Large memory costs for LCAO calculations (Fe with 32 atoms)

[hongriTianqi]

Describe the bug

Job kill by signal 9 in two Fe systems Bohrium links: Fe32: https://bohrium.dp.tech/jobs/detail/11340467 Fe33: https://bohrium.dp.tech/jobs/detail/11340457

Expected behavior

no error

To Reproduce

test_tzdp.tar.gz

One Fe32 system, one Fe33 system.

Environment

No response

Additional Context

No response

Task list for Issue attackers (only for developers)

• [x] Verify the issue is not a duplicate.
• [ ] Describe the bug.
• [ ] Steps to reproduce.
• [ ] Expected behavior.
• [ ] Error message.
• [ ] Environment details.
• [ ] Additional context.
• [ ] Assign a priority level (low, medium, high, urgent).
• [ ] Assign the issue to a team member.
• [ ] Label the issue with relevant tags.
• [ ] Identify possible related issues.
• [ ] Create a unit test or automated test to reproduce the bug (if applicable).
• [ ] Fix the bug.
• [ ] Test the fix.
• [ ] Update documentation (if necessary).
• [ ] Close the issue and inform the reporter (if applicable).

[caic99]

Hi @hongriTianqi , Have you tried instances with larger memory?

[dyzheng]

I have tested this case without delta_spin, it costed 30.8 GB memory with MPI 8 processes. And it costed 51.1 GB memory with MPI 16 processes.