p2y stops with qexml_read_cell IOTK error

[huangj3]

Dear all,

I downloaded and compiled Quantum Espresso 5.0 together with tddfpt trying to calculate UV-vis spectroscopy for an extended system. Tests indicated that everything is fine but tddfpt only works for Gamma point at this moment. Therefore I had to resort to Yambo-3.2.5 . Both QE and Yambo are the the latest version. However, I had a problem with p2y, which is described in detail as follows.

Compliers used are composerxe/2011.5.220 and mpi/openmpi-1.4.5-intel. The architecture is x86_64. Compilation aimed at a parallel build. I tried two approaches: (1) compile Yambo together with QE by "$make pw yambo"; (2) complie QE first (tests indicate it was compiled correctly) and then configure and make Yambo afterwards. It seems to make no difference in terms of the problem of p2y. Taking the second approach as an example, the configuration was done by: $ ./configure --with-blas="-L/opt/intel/mkl/10.0.2.018/lib/em64t -lmkl_intel_lp64 -lmkl_sequential -lmkl_core" --with-netcdf-include=/usr/include --with-netcdf-lib=/usr/lib64 --enable-netcdf-hdf5 --with-iotk=/home/hzu/espresso/iotk --with-p2y=4.0 PFC=mpif90 FC=mpif90 FCFLAGS="-O2 -assume byterecl -g -traceback -par-report0 -vec-report0 -nomodule -fpp -D__INTEL -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK -I../include -nofor_main". And make was done by: $make all.

After an scf run and an nscf run with wave functions collected, in the prefix.save, by typing $p2y, p2y stopped with the following message:
<---> P(W) 2 Y(ambo) Ver. 4.0
<---> DBs path set to .
<---> Index file set to data-file.xml
<---> Header/K-points/Energies...done
<---> Cell data...

[ERROR] STOP signal received while in :
[ERROR]Error in qexml_read_cell IOTK error ierr: 1

I browsed different forums and saw some problems with p2y but those problems came after the cell data stage. I looked at the logfile for make and didn't see anything wrong. The logfile for configure shows the following:

#
# [VER] 3.2.5 r.1015
#
# [SYS] linux@x86_64
# [SRC] /home/hzu/espresso/yambo-3.2.5
# [BIN] /home/hzu/espresso/yambo-3.2.5/bin
# [FFT] Goedecker Fast Fourier transform with 0 cache
#
# [ ] Double precision
# [X] Redundant compilation
# [X] MPI
# [X] PW (4.0) support
# [ ] ETSF I/O support
# [ ] SCALAPACK
# [ ] LibXC
# [ ] NETCDF/HDF5/Large Files
# [XX ] Built-in BLAS/LAPACK/LOCAL
#
# [ CPP ] icc -E -ansi
# [ C ] icc -g -O2 -D_C_US -D_FORTRAN_US
# [MPICC] mpicc -g -O2 -D_C_US -D_FORTRAN_US
# [ F90 ] mpif90 -O2 -assume byterecl -g -traceback -par-report0 -vec-report0 -nomodule -fpp -D__INTEL -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK -I../include -nofor_main
# [MPIF ] mpif90 -O2 -assume byterecl -g -traceback -par-report0 -vec-report0 -nomodule -fpp -D__INTEL -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK -I../include -nofor_main
# [ F77 ] mpif90 -O2 -assume byterecl -g -traceback -par-report0 -vec-report0 -nomodule -fpp -D__INTEL -D__FFTW3 -D__MPI -D__PARA -D__SCALAPACK -I../include -nofor_main
# [Cmain]
# [NoOpt] -O0
#
# [ MAKE ] make
# [EDITOR] vim
#

It can be seen that although NETCDF include and lib paths were specified, somehow it didn't find the files. But I don't think that's the reason. In the config.log file, there are a few warning and error messages associated with conftest. Those may not be the reason either. So I don't know what might be wrong. I'd greatly appreciate your help by examining the attached files including config.log, configure.logfile, make.logfile, silicon.scf.in, silicon.nscf.in, and data-file.xml.

Best regards,

Jingsong Huang
Computer Sciences and Mathematics
Center for Nanophase Materials Sciences
Oak Ridge National Laboratory
Oak Ridge, TN 38831, USA

[Stik Azzi]

Dear Jingsong,
I think the problem is a stupid one (not yours!) related to an unclosed tag in the XML file.
From your data-file.xml, I see you are using Si.vbc.UPF, and if I'm right, it should have the line
Author: Von Barth-Car (<1984)
This "<" confuses the parser. Just remove it from the UPF file (its just a comment anyway).
Let me know if it doesn't work....and sorry that you lost time over this. You are not the first to have this problem - I should really put in some check for it into the code.
Regards,
Conor

[huangj3]

Dear Conor,

Thank you very much for your quick response and your answer . I looked at the UPF file Si.pz-vbd.UPF but did not see the line of "Von Barth-Car (<1984)" . Please see the attached Si.pz-vbd.UPF file in the tarball. I also looked at the data-file.xml but it seems that each tag is closed, at least in the <CELL> ...</CELL> section (note that this is the part that p2y stopped). Thank you again for your help .

Best regards,

Jingsong
Computer Sciences and Mathematics
Center for Nanophase Materials Sciences
Oak Ridge National Laboratory
Oak Ridge, TN 38831, USA

[Stik Azzi]

Dear Jingsong,
Ha - I see there's another version of that file floating around now with "before 1984". Good.
So the problem is the usual one when new versions of Q-E are released - in this case in version 5.0 there seems to be a small change in the writing of the data, in particular, the writing of the tag CELL_SYMMETRY is now missing from Modules/xml_io_bse.f90, while p2y/qexml still looks for it.
So I should really update p2y for this case (users tend to jump to new versions quicker than developers!).

In the short term, I suggest you one of the following:
(1) Comment out the lines 1592-1593 of interfaces/p2y/qexml_v4.0.F
CALL iotk_scan_dat( iunit, "CELL_SYMMETRY", symm_type_, IERR=ierr )
IF ( ierr /= 0 ) RETURN
(assuming there are not more problems with 5.0)
(2) Use Q-E 4.3.2 (a step backwards)
(3) Insert a dummy tag like
<CELL_SYMMETRY type="character" size="1" len="1">
c
</CELL_SYMMETRY>
into your data-file.xml file after BRAVAIS_LATTICE (tedious)

Any of these will work....note that you get another error from using non-symmorphic symmetry operators, so you shoudl set force_symmorphic=.true. in PWSCF.

Let us know if you have any more problems.
Regards,
Conor

[huangj3]

Dear Conor,

There is no way that users can jump to new versions quicker than developers

Thank you very much. It passed p2y and yamobo -D -N test. I will try to get the optical properties next.

One more quick question: I wasn't able to link it to NETCDF/HDF5, although I did specify the library and include paths. I will try to install NETCDF/HDF5 myself. But at the moment, the only thing affected is performance, right?

Thanks a lot and have a great weekend.

Best regards,
Jingsong Huang
Computer Sciences and Mathematics
Center for Nanophase Materials Sciences
Oak Ridge National Laboratory
Oak Ridge, TN 38831, USA

[Daniele Varsano]

Dear Jingsong,
the problem could rely that netcdf and yambo are not compiled with the same compiler, so if you try to compile then by yourself
it could be the solution.

One more quick question: I wasn't able to link it to NETCDF/HDF5, although I did specify the library and include paths. I will try to install NETCDF/HDF5 myself. But at the moment, the only thing affected is performance, right?

Yes, using or not using the netcdf affect the I/O, i.e. the format of the databases generated and read from yambo. The great advantage
of using nectdf is that the produced database are machine independent.

Cheers,

Daniele

[huangj3]

Dear Daniele and Conor,

As described in my earlier posts, without linking with netcdf, yambo can be installed successfully. But if I tried to link yambo with netcdf and hdf5 (both were installed by myself), I could not get yambo installed correctly. Compliers used are composerxe/2011.5.220 and mpi/openmpi-1.4.5-intel. The architecture is x86_64. Compilation aimed at a parallel build.

For hdf5-1.8.8, I performed ./configure --prefix=/home/hzu/local. For netcdf-4.2, I did ./configure CPPFLAGS=-I/home/hzu/local/include LDFLAGS=-L/home/hzu/local/lib --prefix=/home/hzu/local. It appears that both netcdf and hdf5 were installed correctly.

For yambo, configuation was done by ./configure --with-netcdf-include=/home/hzu/local/include --with-netcdf-lib=/home/hzu/local/lib --enable-netcdf-hdf5=yes --with-iotk=/home/hzu/espresso/iotk --with-p2y=4.0. However, even though --enable_netcdf_hdf5 was set to yes, it seems that hdf5 was not correctly linked as can be seen from # [X ] NETCDF/HDF5/Large Files. I also tried without enabling hdf5 but there were still a couple of errors during make, the same as the former case with hdf5 enabled.

Could you please have a look at my tarball which includes the configure logfiles and make logfiles for netcdf, hdf5, and yambo?

Thank you so much for your help.

Best regards,

Jingsong Huang
Computer Sciences and Mathematics
Center for Nanophase Materials Sciences
Oak Ridge National Laboratory
Oak Ridge, TN 38831, USA