Dear Joydev,
i have got a warning message in the r_optics_chi_tddft file that rename the file name ns.kb_pp_pwscf in the SAVE directory for first calculation. After renaming the file name, i have also got a warning message that you are missing the non-locality of the pseudo potential. I don't understand the reason of warning message.
Inclusion of non-local part of psuedo in dipoles calculations is time consuming and not always needed. The code just warns you on that. If you do not want to include it you need to rename the ns.kb_pp_pwscf file. Up to you to judge if it is needed or not, and see how much is cumbersome to keep it.
One more problem, always the job exit with an error message.
\
Having a look at the Log file the code exit because not able to allocate the needed memory:
Code: Select all
<01m-10s> P0055: [M 4.108 Gb] Alloc WF ( 3.725)
It looks you have 4GB per code.
In order to overcome this problem you can:
1) Use lesse cores per node, e.g. 8 core instead of 16, you will have less cpus but more RAM for each.
2) Reduce a bit the cutoff for FFT of Wave functions, this is controlled by using the FFTGvecs variable.
3) Reduced the number of bands in BndsRnXd
For the case 2-3 of course you need to check convergences and control you do not loose much accuracy.
Best,
Daniele
