NumPy multithreading in Baobab

Dear Adrian,

Is there a particular reason to use an old NumPy/Python version?

Are you using the same NumPy version on Baobab and on your pc?

To know more about the version of NumPy on Baobab you are using:

[sagon@login2 ~] $ module load GCC/7.3.0-2.30 OpenMPI/3.1.1 Python/2.7.15
[sagon@login2 ~] $ python -c "import numpy; numpy.show_config()"
lapack_opt_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/opt/ebsofts/Compiler/GCC/7.3.0-2.30/OpenBLAS/0.3.1/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
blas_opt_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/opt/ebsofts/Compiler/GCC/7.3.0-2.30/OpenBLAS/0.3.1/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
openblas_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/opt/ebsofts/Compiler/GCC/7.3.0-2.30/OpenBLAS/0.3.1/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
blis_info:
  NOT AVAILABLE
openblas_lapack_info:
    libraries = ['openblas', 'openblas']
    library_dirs = ['/opt/ebsofts/Compiler/GCC/7.3.0-2.30/OpenBLAS/0.3.1/lib']
    define_macros = [('HAVE_CBLAS', None)]
    language = c
lapack_mkl_info:
  NOT AVAILABLE
blas_mkl_info:
  NOT AVAILABLE
[sagon@login2 ~]

So indeed it seems you can control the parallelism with OPENBLAS_NUM_THREADS

If you set OPENBLAS_NUM_THREADS=1, do you have an expected result?

About this line in your script:

#SBATCH --job-name=asg-$filename

is this realy working? how do get the $filename value, is this a variable set outside of the script?