Transiesta stuck at Title: 'Generated GF file'

Asked by Tony on 2017-03-02

Dear all,
I have a problem when I implement calculations on the cluster at the university computing centre.

When I calculate transiesta, it is always stuck at the step :
 Reading GF file, with title:
   /gpfs/workdir/tony/Siesta/Electrode_2cells/SR.NA5_1vac/AGNR5.TSGFL
 Title: 'Generated GF file'

 Reading GF file, with title:
   /gpfs/workdir/tony/Siesta/Electrode_2cells/SR.NA5_1vac/AGNR5.TSGFR
 Title: 'Generated GF file'

and calculations stopped.

Further checks, I found it shows error:
forrtl: severe (67): input statement requires too much data, unit 15, file /gpfs/workdir/tony/Siesta/Electrode_2cells/SR.NA5_1vac/AGNR5.TSGFR

I checked the size of the file AGNR5.TSGFR and saw it is about 4600 KiB.

I tried many times by setting different number of nodes, cpus, memory but did not succeed.

The siesta version on this cluster is siesta-4.0--500 parallel version and it was compiled with ifort -g.

It is worth to note that when I calculated for a small system with the size of the file AGNR5.TSGFR (AGNR5.TSGFL) is about 1500 KiB, I did not encounter such problem.

Could you please tell me any reason of this error?

Thank you in advance.

best regards,
Tony

Question information

Language:
English Edit question
Status:
Solved
For:
Siesta Edit question
Assignee:
No assignee Edit question
Solved by:
Nick Papior
Solved:
2017-03-24
Last query:
2017-03-24
Last reply:
2017-03-14
Nick Papior (nickpapior) said : #1

Have you tried to delete the TSGF* files and restart the calculation?

Tony (tony2017) said : #2

It does not help, Nick. I already tried, deleted the TSGF* or all files and re-calculate, problems remain the same. I think it is not because of parallel run, I tried just with 1 cpu setup, I still got stuck at this problem.

Nick Papior (nickpapior) said : #3

Your file sizes seem extremely small. Could you attach your fdf file?

Secondly, you may compile transiesta with backtrace mode (see the bug-report advice) which will give better diagnostics. Could you attach the output for such a run?

Tony (tony2017) said : #5

Dear Nick,

Please see all fdf via the link bellow: (I do not see where I can attach files).

https://www.dropbox.com/s/xo5bftza0q3ic36/Electrode_2cells.tar?dl=0

I have no right to compile transiesta with backtrace mode because it was compiled on the computing centre. But I will send a request to administrators.

Thank you.

Best Nick Papior (nickpapior) said : #6

1. Your electrode setup is not a AGNR, your 1st lattice vector is too small, so your system has couplings along that vacuum direction. I suspect this is _not_ what you want?
2. Your scattering region connects directly to the electrodes, this is not adviced, you need some additional layers to screen the bulk electrode from the scattering region. There are many articles discussion this, see e.g. http://www.sciencedirect.com/science/article/pii/S001046551630306X.
3. I can run your example without problems using siesta 4.0. Please contact your system administrator and ensure that your filesystem works as expected.

Tony (tony2017) said : #7

Hi Nick,
Thank you for your information.
1. My electrode is really an AGNR.
The structure is periodic along z direction, not x. I checked by viewing structure with VNL software and exported input file for Siesta run. It looked correctly. So, I do not know why you said the system has couplings along vacuum directions.
I have also learn from the tutorial for zigzag ribbons: http://dipc.ehu.es/frederiksen/tstutorial/index.php/Zig-zag_graphene_nanoribbon, so my setup should be ok.

2. Thank you for giving me this hint/advice.

3. Thank you for this confirmation. It is thus my setup has no errors. I will contact the computing centre to debug this problem.

Regards,
Tony

Nick Papior (nickpapior) said : #8

Regards to 1. , siesta prints out the super-cell information where there are couplings. And it is definitely coupling along the first cell vector. You should add more vacuum. (an geometry viewer does not give you this information).

Tony (tony2017) said : #9

Thank you, Nick. Now I think I understand what you mean.

Tony (tony2017) said : #10

Hi Nick, I would like to ask how did you compile Siesta, transiesta and TBtrans on your cluster? It might help to debug this issue on our cluster. Here we compiled with Intel compiler ifort and INTEL MPI library with mpiifort.

Nick Papior (nickpapior) said : #11

I have successfully compiled with nearly all variations of the intel compiler and the OpenMPI. However, I know that others have used Intel MPI as well, so I think there are no problems.
Here is my arch.make:

<<<<
PP = cpp -E -P -C -nostdinc
LIBS=$(ADDLIB) -lmkl_scalapack_lp64 -lmkl_blacs_openmpi_lp64 -lmkl_lapack95_lp64 -lmkl_blas95_lp64
LIBS+= -lmkl_intel_lp64 -lmkl_core -lmkl_sequential

LDFLAGS=-L/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//mkl/lib/intel64 -Wl,-rpath=/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//mkl/lib/intel64 -L/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//compiler/lib/intel64 -L/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//mkl/lib/intel64 -L/zdata/groups/common/nicpa/2016-oct/generic/numactl/2.0.11/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/zlib/1.2.8/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/libxml2/2.9.4/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hwloc/1.11.4/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/openmpi/2.0.1/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hdf5/1.8.17/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/pnetcdf/1.7.0/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/netcdf/4.4.1/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/flook/0.7.0/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/parmetis/4.0.3/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/mumps/5.0.2/intel-17.0.0/lib -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/fftw/3.3.5/intel-17.0.0/lib -Wl,-rpath=/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//compiler/lib/intel64 -Wl,-rpath=/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//mkl/lib/intel64 -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/generic/numactl/2.0.11/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/zlib/1.2.8/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/libxml2/2.9.4/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hwloc/1.11.4/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/openmpi/2.0.1/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hdf5/1.8.17/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/pnetcdf/1.7.0/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/netcdf/4.4.1/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/flook/0.7.0/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/parmetis/4.0.3/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/mumps/5.0.2/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/fftw/3.3.5/intel-17.0.0/lib
FPPFLAGS= -I/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//compiler/include -I/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//mkl/include -I/zdata/groups/common/nicpa/2016-oct/generic/numactl/2.0.11/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/zlib/1.2.8/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/libxml2/2.9.4/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hwloc/1.11.4/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/openmpi/2.0.1/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hdf5/1.8.17/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/pnetcdf/1.7.0/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/netcdf/4.4.1/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/flook/0.7.0/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/parmetis/4.0.3/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/mumps/5.0.2/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/fftw/3.3.5/intel-17.0.0/include

.SUFFIXES:
.SUFFIXES: .f .F .o .a .f90 .c .F90
SIESTA_ARCH=x86_64-linux-n-62-18-17

FPP=mpif90
FPP_OUTPUT=
CC=icc
FC=mpif90
FC_SERIAL=ifort
AR=xiar
RANLIB=ranlib
SYS=nag
SP_KIND=4
DP_KIND=8
KINDS=$(SP_KIND) $(DP_KIND)

FFLAGS=-m64 -fPIC -O2 -xHost -prec-div -prec-sqrt -fp-model source -qopt-prefetch
FFLAGS += #OMPPLACEHOLDER
FPPFLAGS += -DMPI -DFC_HAVE_FLUSH -DFC_HAVE_ABORT -DCDF -DCDF4

ARFLAGS_EXTRA=

NETCDF_INCFLAGS= -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/netcdf/4.4.1/intel-17.0.0/include
NETCDF_LIBS= -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/netcdf/4.4.1/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/netcdf/4.4.1/intel-17.0.0/lib
ADDLIB=-lnetcdff -lnetcdf -lpnetcdf -lhdf5_hl -lhdf5 -lz
ADDLIB += #OMPPLACEHOLDER
INCFLAGS = -I/opt/intel/2017/compilers_and_libraries_2017.0.098/linux//compiler/include -I/zdata/groups/common/nicpa/2016-oct/generic/numactl/2.0.11/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/zlib/1.2.8/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/libxml2/2.9.4/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hwloc/1.11.4/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/openmpi/2.0.1/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/hdf5/1.8.17/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/pnetcdf/1.7.0/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/netcdf/4.4.1/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/flook/0.7.0/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/parmetis/4.0.3/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/mumps/5.0.2/intel-17.0.0/include -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/fftw/3.3.5/intel-17.0.0/include

MPI_INTERFACE=libmpi_f90.a
MPI_INCLUDE=.

FPPFLAGS += -DSIESTA__METIS -DSIESTA__MUMPS -DTS_NOCHECKS
ADDLIB += -lzmumps -lmumps_common -lpord -lparmetis -lmetis
METIS_LIB = -lmetis
FFTW_PATH = /zdata/groups/common/nicpa/2016-oct/XeonE5-2665/fftw/3.3.5/intel-17.0.0
FFTW_INCFLAGS = -I$(FFTW_PATH)/include
FFTW_LIBS = -L$(FFTW_PATH)/lib -lfftw3 $(METIS_LIB)
LIBS += $(METIS_LIB)
FPPFLAGS += -DNCDF -DNCDF_4 -DNCDF_PARALLEL
COMP_LIBS += libncdf.a libfdict.a
FPPFLAGS += -DSIESTA__FLOOK
FLOOK_LIB = -L/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/flook/0.7.0/intel-17.0.0/lib -Wl,-rpath=/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/flook/0.7.0/intel-17.0.0/lib -lflookall -ldl
INCFLAGS += -I/zdata/groups/common/nicpa/2016-oct/XeonE5-2665/flook/0.7.0/intel-17.0.0/include
LIBS += $(FLOOK_LIB)

.F.o:
 $(FC) -c $(FFLAGS) $(INCFLAGS) $(FPPFLAGS) $<
.F90.o:
 $(FC) -c $(FFLAGS) $(INCFLAGS) $(FPPFLAGS) $<
.f.o:
 $(FC) -c $(FFLAGS) $(INCFLAGS) $<
.c.o:
 $(CC) -c $(CFLAGS) $(INCFLAGS) $(FPPFLAGS) $<
.f90.o:
 $(FC) -c $(FFLAGS) $(INCFLAGS) $<

atom.o: atom.F
 $(FC) -c -O1 $(INCFLAGS) $(FPPFLAGS) $<

m_new_dm.o: m_new_dm.F90
 $(FC) -c -O2 $(INCFLAGS) $(FPPFLAGS) $<

<<<

Tony (tony2017) said : #12

Thank you, Nick. I will check with the administrators and will inform you.

Tony (tony2017) said : #13

Thanks Nick Papior, that solved my question.

Tony (tony2017) said : #14

Dear Nick,

It is indeed due to the arch.make file. After re-compiling siesta on the cluster, it now works as expected.

Thank you very much for your help.

Regards,
Tony