[CPMD-list] PC-PGI-MPI and testresults
Juerg Hutter
hutter at pci.unizh.ch
Wed Mar 5 11:34:06 CET 2003
Dear Veronique
CPMD uses for perfomance reasons initialization and
optimization procedures that depend on the number
of CPU's used. This means that the way a run converges
may be different if you use a serial or parallel
executable. If the convergence criteria is set to be
very tight, the final results should, however, be
identical.
Juerg
----------------------------------------------------------
Juerg Hutter Phone : ++41 1 635 4491
Physical Chemistry Institute FAX : ++41 1 635 6838
University of Zurich E-mail: hutter at pci.unizh.ch
Winterthurerstrasse 190
CH-8057 Zurich, Switzerland
----------------------------------------------------------
On Tue, 4 Mar 2003, Veronique Van Speybroeck wrote:
> Dear all,
>
> I now compiled the code again without the atlas libraries. The problem
> is not related with this.
>
> I then started from the same restart file three times using following
> commands :
> cpmd.x C.inp > C.out
> mpirun -np 1 cpmdmpi.x C.inp > C1.out
> mpirun -np 2 cpmdmpi.x C.inp > C2.out
>
> as inputfile I used :
> &CPMD
> RESTART ALL
> OPTIMIZE WAVEFUNCTION
> MAXSTEP
> 1
> LSD
> &END
>
> &SYSTEM
> SYMMETRY
> 1
> CELL
> 7.0 1.0 1.0 0.0 0.0 0.0
> CUTOFF
> 25.
> MULTIPLICITY
> 3
> &END
>
> &ATOMS
> *C_VDB_U BINARY NEWF
> 1 1 2
> 1
> 0.0 0.0 0.0
> &END
>
> &DFT
> LDA CORRELATION PZ
> FUNCTIONAL BP
> GC-CUTOFF
> 5.D-7
> &END
>
> so I only did one step. In this case I get the same energies three times.
>
> Still when I start from the same inputfile without restarting and also
> doing one step, I get slightly different energies . I still do not know
> which options or settings are different in the three codes (not
> parallel, parallel on one processor, parallel on two processors).
>
> From the above results I must conclude that the initial guess for the
> wavefunctions is determining. Is this correct ?
>
> I think the problem is moreless solved. However I am not able to trace
> the origin of the differences that remain when not restarting from the
> same restart file. Maybe somebody of you knows?
>
> thanks anyway for all the usefull help
>
> veronique
>
> Juerg Hutter wrote:
>
> >Hi
> >
> >have a look at this mail
> >
> >http://www.cpmd.org/pipermail/cpmd-list/2003-February/000693.html
> >
> >If you still get different results for serial and parallel
> >jobs there might be a problem with your system (hardware
> >or software) or you found a new CPMD bug.
> >
> >greetings
> >
> >Juerg
> >
> >----------------------------------------------------------
> >Juerg Hutter Phone : ++41 1 635 4491
> >Physical Chemistry Institute FAX : ++41 1 635 6838
> >University of Zurich E-mail: hutter at pci.unizh.ch
> >Winterthurerstrasse 190
> >CH-8057 Zurich, Switzerland
> >----------------------------------------------------------
> >
> >
> >On Fri, 28 Feb 2003, Veronique Van Speybroeck wrote:
> >
> >
> >
> >> Dear all,
> >>
> >>thanks first for all your help, I now succeeded in compiling the code to
> >>construct a parallel version and also the Vanderbilt pseudopotentials
> >>are properly read in.
> >>
> >>Now I have done some simple test runs on one Carbon atom both with
> >>Vanderbilt and Martin Trouiller pseudopotentials and with different
> >>compilations of the code. The results for the energy are given at the
> >>bottom of this email message. It seems that I get slightly different
> >>energy results depending on the flags I use in the makefile (which I can
> >>understand since, other libraries are used), the number of processors a
> >>parallel code is run on (this seems very strange to me, since I then
> >>start from the same executable), the parallel and not parallel code.
> >>The results are qualitatively the same for the Trouiller Martin PP's,
> >>these results are only given for completeness.
> >>
> >>Are those things normal?
> >>
> >>Thanks in advance
> >>
> >>veronique
> >>
> >>
> >>
> >>C Vanderbilt pseudopotentials :
> >>SOURCENOATLAS
> >>NOT PARALLEL with following flags
> >>SRC = .
> >>DEST = .
> >>BIN = .
> >>#QMMM_FLAGS = -D__QMECHCOUPL
> >>#QMMM_LIBS = -L. -lmm
> >>FFLAGS = -Mr8 -Msignextend -Msecond_underscore
> >>LFLAGS = -llapack -lblas $(QMMM_LIBS)
> >>CFLAGS =
> >>CPP = /lib/cpp -P -C -traditional
> >>CPPFLAGS = -D__Linux -D__PGI -DLAPACK -DFFT_DEFAULT
> >>CC = gcc -O2 -Wall
> >>FC = pgf90 -c -fast
> >>LD = pgf90 -fast
> >>AR =
> >>
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.42206348 A.U.
> >> (K) KINETIC ENERGY = 2.99486758 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.24561528 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -1.90747711 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 0.35662877 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.62046743 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.07545803 A.U.
> >>
> >>SOURCEMPINOATLAS 1 processor :
> >>Parellel version with following flags run on one processor
> >>SRC = .
> >>DEST = .
> >>BIN = .
> >>MPI-LIB= -I/usr/include/lam -L/usr/lib/lam -I/usr/include -L/usr/lib
> >>#QMMM_FLAGS = -D__QMECHCOUPL
> >>#QMMM_LIBS = -L. -lmm
> >>FFLAGS = -Mr8 -Msignextend -Msecond_underscore $(MPI-LIB)
> >>LFLAGS = -llapack -lblas $(QMMM_LIBS) $(MPI-LIB)
> >>CFLAGS =
> >>CPP = /lib/cpp -P -C -traditional
> >>CPPFLAGS = -D__Linux -D__PGI -DLAPACK -DFFT_DEFAULT \
> >> -DPARALLEL -DMP_LIBRARY=__MPI -DMYRINET
> >>CC = mpicc -O2 -Wall
> >>FC = mpif77 -c -fast
> >>LD = mpif77 -fast
> >>AR =
> >>
> >>
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.42196959 A.U.
> >> (K) KINETIC ENERGY = 2.99486758 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.24560347 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -1.90747711 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 0.35662877 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.62038535 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.07541052 A.U.
> >>
> >>SOURCEMPINOATLAS 2 processoren
> >>same as previously but run with two processors
> >>
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.42171352 A.U.
> >> (K) KINETIC ENERGY = 2.99486758 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.24556515 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -1.90747711 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 0.35662877 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.62016761 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.07531574 A.U.
> >>
> >>
> >>SOURCEMPIATLAS 1 processor
> >>Parallel version with following flags
> >>SRC = .
> >>DEST = .
> >>BIN = .
> >>MPI-LIB= -I/usr/include/lam -L/usr/lib/lam -I/usr/include -L/usr/lib
> >>#QMMM_FLAGS = -D__QMECHCOUPL
> >>#QMMM_LIBS = -L. -lmm
> >>FFLAGS = -Mr8 -Msignextend -Msecond_underscore $(MPI-LIB)
> >>LFLAGS = -llapack -lf77blas -latlas $(QMMM_LIBS) $(MPI-LIB)
> >>CFLAGS =
> >>CPP = /lib/cpp -P -C -traditional
> >>CPPFLAGS = -D__Linux -D__PGI -DLAPACK -DFFT_DEFAULT \
> >> -DPARALLEL -DMP_LIBRARY=__MPI -DMYRINET
> >>CC = mpicc -O2 -Wall
> >>FC = mpif77 -c -fast
> >>LD = mpif77 -fast
> >>AR =
> >>
> >>
> >>
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.42201709 A.U.
> >> (K) KINETIC ENERGY = 2.99486758 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.24560828 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -1.90747711 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 0.35662877 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.62042804 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.07544533 A.U.
> >>
> >>
> >>SOURCEMPIATLAS 2 processoren
> >>same as previously but with two processors
> >>
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.42168874 A.U.
> >> (K) KINETIC ENERGY = 2.99486758 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.24556230 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -1.90747711 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 0.35662877 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.62014568 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.07530204 A.U.
> >>
> >>SOURCE
> >>
> >>C Martin Trouiller pseudopotentials
> >>SOURCENOATLAS
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.30456829 A.U.
> >> (K) KINETIC ENERGY = 3.23978713 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.21842956 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -2.82834506 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 1.08020021 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.57778101 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.06456592 A.U.
> >>
> >>
> >>
> >>SOURCEMPINOATLAS 1 processor
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.30472355 A.U.
> >> (K) KINETIC ENERGY = 3.23978713 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.21847556 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -2.82834506 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 1.08020021 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.57789028 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.06456999 A.U.
> >>
> >>
> >>
> >>SOURCEMPINOATLAS 2 processoren
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.30491191 A.U.
> >> (K) KINETIC ENERGY = 3.23978713 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.21851553 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -2.82834506 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 1.08020021 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.57803866 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.06462282 A.U.
> >>
> >>SOURCEMPIATLAS 1 processor
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.30460928 A.U.
> >> (K) KINETIC ENERGY = 3.23978713 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.21844302 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -2.82834506 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 1.08020021 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.57780855 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.06456282 A.U.
> >>
> >>
> >>SOURCEMPIATLAS 2 processoren
> >> (K+E1+L+N+X) TOTAL ENERGY = -5.30469705 A.U.
> >> (K) KINETIC ENERGY = 3.23978713 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.21845926 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -2.82834506 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 1.08020021 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.57788008 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.06459670 A.U.
> >>
> >>
> >>SOURCE
> >>(K+E1+L+N+X) TOTAL ENERGY = -5.30458135 A.U.
> >> (K) KINETIC ENERGY = 3.23978713 A.U.
> >> (E1=A-S+R) ELECTROSTATIC ENERGY = -5.21843460 A.U.
> >> (S) ESELF = 5.31923041 A.U.
> >> (R) ESR = 0.00000000 A.U.
> >> (L) LOCAL PSEUDOPOTENTIAL ENERGY = -2.82834506 A.U.
> >> (N) N-L PSEUDOPOTENTIAL ENERGY = 1.08020021 A.U.
> >> (X) EXCHANGE-CORRELATION ENERGY = -1.57778903 A.U.
> >> GRADIENT CORRECTION ENERGY = -0.06456248 A.U.
> >>
> >>
> >>
> >>
> >>
> >>
> >>
> >>--
> >>-----------------------------------------------------------------------
> >>Dr. ir. Van Speybroeck Veronique
> >>Laboratorium voor Theoretische Fysica
> >>Universiteit Gent
> >>Proeftuinstraat 86
> >>9000 Gent
> >>Tel +32-9-264.65.58 GSM : +32/474/259767
> >>Fax +32-9-264.65.60
> >>email : veronique.vanspeybroeck at rug.ac.be
> >>http://inwfaxp2.rug.ac.be/~web/onderzoek/proj_3.html
> >>----------------------------------------------------------
> >>
> >>-----------------------------------------------------------------------
> >>Dr. ir. Van Speybroeck Veronique
> >>Laboratorium voor Theoretische Fysica
> >>Universiteit Gent
> >>Proeftuinstraat 86
> >>9000 Gent
> >>Tel +32-9-264.65.58 GSM : +32/474/259767
> >>Fax +32-9-264.65.60
> >>email : veronique.vanspeybroeck at rug.ac.be
> >>http://inwfaxp2.rug.ac.be/~web/onderzoek/proj_3.html
> >>-----------------------------------------------------------------------
> >>
> >>
> >>
> >>_______________________________________________
> >>CPMD-list mailing list
> >>CPMD-list at cpmd.org
> >>http://www.cpmd.org/mailman/listinfo/cpmd-list
> >>
> >>
> >>
>
> --
> -----------------------------------------------------------------------
> Dr. ir. Van Speybroeck Veronique
> Laboratorium voor Theoretische Fysica
> Universiteit Gent
> Proeftuinstraat 86
> 9000 Gent
> Tel +32-9-264.65.58 GSM : +32/474/259767
> Fax +32-9-264.65.60
> email : veronique.vanspeybroeck at rug.ac.be
> http://inwfaxp2.rug.ac.be/~web/onderzoek/proj_3.html
> -----------------------------------------------------------------------
>
>
>
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://www.cpmd.org/mailman/listinfo/cpmd-list
>
More information about the CPMD-list
mailing list