get_s producing super large overlap matrix failed

[JTaozhang]

Describe the bug

Hi there,

I have a system which have around 12w atoms. I just want to produce the overlap matrix. It works initially. and the overlap matrix is sucessfully created. But when I open the overlap lap matrix, I found the total number of effective matrix element is a negative number for R=[0,0,0] matrix, according to the format of overlap matrix. I think it is due to the overflow of integer number. moreover, I found some elements can't be reached, for example, in R[0,0,0], the matrix elements of [631800:631813,0:13] can not be found.

could you help me to check this problem?

you can barely take out the R[0,0,0] matrix and check those matrix elements, I mentioned here.

Best,
Tao

Expected behavior

abacus produce complete overlap matrix

large_overlap.zip

To Reproduce

No response

Environment

No response

Additional Context

No response

Task list for Issue attackers (only for developers)

• 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).

[mohanchen]

Thanks, we will check the issue.

[JTaozhang]

Hi,

Sorry to bother again.

I have tried the latest vesion of abacus. it indeed has fixed some problem. But it still contains the overflow of integer number in the in row_pointer of csr_matrix.

Here, I still take the sparse matrix of R[0,0,0] as an example, which I have said early message. I have token out the row_pointer of this sparse matrix, which has attached in this message. you can find some negative integer number, which is not reasonable.

Could you check the code again to fix these problem?

Best,
Tao

indptr.zip

[JTaozhang]

Moreover, the value of row_pointer should satisfy the condition of monotonic increase. But here it is not.

[ErjieWu]

@JTaozhang I fixed some missing points from last time. You can try #6803 version to see if it can output normally.

[JTaozhang]

Hi There,

thanks all developer's help a lot. It works now.

Best,
Tao