Abnormal point in band structure calculation of Li

[pxlxingliang]

Describe the bug

In my band structure calculation of Li, there has an abnormal value appears near the gamma point. I have try to do the calculation by LCAO, the abnormal point is dispeared.

BANDS_1.dat

Li.zip

[dyzheng]

Can you try ks_solver cg for this case?

[pxlxingliang]

The results by CG looks fine.

[mohanchen]

It seems that it is caused by linear algebra solver, @Cstandardlib could you take a look at this issue?

[Cstandardlib]

I've tested this case on my machine and did not reproduce this result.
Test env:

ABACUS v3.9.0.2
Commit: 35448cbe7 (Mon Mar 31)
dav_subspace with both CPU and GPU

and also the given version ABACUS v3.9.0

ABACUS v3.9.0
Commit: 19a1b13a7 (Tue Mar 4 08:58:25 2025 +0800)
dav_subspace, CPU

Could you provide more information or reproduce it on another machine?
Note that a PR #6020 has been merged, which did some code changes about dav_subspace.

[Cstandardlib]

If you encounter this problem again, please provide more information and reopen the issue, thanks!