How to disable HDF5 in yambo5 configuration

[jasonhan0710]

Dear Developers,

I have calculate the quasi particle energy with yambo-5.0. However, the report file "o-GW.qp" gives "NaN" in the quasiparticle energy.

Code: Select all

#    K-point            Band               Eo [eV]            E-Eo [eV]          Sc|Eo [eV]          
#
        1                  30                -14.31931                   NaN                NaN
        1                  31                -14.30590                   NaN                NaN
        1                   32               -14.30590                   NaN                NaN

When I use the same input file with yambo-4.3, it will give the quasi-particle energy. The report files show that

Code: Select all

#    K-point      Band         Eo           E-Eo         Sc|Eo
     1.00000     30.00000    -14.31931      2.39237      6.93869
     1.00000     31.00000    -14.30590      2.42014      6.91741
     1.00000     32.00000    -14.30590      2.42031      6.91741

The difference between two version of yambo is "hdf5_io".

Code: Select all

   Version 5.0.3 Revision 19584 Hash 1cce2ba63  
   Branch is
                 MPI+OpenMP+SLK+HDF5_IO Build 

and

Code: Select all

  GPL Version 4.5.3 Revision 9122. (Based on r.17117 h.03e635bbc)       
                      MPI+SLK+OpenMP Build

Would you please tell me how to disable hdf5 in yambo-5.0?

Best,
Jason

[Daniele Varsano]

Dear Jason,
it is not possible to disable hdf5 as they are needed to compile recent version of netcdf.
Anyway, the problem of the NaN you obtain is unlikely due to the hdf5 library.
You have already NaN in the Hartree-Fock part or in the correlation (you can spot it in the report).
Can you check if you have NaN in the dipole or screening database? You can read them by using the ncdump command.

Best,
Daniele

[jasonhan0710]

Dear Daniele,

The report file gives the results in the Hartree-Fock part. Only GW process reports NaN. Here attached the report files with both yambo4 and yambo5. Would you please help me check it?

Best,
Jason

[Daniele Varsano]

Dear Jason,

can you check if NaN appear in the ndb.dipoles*, ndb.pp* or none/both?

Best,
Daniele

[jasonhan0710]

Dear Daniele,

Both ndb.dipoles*, ndb.pp* show NaNf. For example, the data part of ndb.dipoles_fragment_1 gives that

Code: Select all

data:

 DIP_iR_k_0001_spin_0001 =
  NaNf, NaNf,
  -2.367106e+36, NaNf,
  -4.005494e+34, NaNf,
  NaNf, -1.345728e+34,
  -2.367106e+36, NaNf,
  9.794995e+37, NaNf,

and ndb.pp_fragment_1 shows

Code: Select all

data:

 FREQ_PARS_sec_iq1 = 0.9999999 ;

 FREQ_sec_iq1 =
  1.192093e-07, 0,
  0, 0.9999999 ;

 X_Q_1 =
  NaNf, NaNf,
  NaNf, NaNf,
  NaNf, NaNf,
  NaNf, NaNf,
  NaNf, NaNf,

Would you please tell me how to fix it in yambo-5.0? Thank you!

Best,
Jason

[Daniele Varsano]

Dear Jason,

it is not easy to spot the problem.
Some tests that can be useful:

1) Try to see if you have problems also in running a very simple test e.g. a GW calculation from the tutorial:
http://www.yambo-code.org/wiki/index.ph ... 21_version

the database can be downloaded here:
http://www.yambo-code.org/wiki/index.ph ... _CECAM2021

2) If it is the case, can you try to recompile the code using in the configure the command line:
--enable-int-linalg

Best,
Daniele

[jasonhan0710]

Dear Daniele,

Thanks for your reply.

I have did the two test

1) Try to see if you have problems also in running a very simple test e.g. a GW calculation from the tutorial:
http://www.yambo-code.org/wiki/index.ph ... 21_version

The sample test gives the NaNf in ndb.pp* files.

the database can be downloaded here:
http://www.yambo-code.org/wiki/index.ph ... _CECAM2021

The downloaded database also gives NaNf.

2) If it is the case, can you try to recompile the code using in the configure the command line:
--enable-int-linalg

The recompilation did not solve the problem, either.

Are there any other situations may cause this problem?

Best,
Jason

[Daniele Varsano]

Dear Jason,
can you post your config.log file?
Let's see if we can spot something that went wrong during the compilation.

Best,
Daniele

[jasonhan0710]

Dear Daniele,

Please check the config.log file as attached.

Best,
Jason

[Daniele Varsano]

Dear Jason,

a possibility of the problem could be the lack of the include directory of the FFT libraries.
Can you try to add

--with-fft-includedir="${MKLROOT}/include" \
--with-fft-libs="-L${MKLROOT}/lib/intel64 -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -liomp5 -lpthread -lm -ldl" \

It is possible you need to define the $MKLROOT that should be:
/data/home/djy3/software/intel2019/compilers_and_libraries_2019.5.281/linux/mkl

you can also remove the --enable-int-linalg

If it does not work I hope someone else can provide you some advise.

Best,
Daniele