Large memory required

[iduygnay]

Details

I just want to know how much memory is required for my task. I applied for 3400G memory but still oom. My calculation is using PBE functional, considering SOC, LCAO basis, 100 ecut, 16928 atoms for twisted structure. I know there are many different users who employ ABACUS to calculate twisted structures. I attached the settings of my task.

Task list for Issue attackers (only for developers)

• [ ] Reproduce the performance issue on a similar system or environment.
• [ ] Identify the specific section of the code causing the performance issue.
• [ ] Investigate the issue and determine the root cause.
• [ ] Research best practices and potential solutions for the identified performance issue.
• [ ] Implement the chosen solution to address the performance issue.
• [ ] Test the implemented solution to ensure it improves performance without introducing new issues.
• [ ] Optimize the solution if necessary, considering trade-offs between performance and other factors (e.g., code complexity, readability, maintainability).
• [ ] Review and incorporate any relevant feedback from users or developers.
• [ ] Merge the improved solution into the main codebase and notify the issue reporter.

[xdzhu]

I think it's too big for this kind of large system.

1. Reduce mpi process by mpirun -np 1, and enlarge the omp at the same time. 4mpi cost 4 times memory more than 1mpi.
2. Test 7au, 8au, 9au, 10au DZP .orb files on small systems first, check the convergence.
3. Reduce ecut. ecut test on small system is encouraged, you can refer to this notebook to test the reasonable ecut for your system https://bohrium.dp.tech/notebooks/55473361357
4. Test your memory cost of small systems of choosen .orb and ecut, and then you can estimate the cost of your real twisted systems.

Notice: Do not use gamma_only because there is no Hamitonian output in this mode.

[QuantumMisaka]

@iduygnay The newest version of ABACUS is recommended

[iduygnay]

I don't think so the newest version supports noncolinear SOC, Like it won't converge.

[LiuXiaohui123321]

@iduygnay Could you please recompile ABACUS with the option “-DEBUG_INFO=ON” and add “out_alllog 1” to the INPUT file? In that case it will give some useful debugging information.

[mohanchen]

Since the system is too large, we will first solve the memory issue and then see if it fixes the issue. I will close it, reopen if needed.