[Pw_forum] Build QE for Intel Phi native run

Hideaki Kuraishi Hideaki.Kuraishi at uk.fujitsu.com
Fri Feb 6 18:54:26 CET 2015


Dear Filippo,

Yes, sure. 
What I did is as follows,

[Compiler and MPI Version]
Intel compiler    : composer_xe_2013_sp1.3.174
Intel MPI         : 5.0.0.028

[Machine environment]
CentOS Linux release 7.0.1406 (Core)

[Procedure]
1) Configure

$ export FC=mpiifort
$ export F90=$FC
$ export F77=$FC
$ export MPIF90=$FC
$ export FCFLAGS="-mmic -openmp -mkl"
$ export FFLAGS=$FCFLAGS
$ export CC=mpiicc
$ export CPP="icc -E"
$ export CFLAGS=$FCFLAGS
$ export AR=xiar

$ ./configure --enable-openmp --enable-parallel --with-internal-blas --with-internal-lapack --without-scalapack

2) Build
2-1) Compile internal BLAS

$ vi make.inc

FORTRAN  = mpiifort
OPTS     = -mmic -openmp -mkl
DRVOPTS  = $(OPTS)
NOOPT    = -mmic -openmp -mkl
LOADER   = mpiifort
LOADOPTS =

$  make

2-2) Compile internal LAPACK

$ vi make.inc

FORTRAN  = mpiifort
OPTS     = -mmic -openmp -mkl
DRVOPTS  = -mmic -openmp -mkl
NOOPT    = -mmic -openmp
LOADER   = mpiifort
LOADOPTS = -mmic -openmp

BLASLIB      = $HOME/espresso-5.0.2/BLAS/blas.a

$ make

2-3) Modify "make.sys"

$ vi make.sys

CFLAGS         = -mmic -openmp -mkl $(DFLAGS) $(IFLAGS)
F90FLAGS       = $(FFLAGS) -nomodule -openmp -fpp $(FDFLAGS) $(IFLAGS) $(MODFLAGS)
FFLAGS         = -mmic -openmp -mkl

FFLAGS_NOOPT   =

LDFLAGS        = -static-intel -mmic -openmp

BLAS_LIBS      = $HOME/espresso-5.0.2/BLAS/blas.a
BLAS_LIBS_SWITCH = internal

LAPACK_LIBS = $HOME/espresso-5.0.2/lapack-3.2/lapack.a
LAPACK_LIBS_SWITCH = internal

LAPACK_LIBS    = $HOME/espresso-5.0.2/lapack-3.2/lapack.a
LAPACK_LIBS_SWITCH = internal

FFT_LIBS       =

MPI_LIBS       = -L/opt/intel/impi/5.0.0/mic/lib –lmpi_mt -ltmi

$ make all 2>&1 | tee LOG.make

3) Run

$ $QE_DIR/pw.x -inp ausurf.in

Hope this is helpful...

Best regards
Kai

> -----Original Message-----
> From: Filippo Spiga [mailto:spiga.filippo at gmail.com]
> Sent: Friday, February 06, 2015 4:31 PM
> To: Hideaki Kuraishi
> Cc: <q-e-developers at qe-forge.org>; PWSCF Forum
> Subject: Re: [Pw_forum] Build QE for Intel Phi native run
> 
> Dear Hideaki,
> 
> can you share information about compiler versions, flags, changes in the
> make.sys so we can add in the regular distribution instructions how to
> get up&running on Intel Phi? We will acknowledge you for the help :-)
> 
> Thanks in advance.
> 
> Regards,
> Filippo
> 
> On Feb 6, 2015, at 8:18 AM, Hideaki Kuraishi
> <Hideaki.Kuraishi at uk.fujitsu.com> wrote:
> > Hi Filippo,
> >
> > Thanks to the comment below,
> > I now able to compile and run the code.
> > Thanks a lot for the advice!
> >
> > Best regards
> > Kai
> >
> >> -----Original Message-----
> >> From: pw_forum-bounces at pwscf.org
> [mailto:pw_forum-bounces at pwscf.org]
> >> On Behalf Of Filippo Spiga
> >> Sent: Friday, February 06, 2015 4:03 PM
> >> To: PWSCF Forum
> >> Cc: <q-e-developers at qe-forge.org>
> >> Subject: Re: [Pw_forum] Build QE for Intel Phi native run
> >>
> >> Dear Hideaki,
> >>
> >> please keep us posted if you succeed to run or not so we can
> >> internally track the problem and notify future users if a bug exist.
> >>
> >> Thanks,
> >> Filippo
> >>
> >> On Jan 26, 2015, at 8:16 AM, Hideaki Kuraishi
> >> <Hideaki.Kuraishi at uk.fujitsu.com> wrote:
> >>> Dear Filippo, Fabio
> >>>
> >>> Thank you for the comment.
> >>> Yes, core file is generated and will analyze this.
> >>>
> >>> So far I have tried your suggestions but situation didn’t change and
> >>> the same error occurred. Even when I used older versions of Intel
> >>> compiler, v13 and 14, the error was reproduced. I will try this on
> >>> more
> >> using another system this week.
> >>>
> >>> Best regards
> >>> Kai
> >>>
> >>>
> >>> From: pw_forum-bounces at pwscf.org
> [mailto:pw_forum-bounces at pwscf.org]
> >>> On Behalf Of Filippo Spiga
> >>> Sent: Sunday, January 25, 2015 11:07 PM
> >>> To: PWSCF Forum
> >>> Cc: <q-e-developers at qe-forge.org>
> >>> Subject: Re: [Pw_forum] Build QE for Intel Phi native run
> >>>
> >>> Dear Hideaki,
> >>> ,
> >>> Fabio Affinito can advise for the Intel Phi part however I do
> >>> suggest
> >> to run configure in this way:
> >>>
> >>> ./configure --enable-openmp --enable-parallel
> --with-scalapack=intel
> >>>
> >>> and only after edit the make.sys to eventually modify the libraries
> >> patch and names. The configure should be able to pickup MKL and
> >> ScaLAPACK if the environment is set correctly.
> >>>
> >>> The problem you see is additional to the one I noticed while ago
> >>> with
> >> the latest 15.x Intel compiler. If you can systematically reproduce
> >> this compiler seg fault then it is worth to file a bug report to
> >> Intel. It mention a core file... do you have one in the PW/src directory?
> >>>
> >>> F
> >>>
> >>> On Jan 23, 2015, at 3:59 PM, Hideaki Kuraishi
> >> <Hideaki.Kuraishi at uk.fujitsu.com> wrote:
> >>>
> >>> Hello,
> >>>
> >>> I have been trying to build QE 5.0.2 with Intel Compiler 15.0.0 and
> >> Intel MPI 5.0 to run in native mode. However, though configure is
> >> fine, it fails at make phase with segmentation fault error as follows.
> >>>
> >>> --------------------------------------
> >>>           :
> >>> mpiifort -openmp -o bands_FS.x bands_FS.o -L$MKLROOT/lib/mic
> >>> -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 -liopm5
> >>> $MKLROOT/lib/mic/libmkl_lapack95_lp64.a -liomp5
> >>> $HOME/mkl/lib/libfftw3xf_intel.a -L$MKLROOT/lib/mic
> -lmkl_intel_lp64
> >>> -liomp5
> >>>
> >>> /opt/intel/impi/5.0.0.028/intel64/bin/mpiifort: line 729:  5630
> >> Segmentation fault      (core dumped) $Show $FC $FCFLAGS
> >> "${allargs[@]}" $FCMODDIRS $FCINCDIRS -L${libdir}${MPILIBDIR}
> >> -L$libdir $rpath_opt $mpilibs $I_MPI_OTHERLIBS $LDFLAGS
> >> $MPI_OTHERLIBS
> >>
> >>>
> >>> ( cd ../../bin ; ln -fs ../PW/tools/bands_FS.x . )
> >>>
> >>> mpiifort -openmp -o kvecs_FS.x kvecs_FS.o -L$MKLROOT/lib/mic
> >>> -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 -liopm5
> >>> $MKLROOT/lib/mic/libmkl_lapack95_lp64.a -liomp5
> >>> $HOME/mkl/lib/libfftw3xf_intel.a -L$MKLROOT/lib/mic
> -lmkl_intel_lp64
> >>> -liomp5
> >>>
> >>> /opt/intel/impi/5.0.0.028/intel64/bin/mpiifort: line 729:  5656
> >> Segmentation fault      (core dumped) $Show $FC $FCFLAGS
> >> "${allargs[@]}" $FCMODDIRS $FCINCDIRS -L${libdir}${MPILIBDIR}
> >> -L$libdir $rpath_opt $mpilibs $I_MPI_OTHERLIBS $LDFLAGS
> >> $MPI_OTHERLIBS
> >>
> >>>
> >>> ( cd ../../bin ; ln -fs ../PW/tools/kvecs_FS.x . )
> >>>           :
> >>> --------------------------------------
> >>>
> >>> If someone has experienced the same error, could you please share
> >>> the
> >> information on how to solve this?
> >>>
> >>> My procedure is...
> >>>
> >>> (1) Configure
> >>> export CC=mpicc
> >>> export FC=mpiifort
> >>> export F90=$FC
> >>> export MPIF90=$FC
> >>> export AR="xiar"
> >>> export CFLAGS="-mmic -openmp"
> >>> export FFLAGS=$CFLAGS
> >>> export FCFLAGS=$CFLAGS
> >>> export BLAS_LIBS="-L$MKLROOT/lib/mic -lmkl_intel_lp64 -liomp5"
> >>> export LAPACK_LIBS="$MKLROOT/lib/mic/libmkl_lapack95_lp64.a
> -liomp5"
> >>> export SCALAPACK_LIBS="-L$MKLROOT/lib/mic -lmkl_scalapack_lp64
> >> -lmkl_blacs_intelmpi_lp64 -liopm5"
> >>>
> >>> $ ./configure --enable-openmp --enable-parallel
> >>>
> >>> (2) Build
> >>> $ make pw 2>&1 | tee LOG.make
> >>>
> >>> Best regards
> >>> Kai
> >>>
> >>>
> >>>
> >>>
> >>> _______________________________________________
> >>> Pw_forum mailing list
> >>> Pw_forum at pwscf.org
> >>> http://pwscf.org/mailman/listinfo/pw_forum
> >>>
> >>> --
> >>> Mr. Filippo SPIGA, M.Sc.
> >>> http://filippospiga.info ~ skype: filippo.spiga
> >>>
> >>> «Nobody will drive us out of Cantor's paradise.» ~ David Hilbert
> >>>
> >>> *****
> >>> Disclaimer: "Please note this message and any attachments are
> >> CONFIDENTIAL and may be privileged or otherwise protected from
> disclosure.
> >> The contents are not to be disclosed to anyone other than the addressee.
> >> Unauthorized recipients are requested to preserve this
> >> confidentiality and to advise the sender immediately of any error in
> transmission."
> >>>
> >>>
> >>>
> >>>
> >>> _______________________________________________
> >>> Pw_forum mailing list
> >>> Pw_forum at pwscf.org
> >>> http://pwscf.org/mailman/listinfo/pw_forum
> >>
> >> --
> >> Mr. Filippo SPIGA, M.Sc.
> >> http://filippospiga.info ~ skype: filippo.spiga
> >>
> >> «Nobody will drive us out of Cantor's paradise.» ~ David Hilbert
> >>
> >> *****
> >> Disclaimer: "Please note this message and any attachments are
> >> CONFIDENTIAL and may be privileged or otherwise protected from
> disclosure.
> >> The contents are not to be disclosed to anyone other than the addressee.
> >> Unauthorized recipients are requested to preserve this
> >> confidentiality and to advise the sender immediately of any error in
> transmission."
> >>
> >>
> >>
> >> _______________________________________________
> >> Pw_forum mailing list
> >> Pw_forum at pwscf.org
> >> http://pwscf.org/mailman/listinfo/pw_forum
> >
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://pwscf.org/mailman/listinfo/pw_forum
> 
> --
> Mr. Filippo SPIGA, M.Sc.
> http://filippospiga.info ~ skype: filippo.spiga
> 
> «Nobody will drive us out of Cantor's paradise.» ~ David Hilbert
> 
> *****
> Disclaimer: "Please note this message and any attachments are
> CONFIDENTIAL and may be privileged or otherwise protected from disclosure.
> The contents are not to be disclosed to anyone other than the addressee.
> Unauthorized recipients are requested to preserve this confidentiality
> and to advise the sender immediately of any error in transmission."
> 


______________________________________________________________________
                                        
 Fujitsu Laboratories of Europe Limited
 Hayes Park Central, Hayes End Road, Hayes, Middlesex, UB4 8FE
 Registered No. 4153469
 
 This e-mail and any attachments are for the sole use of addressee(s) and
 may contain information which is privileged and confidential. Unauthorised
 use or copying for disclosure is strictly prohibited. The fact that this
 e-mail has been scanned by Trendmicro Interscan does not guarantee that 
 it has not been intercepted or amended nor that it is virus-free.




More information about the users mailing list