error for run yambo

[damao4361556]

Dear Daniele,

I now come across with the error for built the BSE and then run it.

I run the yambo with either one cpu or many cpu cores, it has the error like:

Code: Select all

<---> [WARNING] Empty workload for CPU 1*** glibc detected *** yambo: double free or corruption (out): 0x0000000004fa50a0 ***
======= Backtrace: =========
/lib64/libc.so.6[0x318e275e66]
/lib64/libc.so.6[0x318e2789b3]
yambo[0x9172ac]
yambo[0x55c800]
yambo[0x55c1a0]
yambo[0x476f47]
yambo[0x409b33]
yambo[0x40525f]
/lib64/libc.so.6(__libc_start_main+0xfd)[0x318e21ed5d]
yambo[0x4041c9]
======= Memory map: ========

for one cpu, run in the terminal

or

Code: Select all

forrtl: severe (104): incorrect POSITION= specifier value for connected file, unit 11, file /home/weili/YAMBO-test/new/YAMBO_TUTORIALS/hBN/PWSCF/hBN.save/./r_optics_bse_bsk_em1s
Image              PC                Routine            Line        Source
yambo              0000000000978C7E  Unknown               Unknown  Unknown

for many cpu cores, with computing nodes

the input, report, log files are attached.

Could you please help me with that. Not sure if it is related to the memory issue, but I have use "ulimit -c unlimited", still not working.
Or for many cpu cores case, the output is CPU dependent, then something wrong happens?

thank you

[Daniele Varsano]

Dear Wei Li,
it is not easy to understand what is going on.

1) Please note you are not doing a BSE calculations as you have

Code: Select all

BSKmod= "Hartree"   

BSE kernel is built with

Code: Select all

BSKmod= "SEX"   

and relative variables.
This input is rather inconsistent as you ask for calculating the screening but then you do not use it.
Next:
2)

Code: Select all

% BSEBands
   1 | 100 |                   # [BSK] Bands range
%

usually, you need to include few bands around the gap. By using this setting you end up with an enormous matrix and relative memory problem.
3)

Code: Select all

NGsBlkXs= 1            RL

Here you are including the screening with only 1 G vectors, i.e. scalar, you are totally neglecting local field effects. In any case, finally, the screening is not used in an Hartree calculation.
4)
Please note you have previous calculations of the screening which are inconsistent with the present input, you can see the ERR in the report file after reading the ndb.em1s database. You can try to remove it and run yambo again, fixing first the input file.
5)
Your parallel yambo is not working, you should have just one report and one LOG directory. If you open any of the reports you will see CPU=1, so something went wrong in the parallel compilation of yambo.

Finally, I strongly suggest you follow a step by step tutorial for the BSE (http://www.yambo-code.org/wiki/index.ph ... step_guide).
In any case, it is not clear to me why the serial calculation stops, I should reproduce your error. Anyway I suggest first to try with a meaningful BSE input.

Best,
Daniele

[damao4361556]

Dear Daniele,

thanks again for your suggestions.

I will try the method you suggested to run Yambo with single cpu.
but, at the same time, I plan to install the YAMBO with mpirun functionality.

The is the command for configure,

Code: Select all

export LD_LIBRARY_PATH=/home/weili/soft/openmpi-1.8/lib:$LD_LIBRARY_PATH
./configure FC=gfortran F77=gfortran PFC=/home/weili/soft/openmpi-1.8/bin/mpif90 CC=gcc CPP=cpp  MPIF90=/home/weili/soft/openmpi-1.8/bin/mpif90  MPICC=/home/weili/soft/openmpi-1.8/bin/mpicc \
--enable-open-mpi \
--enable-momory-profile \
--with-netcdf-lib="-L/home/weili/soft/yambo/netcid-work/lib" --with-netcdf-include="/home/weili/soft/yambo/netcid-work/include" \
--with-p2y-version=5.1 \
--with-iotk-path=/home/weili/soft/espresso-5.1.1/espresso-5.1.1/bin/iotk

All with gfortran or gcc-mpif90 compiler.

Code: Select all

But get the error:
checking fortran 90 modules inclusion flag... -I
configure: testing C-preprocessor cpp -traditional
checking if C precompiler works on C source... yes
configure: testing FC-preprocessor gfortran -E -P -cpp
checking if FC precompiler works on FC source... no
configure: error: Found FC precompiler problems in processing FC source.

Do you have ways to fix it?
the config.log file is attached.

[Daniele Varsano]

Dear Wei Li,
you can try to set:

Code: Select all

CPP="gcc -E"

and use:

MPIFC instead of MPIF90

Daniele