assert(psi_norm>0) error with ks_solver dav about abacus-develop

[dyzheng]

I think it is not a bug, often caused by accuracy loss for Schmidt orthogonalization method.

[WHUweiqingzhou]

@haozhihan could you have a look?

[haozhihan]

#3643 FYI

[WHUweiqingzhou]

So, you do not think it is a bug for old dav method?

[haozhihan]

I think, this should be a numerical problem for Schmidt orthogonalization, not a bug for old dav.

Currently, Schmidt orthogonalization is used in both the cg and dav methods of abacus, which leads to this issue.

[pxlxingliang]

More examples that suffer this error.
relaxation.zip

[WHUweiqingzhou]

@mohanchen and @dyzheng,
what is your opinion? Could we just leave this issue as a feature instead of a bug.

[haozhihan]

More examples that suffer this error. relaxation.zip

@pxlxingliang Can you provide related input files instead of just log and OUTPUT?

[pxlxingliang]

More examples that suffer this error. relaxation.zip

@pxlxingliang Can you provide related input files instead of just log and OUTPUT?

Please check the attached file of V inputs
v.zip