescor/lib \ -lscalapack \ $(HOME_LIB)/blacsCinit_MPI-LINUX-0.a $(HOME_LIB)/blacsF77init_MPI-LINUX-0.a $(HOME_LIB)/blacs_MPI-LINUX-0.a \ -llapack -lblas \ # -l1upblas -l1utools -l.pgi.aux -lredist \ # -lfblacs -llapack -lblas \ # -l1umpich -lpgiarg $(NETCDF_LIBS) SYS=cpu_time DEFS= $(DEFS_CDF) $(DEFS_MPI) # # # Important (at least for V5.0-1
Re: [SIESTA-L] (瑞星提示-此邮件可 能是垃圾邮件)Re: [SIESTA-L] Test error?
Hi, Nestor Correia, Thanks for your kind help. Unfortunately, I still cannot solve the problem. I have spent many days on this strange problem, tried to use mpich-pgi, mpich-intel, intel-mkl, mpich-gm. etc. and feel really.. I wonder if it is impossible to run parallel siesta on our cluster! Thank you anyway. Chun - Original Message - From: Nestor Correia [EMAIL PROTECTED] To: SIESTA-L@listserv.uam.es Sent: Thursday, September 07, 2006 10:23 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] Test error? Hi Chun Li, This one works fine. I followed the instructions from Sebastien Le Roux look in archives. I needed to recompile all libraries. [EMAIL PROTECTED] Src]$ more arch.make # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia, J.Junquera, P.Ordejon, D.Sanchez-Portal # and J.M.Soler, 1996-2006. # # Use of this software constitutes agreement with the full conditions # given in the SIESTA license, as signed by all legitimate users. # SIESTA_ARCH=pgf90-mpich # #FC=pgf90 FC=mpif90 FC_ASIS=$(FC) # FFLAGS= -fast FFLAGS_DEBUG= -g -O0 RANLIB=echo COMP_LIBS=dc_lapack.a # NETCDF_LIBS= # /usr/local/netcdf-3.5/lib/pgi/libnetcdf.a NETCDF_INTERFACE=# libnetcdf_f90.a DEFS_CDF=# -DCDF # MPI_INTERFACE=libmpi_f90.a #MPI_INCLUDE=/usr/local/include MPI_INCLUDE=/usr/local/ibg2/mpi/pgi/mvapich-0.9.5-mlx2.0.1/include DEFS_MPI=-DMPI # # There are (were?) some problems with command-line processing compatibility # that forced the extraction of pgi.aux and pgiarg as independent # libraries (details unfortunately lost) # #LIBS= -L/usr/local/lib/pgi \ # -lscalapack -l1upblas -l1utools -l.pgi.aux -lredist \ # -lfblacs -llapack -lblas \ # -l1umpich -lpgiarg $(NETCDF_LIBS) HOME_LIB=/home/ranescor/lib LIBS= -L/home/ranescor/lib \ -lscalapack \ $(HOME_LIB)/blacsCinit_MPI-LINUX-0.a $(HOME_LIB)/blacsF77init_MPI-LINUX-0.a $(HOME_LIB)/blacs_MPI-LINUX-0.a \ -llapack -lblas \ # -l1upblas -l1utools -l.pgi.aux -lredist \ # -lfblacs -llapack -lblas \ # -l1umpich -lpgiarg $(NETCDF_LIBS) SYS=cpu_time DEFS= $(DEFS_CDF) $(DEFS_MPI) # # # Important (at least for V5.0-1 of the pgf90 compiler...) # Compile atom.f and electrostatic.f without optimization. # atom.o: $(FC) -c $(FFLAGS_DEBUG) atom.f # electrostatic.o: $(FC) -c $(FFLAGS_DEBUG) electrostatic.f # .F.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ .F90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ # [EMAIL PROTECTED] Src]$ On Thu, 2006-09-07 at 21:30 +0800, Chun Li wrote: Hi, is there anyone can give me a arch.make file for parallel siesta? Thanks a lot! I am really confused about the problem. Regards. Chun - Original Message - From: Chun Li To: SIESTA-L@listserv.uam.es Sent: Wednesday, September 06, 2006 9:10 AM Subject: (瑞星提示-此邮件可能是垃圾邮件)[SIESTA-L] Test error? Dear all, I just compiled parallel siesta-2.0 on our 64bit 2G Opteron CPU cluster. But when I tried to make the examples in the Test folder, the following error appeared (e.g. Fe): Running fe test... == Copying pseudopotential file for Fe... == Running SIESTA as ../../../Src/siesta MPICH-GM Error: Need to obtain the job magic number in GMPI_MAGIC ! make: *** [completed] Error 141 I tried other test examples, the same error also occured! I also tried to submit my own calculation test to our cluster with 4 or 8 CPUs, it can run normally in parallel mode just in the beginning, but hangs there without any error or output after about two minites. The last lines in the output file are: siesta: Simulation parameters siesta: siesta: The following are some of the parameters of the simulation. siesta: A complete list of the parameters used, including default values, siesta: can be found in file out.fdf siesta: coor: Atomic-coordinates input format = Cartesian coordinates coor: (in Angstroms) Could anyone give me some ideas? My arch.make file is as follows: # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia, J.Junquera, P.Ordejon, D.Sanchez-Portal # and J.M.Soler, 1996-2006. # # Use of this software constitutes agreement
Re: [SIESTA-L] Test error?
Hi, is there anyone can give me a arch.make file for parallel siesta? Thanks a lot! I am really confused about the problem. Regards. Chun - Original Message - From: Chun Li To: SIESTA-L@listserv.uam.es Sent: Wednesday, September 06, 2006 9:10 AM Subject: (瑞星提示-此邮件可能是垃圾邮件)[SIESTA-L] Test error? Dear all, I just compiled parallel siesta-2.0 on our 64bit 2G Opteron CPU cluster. But when I tried to make the examples in the Test folder, the following error appeared (e.g. Fe): Running fe test... == Copying pseudopotential file for Fe... == Running SIESTA as ../../../Src/siesta MPICH-GM Error: Need to obtain the job magic number in GMPI_MAGIC ! make: *** [completed] Error 141 I tried other test examples, the same error also occured! I also tried to submit my own calculation test to our cluster with 4 or 8 CPUs, it can run normally in parallel mode just in the beginning, but hangs there without any error or output after about two minites. The last lines in the output file are: siesta: Simulation parameters siesta: siesta: The following are some of the parameters of the simulation. siesta: A complete list of the parameters used, including default values, siesta: can be found in file out.fdf siesta: coor: Atomic-coordinates input format = Cartesian coordinates coor: (in Angstroms) Could anyone give me some ideas? My arch.make file is as follows: # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia, J.Junquera, P.Ordejon, D.Sanchez-Portal # and J.M.Soler, 1996-2006. # # Use of this software constitutes agreement with the full conditions # given in the SIESTA license, as signed by all legitimate users. # SIESTA_ARCH=pgf90-matterhorn-gm # FC=/share/data/software/mpich-gm/pgi/bin/mpif90 FC_ASIS=$(FC) # FFLAGS= -fastsse FFLAGS_DEBUG= -g -O0 RANLIB=echo COMP_LIBS=dc_lapack.a # NETCDF_LIBS= # /usr/local/netcdf-3.5/lib/pgi/libnetcdf.a NETCDF_INTERFACE=# libnetcdf_f90.a DEFS_CDF=# -DCDF # MPI_ROOT_MYRINET=/share/data/software/mpich-gm/pgi MYRINET_LIBS=-L/opt/gm/lib64 -lgm MPI_LIBS_MYRINET= -L$(MPI_ROOT_MYRINET)/lib -lmpich $(MYRINET_LIBS) # MPI_ROOT_MPICH=//share/data/software/mpich-gm/pgi MPI_LIBS_MPICH= -L$(MPI_ROOT_MPICH)/lib -lmpich # MPI_ROOT=$(MPI_ROOT_MYRINET) MPI_LIBS=$(MPI_LIBS_MYRINET) # MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=$(MPI_ROOT)/include DEFS_MPI=-DMPI # LIBS= -L/share/data/software/acml270/acml2.7.0/pgi64/lib -lscalapack \ -lblacs -lblacsF77init -lblacsCinit -lblacsF77init -lblacs \ -llapack -lblas \ $(MPI_LIBS) $(NETCDF_LIBS) SYS=cpu_time DEFS= $(DEFS_CDF) $(DEFS_MPI) # # # Important (at least for V5.0-1 of the pgf90 compiler...) # Compile atom.f and electrostatic.f without optimization. # atom.o: $(FC) -c $(FFLAGS_DEBUG) atom.f # electrostatic.o: $(FC) -c $(FFLAGS_DEBUG) electrostatic.f # .F.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ .F90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ # Thank you and best regards. Chun
[SIESTA-L] Test error?
Dear all, I just compiled parallel siesta-2.0 on our 64bit 2G Opteron CPU cluster. But when I tried to make the examples in the Test folder, the following error appeared (e.g. Fe): Running fe test... == Copying pseudopotential file for Fe... == Running SIESTA as ../../../Src/siesta MPICH-GM Error: Need to obtain the job magic number in GMPI_MAGIC ! make: *** [completed] Error 141 I tried other test examples, the same error also occured! I also tried to submit my own calculation test to our cluster with 4 or 8 CPUs, it can run normally in parallel mode just in the beginning, but hangs there without any error or output after about two minites. The last lines in the output file are: siesta: Simulation parameters siesta: siesta: The following are some of the parameters of the simulation. siesta: A complete list of the parameters used, including default values, siesta: can be found in file out.fdf siesta: coor: Atomic-coordinates input format = Cartesian coordinates coor: (in Angstroms) Could anyone give me some ideas? My arch.make file is as follows: # # This file is part of the SIESTA package. # # Copyright (c) Fundacion General Universidad Autonoma de Madrid: # E.Artacho, J.Gale, A.Garcia, J.Junquera, P.Ordejon, D.Sanchez-Portal # and J.M.Soler, 1996-2006. # # Use of this software constitutes agreement with the full conditions # given in the SIESTA license, as signed by all legitimate users. # SIESTA_ARCH=pgf90-matterhorn-gm # FC=/share/data/software/mpich-gm/pgi/bin/mpif90 FC_ASIS=$(FC) # FFLAGS= -fastsse FFLAGS_DEBUG= -g -O0 RANLIB=echo COMP_LIBS=dc_lapack.a # NETCDF_LIBS= # /usr/local/netcdf-3.5/lib/pgi/libnetcdf.a NETCDF_INTERFACE=# libnetcdf_f90.a DEFS_CDF=# -DCDF # MPI_ROOT_MYRINET=/share/data/software/mpich-gm/pgi MYRINET_LIBS=-L/opt/gm/lib64 -lgm MPI_LIBS_MYRINET= -L$(MPI_ROOT_MYRINET)/lib -lmpich $(MYRINET_LIBS) # MPI_ROOT_MPICH=//share/data/software/mpich-gm/pgi MPI_LIBS_MPICH= -L$(MPI_ROOT_MPICH)/lib -lmpich # MPI_ROOT=$(MPI_ROOT_MYRINET) MPI_LIBS=$(MPI_LIBS_MYRINET) # MPI_INTERFACE=libmpi_f90.a MPI_INCLUDE=$(MPI_ROOT)/include DEFS_MPI=-DMPI # LIBS= -L/share/data/software/acml270/acml2.7.0/pgi64/lib -lscalapack \ -lblacs -lblacsF77init -lblacsCinit -lblacsF77init -lblacs \ -llapack -lblas \ $(MPI_LIBS) $(NETCDF_LIBS) SYS=cpu_time DEFS= $(DEFS_CDF) $(DEFS_MPI) # # # Important (at least for V5.0-1 of the pgf90 compiler...) # Compile atom.f and electrostatic.f without optimization. # atom.o: $(FC) -c $(FFLAGS_DEBUG) atom.f # electrostatic.o: $(FC) -c $(FFLAGS_DEBUG) electrostatic.f # .F.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ .F90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $(DEFS) $ .f90.o: $(FC) -c $(FFLAGS) $(INCFLAGS) $ # Thank you and best regards. Chun
Re: [SIESTA-L] A suggestion to all chinese users
Dear SIESTA users, I think the original idea of Dr. Wu is positive, but his expression is short of consideration indeed. Dr. Wu is a warm-hearted person, who used to help me for using the code ABINIT when I was a beginner. But this time I think Dr. Wu should consider netiquette as it stands, but not involve other factors, especially the nationality. In addition, frankly, I am also a beginner of SIESTA. I can imagine that any beginner will not paste their questions without any thought. So I hope anyone with experience can give beginners more understandings. Of course, we beginers should also notice the netiquette to furthest take advantage of the forum resource. Anyway, hope all the users will enjoy this forum and learn more. Thanks for your attention. Chun - Original Message - From: Mingsu Si To: SIESTA-L@listserv.uam.es Sent: Sunday, September 03, 2006 6:20 PM Subject: (瑞星提示-此邮件可能是垃圾邮件)Re: [SIESTA-L] A suggestion to all chinese users I am sorry for my question. But there are some thing to be mentioned. In the course of calculating the BN graphitic sheet, I design some structures but the result is very strange. The pure BN graphitic sheet behaves magnetic moment. So I think the structure may be wrong. I want to know the lattice parameters and test the results. Howere, I solve this problem by reading some papers now. In the begining I think the forum is free for any questions. As a beginner, We maybe only put some poor quality questions. In fact, not all the difficult and high quality questions are able to be sloved by Dr. Wu. I think so. Don't you think so? I think the puzzled question which may be simple to you is valueble to issue in the forum. Finally, I feel woeful and vague to Dr. Wu's suggestions. - Original Message - From: Wu Rongqin To: SIESTA-L@listserv.uam.es Sent: Sunday, September 03, 2006 10:24 AM Subject: [SIESTA-L] A suggestion to all chinese users Dear all, I, a Chinese PhD doing first principles calculation and a user of Siesta (I have a publication using Siesta), am very glad that in recent years so many Chinese students are devoted to first principle calculation. I am also glad that Siesta is favored by them (one can see this from questions rised in the forum). Many of them feel quite free to discuss problems in this forum, this is a good news. HOWEVER, I WANT GIVE ALL CHINESE USERS OF SIESTA. WHEN YOU WANT TO PASTE YOUR PROBLEMS TO THE FORUM, PLEASE GIVE YOUR PROBLEMS A SECOND THOUGHT! ACTUALLY YOU MIGHT FIND THAT YOU YOURSELF CAN SOLVE YOUR PROBLEMS AND THUS THERE IS NO NEED TO PUT YOUR QUESTIONS TO THIS FORUM. QUESTIONS ARE WELCOME BUT WHAT AN EVALUATION OTHERS WILL GIVE YOU IF YOUR QUESTIONS COME UP WITH POOR QUALITY? HOW WESTERN PEOPLE WILL EVALUTE CHINISE STUDENT? JUST THINK OF THIS. -Original Message- From: Siesta, Self-Consistent DFT LCAO program, http://www.uam.es/siesta [mailto:[EMAIL PROTECTED] On Behalf Of Mingsu Si Sent: Sunday, September 03, 2006 9:51 AM To: SIESTA-L@listserv.uam.es Subject: [SIESTA-L] about graphitic BN nanotube Dear all, How to write the lattice parameters of graphitic BN nanotube?
[SIESTA-L] pseudopotential problem?
Dear all: I need help eagerly. I encouter a problem after generating the Pseudopotential of Carbon, which makes the calculation can't run rightly. A error [forrtl: severe (41): insufficient virtual memory] always arises.My C.tm2.inp is : pg C Pseudopotencial tm2 2.00 Cca 0 12 20 2.00 21 2.00 1.24 1.24 0.0 0.0 0.0 # #2345678901234567890123456789012345678901234567890 Ruler Please help me! Regards Zhuhua Zhang
Re: [SIESTA-L] alloc_err: allocate status error 4205
Hi, Joaquin Ortega, Have you solved this problem? I met the similar problem when I performed polarization calculations (108 atoms). But the calculations can be completed without problem when smaller systems are studied. Could you give me some ideas? Thanks a lot. Best regards. Chun - Original Message - From: Joaquin Ortega To: SIESTA-L@LISTSERV.UAM.ES Sent: Monday, December 19, 2005 6:51 PM Subject: [SIESTA-L] alloc_err: allocate status error 4205 Hi all. I am trying to run the ferripyrophyllite (160 atoms, 300 cutoff, etc) in a parallel sgi cluster but I receive the next error message for some calculations: alloc_err: allocate status error 4205 alloc_err: array unknown requested by initatomlists alloc_err: dim, lbound, ubound: 1 1 0 alloc_err: dim, lbound, ubound: 2 48 alloc_err: allocate status error 4205 alloc_err: array unknown requested by initatomlists alloc_err: dim, lbound, ubound: 1 1 0 alloc_err: dim, lbound, ubound: 2 96 alloc_err: allocate status error 4205 Does anyone have a tip or which is the solution about this problem? Thanks in advance, Joaquin. PhD. Joaquin Ortega Castro. Estación Experimental del Zaidin (CSIC). C/ Profesor Albareda n 1 - CP 18008. Granada, Spain Telephone: 958 181600 (Ext. 245).