Seg. fault. on yambo-4.2.1

[robertap]

Dear administrators,

I have compiled yambo-4.2.1 successfully (I guess) using the following options:

#!/bin/bash
make clean_all

source /applis/site/env.bash
module purge
module load intel-devel/14
module load netcdf/4.3.2_intel-14.0.2
module load fftw/3.3.4_intel-14.0.2

export FC=ifort
export MPIF77=mpif90
export F77=ifort
export CC=icc

./configure --enable-open-mp --with-netcdf-path=/applis/site/stow/intel_14.0.2/netcdf_4.3.2 --with-netcdff-pat\
h=/applis/site/stow/intel_14.0.2/netcdf-fortran_4.4.1 --with-netcdf-libdir=/applis/site/stow/intel_14.0.2/netcd\
f_4.3.2/lib --with-netcdf-includedir=/applis/site/stow/intel_14.0.2/netcdf_4.3.2 --with-netcdff-libdir=/applis\
/site/stow/intel_14.0.2/netcdf-fortran_4.4.1/lib --with-netcdff-includedir=/applis/site/stow/intel_14.0.2/netc\
df-fortran_4.4.1/include --with-blas-libs="-lmkl_intel_lp64 -lmkl_sequential -lmkl_core" --with-lapack-libs="-\
lmkl_intel_lp64 -lmkl_sequential -lmkl_core" --with-blacs-libs="-lblacs" --with-fft-path=/applis/site/stow/in\
tel_14.0.2/fftw_3.3.4

make yambo interfaces 1> make.log 2>make.err

--- Then I have successfully converted pwscf output ysing p2y and then sucessfully initialized yambo.

Then, I want to compute the exchange self energy for the hBN (example yambo webpage)
HF_and_locXC # [R XX] Hartree-Fock Self-energy and Vxc
SE_Threads=0 # [OPENMP/GW] Number of threads for self-energy
EXXRLvcs= 40 Ry # [XX] Exchange RL components
%QPkrange # [GW] QP generalized Kpoint/Band indices
7| 7| 8| 9|
%

And I get a segmentation fault error:

<---> [01] CPU structure, Files & I/O Directories
<---> CPU-Threads:1(CPU)-12(threads)
<---> [02] CORE Variables Setup
<---> [02.01] Unit cells
<---> [02.02] Symmetries
<---> [02.03] RL shells
<---> [02.04] K-grid lattice
<---> [02.05] Energies [ev] & Occupations
<---> [03] Transferred momenta grid
<01s> [04] Bare local and non-local Exchange-Correlation
<01s> [WF] Performing Wave-Functions I/O from ./SAVE
<01s> [FFT-HF/Rho] Mesh size: 12 12 32Segmentation fault (core dumped)

Attached the conf.log (and summary of compilation). Btw, I compiled successfully also using internal netcdf and internal fftw but eventually get the same error.

Also, on my local machine everything works fine, same hBN calculations, same input ...

Thanks for any hint you can provide...

Roberta

[Daniele Varsano]

Dear Roberta,
please add in your signature your complete affiliation: this is a rule of the forum. You can do it once for all by filling the signature field in your profile.

The segmentation fault can be due to the incompatibility of your internal linear algebra libraries (bias) with the rest of the code, e.g. different compilers, compiler versions etc.
You can try to recompile by adding in your configure line the --enable-int-linalg option.
It is recommendable to recompile from scratch so do a make clean_all before reconfiguring.

Best,
Daniele

[robertap]

Dear Daniele,
that actually solved the problem.
Thanks a lot
Roberta

[robertap]

Dear Daniele,

today I run into another problem with the same version of yambo but on a different machine.

This time, it is the p2y.sh script that doesn't work.

This is the error:

<---> DBs path set to .
<---> Index file set to data-file.xml
<---> Header/K-points/Energies... done
<---> Cell data... done
<---> Atomic data... done
<---> Symmetries...[SI no]......[-I no]...[TR yes]
<---> XC functional...Perdew, Burke & Ernzerhof(X)+Perdew, Burke & Ernzerhof(C)
<---> EXX fraction... 0.000000
<---> EXX screening... 0.000000
<---> K-points mesh... done
<---> RL vectors...
[ERROR]Error in qexml_read_planewaves fmt error ierr: 2

Please note that IT DOES work on a simple calculation like hBN. The system now is much bigger. Please note also that after solving the problem yesterday of blas (following your suggestion) I could run properly this system on that machine (wf. collect and force_symmorphic are set true).

Could the issue be that the blas libraries used to compile pwscf are external while for yambo I used the internal ones?

Attached the config.log (all internal libraries).

Thank you in advance for any suggestion you could provide.

Roberta

[Daniele Varsano]

Dear Roberta,
it does not look it is a problem of compilation.
Can you check if you have the file gvectors.dat in the .save directory of quantum espresso and gkvectors.dat in each Kpoints directory?

Could the issue be that the blas libraries used to compile pwscf are external while for yambo I used the internal ones?

This should not be a problem.

Other question, which version of QE are you using? Please note that in the last qe version (6.2) the data format has changed and it is needed to specify in the yambo configuration the option:
--with-p2y-version=qexsd
Anyway, I can see that p2y found the data-file.xml file so it does not seem to be the case, unless you used different version of qe in the same directory.

Just to be sure, the variable to be set in qe is wf_collect and not wf.collect but I presume that this is just a misspelling in your message.

Best,
Daniele

[robertap]

Dear Daniele,

thanks for the help.

I do have the gvectors.dat file. Since its binary I dont know if it's well written, though it looks it contains all sections (by comparing with other cases).

I only have Gamma since it is a big molecular crystal with no dispersion in the bands and I am using qe-6.1.

As you say, I misspelled the wf collect label in the message.

Please find attached the scf and nscf files.

Thanks again
Roberta

[Daniele Varsano]

Dear Roberta,

we will try to reproduce your error, can you please upload the pseudo you are using for your calculation?
anyway if I have well understood you were able to run p2y for this system on another machine right?

Daniele

[robertap]

Dear Daniele,
thank you.
Please find attached the pseudos. Yes, the same inputs work on my local server and on another supercomputer. That's why I thought it was a compilation issue.
Since it is a relatively big system I also tried reducing drastically the plane waves energy cutoff and the number of unoccupied band (considering that hBN works well with the same yambo and p2y executables), but the error on p2y persist.
Please let me know if you have any suggestion I could try.
Thanks
Roberta

[Daniele Varsano]

Dear Roberta,
in this case, I'm not sure I will be able to reproduce the error, I will try and let you know.

Daniele

[Daniele Varsano]

Dear Roberta,
I tried to reproduce the error but was not successful, I used QE6.1 and yambo4.2 and the p2y run went smooth.
As already asked, in your .save directory do you have the gvectors.dat and also the K00001/gkvectors.dat file?
Just to be sure that the QE run terminated correctly.
Best,
Daniele