Issues with 2D Coulomb cutoff with parallelization

Running Abinit Ground state
1

Hello everyone!

I’m currently testing the icutcoul parameter using Abinit v10.0.3 in order to be able to use a 2D Coulomb interaction cutoff for my systems. While running the test example t132.abi, everything seems to work fine until I added parallelization.

In the sequential run, I get a total energy of -13.362331651 Ha.
In the parallelized run, I get a total energy of +2.8202648155 Ha.

Does anyone have an idea of what might be the problem? I am providing a folder with the input and output files for both runs.
BN_seq.abi (1.4 KB)
BN_seq.abo (25.8 KB)
BN_para.abi (1.5 KB)
BN_para.abo (25.4 KB)

Thank you for your time,

Tony Vuu

1 Like
2

I have done some testing with different FFT library and it seems that the parallelization of FFT is problematic when a Coulomb cutoff is used. Interestingly, this problem seems to be also present in v9.6.2 of Abinit. Here are my findings:

seq_vs_para_cutoff
seq_vs_para_cutoff1585×587 60 KB