Yambo Community Forum

Thank you! This is just a test run so I will converge the results with respect to parameters such as response block size. QE is running on multiple nodes w/o a problem, but Yambo is not. I compiled both by using the same environment. Here are the modules I loaded.

module load gcc/10.3.0
module load lapack/3.8.0_gcc81
module load fftw3/3.3.5
module load intel/mpi/2021.3.0
module load intel/2021.3.0
module load intel/mkl/64/2020.0.166
module load scalapack/intelmpi/2.0.2

Burak

Dear Burak,

yambo is not aware to the number of nodes but only to the number of cpus so I do not know how to help you.
Pay attention to the fact you are loading mixed gnu and intel libraries, be sure you using mkl (intel) and not lapack (gcc) anyway this is not the reason of your problem, and probably now you are using internal linear algebra.
Can you check if all dipole files have been written in the SAVE directory?
Can you also try to parallalelize over k points only?

Best,
Daniele

Dear Daniele,

I solved the problem by running Yambo inside scratch directory.

Best,
Burak