[Pw_forum] Can I get pw.x and memory.x for my linux pc cluster

Axel Kohlmeyer axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Tue Feb 8 17:24:33 CET 2005


On Tue, 8 Feb 2005, Abdelouahab O. wrote:

AO> Hello
AO> Thanks for your relpay;
AO> This is make.sys

hi,

the current version of pwscf (or Quantum-ESPRESSO) is 2.1.2.
with many new features, bugfixes and improvements.

if you have installed you LAM/MPI properly for the intel
compilers you should be able to just use

CC=mpicc
F77=mpif77
F90=mpif90

alternatively you can try to use 
CC=env LAMHCC=icc mpicc
F77=env LAMHF77=ifort mpif77
F90=env LAMHF77=ifort mpif77

using the mpiXXX wrappers will make sure, that you include
the proper headers that are matching your LAM installation.


AO> =======================================================
AO> # make.sys.  Generated from make.sys.in by configure.
AO> CC             = gcc
AO> CCFLAGS        = -O3 -fomit-frame-pointer $(DFLAGS)
AO> $(IFLAGS)
AO> # See include/machine.h.README for a list of
AO> precompilation options
AO> # (possible arguments to -D or -U) and their meaning
AO> DFLAGS         = -D__LINUX -D__INTEL -D__LAM -D__PARA
AO> -DHAS_ZHEGVX -D__FFTW -D__USE_INTERNAL_FFTW
AO> IFLAGS         = -I. -I../include -I../Modules -I../PW
AO> -I../PH 
AO> CPP            = gcc -E
AO> CPPFLAGS       =  $(DFLAGS) $(IFLAGS)
AO> F77            = ifort
AO> F90            = ifort
AO> FFLAGS         = -Vaxlib -O3 -tpp7 $(IFLAGS)
AO> F77FLAGS       = $(FFLAGS)
AO> F90FLAGS       = $(FFLAGS) -nomodule -fpp $(DFLAGS)
AO> F77FLAGS_NOOPT = -O0
AO> LD             = ifort
AO> LDFLAGS        = -Vaxlib -O3 -tpp7 $(LIBOBJS) $(LIBS)
AO> LIBOBJS        = ../flib/ptools.a ../flib/flib.a
AO> ../clib/clib.a
AO> # LIBS must contain the location of all needed
AO> external libraries
AO> LIBS           = -L/opt/intel/mkl72/lib/32
AO> -lmkl_lapack -lmkl_ia32 -lguide -L/usr/lib/lam
AO> # MYLIB can be one of the following (depending on
AO> LIBS):
AO> # blas       : compile the local copy of blas routines
AO> # lapack     : compile the local copy of lapack
AO> routines
AO> # blas_and_lapack : all of the above - use this for a
AO> quick test
AO> #              or if you don't have an optimized
AO> blas/lapack library
AO> # lapack_ibm : compile only lapack routines not
AO> present in IBM ESSL
AO> #              use this together with IBM ESSL
AO> # lapack_t3e : compile only lapack routines not
AO> present in T3E scilib
AO> #              use this together with T3E scilib
AO> # lapack_mkl : compile only lapack routines not
AO> present in Intel MKL
AO> #              use this together with Intel MKL
AO> MYLIB          = lapack_mkl
AO> AR             = ar
AO> ARFLAGS        = ruv
AO> =======================================================
AO> 
AO> And make all gives:
AO> %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
AO> make[1]: Entering directory
AO> `/home1/wahab/pwscf2.0.4/pwscf/Modules'

[...]

AO> fortcom: Warning: mp_wave.f90, line 130: A dummy
AO> argument with an expl
AO> icit INTENT(OUT) declaration is not given an explicit
AO> value.   [PW]
AO>       SUBROUTINE splitwf ( pw, pwt, ngwl, ig_l2g,
AO> mpime, nproc, root,
AO> comm )
AO> ...........................^
AO> fortcom: Warning: mp_wave.f90, line 23: A dummy
AO> argument with an expli

this looks like a 'feature', that only the recent intel
fortran (and the DEC/Compaq) compilers choke on. this
may have been fixed in the newer releases. i remember
to have seen several of those with version 2.1 which
are now fixed.

AO> -D__USE_INTERNAL_FFTW  -c broadcast.f90
AO> fortcom: Error: broadcast.f90, line 146: This name has
AO> already been assigned a data type.  
AO> [MPI_DOUBLE_PRECISION]
AO>       INTEGER MPI_REAL4, MPI_DOUBLE_PRECISION,
AO> MPI_REAL16
AO> .........................^
AO> compilation aborted for broadcast.f90 (code 1)
AO> make[1]: *** [broadcast.o] Error 1
AO> make[1]: Leaving directory
AO> `/home1/wahab/pwscf2.0.4/pwscf/PW'
AO> make: *** [pw] Error 2


i cannot reproduce this on my desktop with pwscf v2.0.4,
my installation of LAM/MPI version 7.0.6 and the intel 
compiler version 8.1. in my distribution broadcast.f90
only has 41 lines, so your error on line 146 looks 
very suspicious. same is for the content, 'MPI_REAL4'
is only present in MPICH installations and not in LAM/MPI.

if you have administrator priviledge on your machine,
you can download my LAM/MPI package as rpm from:
http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/cpmd-linux.html#mpi
these rpm are compiled special to be compatible to the intel
compilers as well as g77 or the pgi compilers. you simply need 
to use mpiifort/mpiicc instead of mpif77/mpicc.

regards,
	axel.


AO> 
AO>  --- Paolo Giannozzi <giannozz at nest.sns.it> a écrit : 
AO> > On Sunday 06 February 2005 10:46, Abdelouahab O.
AO> > wrote:
AO> > 
AO> > > I have a pc linux cluster with ifort8.0, mkl7.2,
AO> > and
AO> > > LAM7.0.6: the parallel compilation cant be
AO> > finilized
AO> > > using the -D__LAM -D__PARA, the error apppears
AO> > when
AO> > > compiling broadcast.f90
AO> > 
AO> > which error?
AO> > _______________________________________________
AO> > Pw_forum mailing list
AO> > Pw_forum at pwscf.org
AO> > http://www.democritos.it/mailman/listinfo/pw_forum
AO> >  
AO> 
AO> 
AO> 
AO> 	
AO> 
AO> 	
AO> 		
AO> Découvrez le nouveau Yahoo! Mail : 250 Mo d'espace de stockage pour vos mails ! 
AO> Créez votre Yahoo! Mail sur http://fr.mail.yahoo.com/
AO> _______________________________________________
AO> Pw_forum mailing list
AO> Pw_forum at pwscf.org
AO> http://www.democritos.it/mailman/listinfo/pw_forum
AO> 
AO> 

-- 

=======================================================================
Dr. Axel Kohlmeyer   e-mail: axel.kohlmeyer at theochem.ruhr-uni-bochum.de
Lehrstuhl fuer Theoretische Chemie          Phone: ++49 (0)234/32-26673
Ruhr-Universitaet Bochum - NC 03/53         Fax:   ++49 (0)234/32-14045
D-44780 Bochum  http://www.theochem.ruhr-uni-bochum.de/~axel.kohlmeyer/
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.




More information about the Pw_forum mailing list