Hi Shihzheng,
I suppose it has worked, then :) The problem when you update a whole system
is that sometimes libraries change version and therefore features inside it
change as well. Ideally you should stick to a certain distribution for as
long as you can and install the updates forf that specific version.
What I would change in your arch.make is the level of optimization to -O2,
which is a relatively safe one in ifort and, if you are not too
knowledgeable about debugging, quit the -g flag altogether. The -g flag
produces debugging information, and this slows down the code. One thing you
could do is to keep a second compilation for use in case you have a big
problem and can debug, or if you can easily get help from a knowledgeable
person for debugging. There is also an option, which is padding variables
with zeroes to get them properly aligned, which improved my performance
quite a bit.
Marcos
On Thu, Jun 10, 2010 at 2:59 AM, shizheng wen <chsz...@gmail.com> wrote:
> Dear Marcos,
> Thank you very much for your guideness.
> The OS is reinstalled instead of upgraded. So, after install the OS,
> the mpich, mkl lib et. al. all are recompiled. (I use the MKL).
> It's strange that this may due to the OS. We use the same type of OS,
> only a new version.
> We export the path of the mpi and mkl lib. Set the 'ulimit -s
> unlimited' in my .bashrc.
> Do you think that I should change some FLAGS? Do this may effect the
> result?
> For convenient, the arch.make is also attached again.
> Thank you for all of you and the your time.
> Best regards,
> Shizheng
>
> PS:
> the mpi is mpich2-1.0.8, ifor-v10.1 and mkl-10.1.1.019
>
> the arch.make is:
>
> #
> # 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- .
> #
> # 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=intel64_RHEL5.4
>
> FPP=
> FPP_OUTPUT=
> FC=mpif90
> RANLIB=ranlib
>
> SYS=nag
>
> SP_KIND=4
> DP_KIND=8
> KINDS=$(SP_KIND) $(DP_KIND)
>
> FFLAGS=-g -O1 -static
> FFLAGS_DEBUG=-g
> FFLAGS_CHECKS=-g -O0 -debug full -traceback -C
> FPPFLAGS= -DMPI -DFC_HAVE_FLUSH -DFC_HAVE_ABORT
> LDFLAGS=-Vaxlib -static
>
> ARFLAGS_EXTRA=
>
> FCFLAGS_fixed_f=
> FCFLAGS_free_f90=
> FPPFLAGS_fixed_F=
> FPPFLAGS_free_F90=
>
> SCALAPACK=-L/opt/intel/cmkl/10.1.1.019/lib/em64t -lmkl_scalapack_ilp64
> SOLVER=-L/opt/intel/cmkl/10.1.1.019/lib/em64t -lmkl_solver_ilp64
> BLACS= -L/opt/intel/cmkl/10.1.1.019/lib/em64t -lmkl_blacs_ilp64
> LAPACK=-L/opt/intel/cmkl/10.1.1.019/lib/em64t -lmkl_lapack
> ITHREAD=-L/opt/intel/cmkl/10.1.1.019/lib/em64t -lmkl_sequential
> ICORE=-L/opt/intel/cmkl/10.1.1.019/lib/em64t -lmkl_core
> BLAS= -L/opt/intel/cmkl/10.1.1.019/lib/em64t -lmkl_intel_ilp64
> GUIDE=-L/opt/intel/cmkl/10.1.1.019/lib/em64t -lguide
> PTHREAD=-lpthread
>
> COMP_LIBS=
>
> NETCDF_LIBS=
> NETCDF_INTERFACE=
>
> LIBS=$(SCALAPACK) $(SOLVER) $(BLACS) $(LAPACK) $(ITHREAD) $(ICORE)
> $(BLAS) $(GUIDE) $(PTHREAD)
>
> #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=/usr/local/include
>
> #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) $<
>
> 2010/6/9 Marcos Veríssimo Alves <marcos.verissimo.al...@gmail.com>:
> > Shihzheng,
> >
> > If you have upgraded your OS then it is likely that you could have to
> > recompile mpich, scalapack and blacs from scratch. As far as I know
> siesta
> > has no problems with mpich in general.
> >
> > Cheers,
> >
> > Marcos
> >
> > On Jun 9, 2010 1:58 PM, "shizheng wen" <chsz...@gmail.com> wrote:
> >
> > Dear Marcos,
> > Thank you very much!
> > First, I try to run with the small system, C6H4N2H4, with mpirun -np 2
> > siesta < input.fdf > output.out.
> > The same error appeared.
> > Then, I recompiled the program, and run it. The same!.
> > The only different from before is that we use the newer version of
> > Linux--Red Hat Enterprise Linux Server release 5.3。
> > Is it that siesta has some problem with the mpich2-1.0.8?
> > In the mail-list it said that it has the same problem with Intel
> > compiler of version 11, but not 10.1.
> >
> > Thank you very much in advance!
> >
> > Best regards,
> > Shizheng
> >
> > PS:
> >
> > The input file:
> >
> > # test mpi
> >
> > SystemName diam
> > SystemLabel diam_opt
> >
> >
> > NumberOfAtoms 16
> > NumberOfSpecies 3
> >
> > %block ChemicalSpeciesLabel
> > 1 6 C
> > 2 7 N
> > 3 1 H
> > %endblock ChemicalSpeciesLabel
> >
> > PAO.BasisSize DZP
> >
> > #LatticeConstant 1. Ang
> >
> > #%block LatticeVectors
> > # 4.16 0.0000 0.0000
> > # 0.0000 4.16 0.0000
> > # 0.0000 0.0000 8.32
> > #%endblock LatticeVectors
> >
> > AtomicCoordinatesFormat Ang
> >
> > %block AtomicCoordinatesAndAtomicSpecies
> > 0.696236 -1.200334 -0.005944 1
> > -0.696186 -1.200391 0.005941 1
> > -1.421078 -0.000046 0.011328 1
> > -0.696242 1.200335 0.005889 1
> > 0.696193 1.200390 -0.005947 1
> > 1.421078 0.000053 -0.011397 1
> > 1.229238 -2.148265 -0.017528 3
> > -1.229144 -2.148338 0.017781 3
> > -1.229228 2.148274 0.017497 3
> > 1.229131 2.148350 -0.017724 3
> > 2.827689 0.000037 -0.090437 2
> > 3.253412 -0.829442 0.304092 3
> > -2.827680 -0.000084 0.090510 2
> > -3.253443 0.829583 -0.303581 3
> > -3.253580 -0.828983 -0.305057 3
> > 3.253544 0.829123 0.304785 3
> > %endblock AtomicCoordinatesAndAtomicSpecies
> >
> >
> > #%block kgrid_Monkhorst_Pack
> > # 20 0 0 0.0
> > # 0 20 0 0.0
> > # 0 0 20 0.0
> > #%endblock kgrid_Monkhorst_Pack
> >
> > xc.functional GGA # 'LDA', 'GGA'
> > xc.authors PBE # 'CA'='PZ', 'PW92', 'PBE'
> > SpinPolarized F
> > FixSpin F
> > TotalSpin 0.0
> > NonCollinearSpin F
> > MeshCutoff 300.0 Ry
> > MaxSCFIterations 500
> >
> > DM.MixingWeight 0.1
> > DM.NumberPulay 7 # Pulay convergency accelerator
> > DM.MixSCF1 F
> > DM.PulayOnFile F # Store in memory ('F') or in files
> ('T')
> > DM.Tolerance 5.0E-5
> > DM.UseSaveDM T
> >
> >
> > NeglNonOverlapInt T # F does not neglect
> > SolutionMethod diagon
> > ElectronicTemperature 300 K # Default value
> >
> >
> > MD.TypeOfRun CG
> > #MD.VariableCell T
> > #MD.PreconditionVariableCell 5. Ang
> > MD.NumCGsteps 500
> > MD.MaxCGDispl 0.01 Ang
> > MD.MaxForceTol 0.02 eV/Ang
> >
> >
> > WriteCoorXmol F
> > SaveElectrostaticPotential T
> > SaveHS F # Save the Hamiltonian and
> > Overlap matrices
> > SaveRho F # Save the valence pseudocharge
> > density
> > SaveDeltaRho F
> > WriteDenchar F # Write Denchar output
> > WriteDMT T
> > WriteEigenvalues T
> > WriteMullikenPop 1
> > LongOutput T
> > UseSaveData T
> > WriteDM T
> >
> > last part of the output file:
> >
> > InitMesh: MESH = 144 x 108 x 72 = 1119744
> > InitMesh: Mesh cutoff (required, used) = 300.000 309.331 Ry
> >
> > * Maximum dynamic memory allocated = 68 MB
> > rank 1 in job 3 pc2_40434 caused collective abort of all ranks
> >
> > exit status of rank 1: return code 1
> >
> > rank 0 in job 3 pc2_40434 caused collective abort of all ranks
> >
> > exit status of rank 0: return code 1
> >
> > And in the screen is:
> >
> > Fatal error in MPI_Comm_size: Invalid communicator, error stack:
> > MPI_Comm_size(112): MPI_Comm_size(...
> >
> > MPI_Comm_size(70).: Invalid communicator 98.9 0.4 0:06.55
> >
> > Fatal error in MPI_Comm_size: Invalid communicator, error stack:
> >
> > MPI_Comm_size(112): MPI_Comm_size(comm=0x5b, size=0x13c46e8) failedion/0
> > MPI_Comm_size(70).: Invalid communicator[cli_1]: aborting job:softirqd/0
> > Fatal error in MPI_Comm_size: Invalid communicator, error stack:chdog/0
> > MPI_Comm_size(112): MPI_Comm_size(comm=0x5b, size=0x13c46e8) failedion/1
> > MPI_Comm_size(70).: Invalid communicator 0.0 0.0 0:00.00
> >
> > 2010/6/9 Marcos Veríssimo Alves <marcos.verissimo.al...@gmail.com>:
> >
> >> Shizheng, > Have you tried it for small systems? 12 GB Ram for Siesta is
> >> enough for > running syst...
> >
> > --
> >
> > Sincerely, Shizheng Wen Dept. of Chem., NENU. Changchun, Jilin, China
>
>
>
> --
> Sincerely,
> Shizheng Wen
> Dept. of Chem., NENU.
> Changchun, Jilin, China
>
