void hsolver::Diago_DavSubspace<T, Device>::cal_grad error

[OutisLi]

Describe the bug

When I run abacus on Bohrium, using image “registry.dp.tech/dptech/abacus-stable:LTSv3.10” with command “OMP_NUM_THREADS=2 mpirun -np 48 abacus > abacus.log 2>&1” on machine “c96_m384_cpu”, I encountered this error.

Expected behavior

No response

To Reproduce

No response

Environment

No response

Additional Context

Related INPUT files are provided in the zip file and output was stored in abacus.log in the zip file.

SiC_3C_3x3x3_C_Si.zip

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).

[Cstandardlib]

I run this example on c52_m384_cpu machine with
OMP_NUM_THREADS=1 mpirun -np 48 abacus | tee abacus.log and it worked well.
While OMP_NUM_THREADS=2 mpirun -np 26 abacus | tee abacus2.log will report the same error after start of STEP OF RELAXATION : 6
seems to be a parallel problem.

[OutisLi]

I run this example on c52_m384_cpu machine with
OMP_NUM_THREADS=1 mpirun -np 48 abacus | tee abacus.log and it worked well.
While OMP_NUM_THREADS=2 mpirun -np 26 abacus | tee abacus2.log will report the same error after start of STEP OF RELAXATION : 6
seems to be a parallel problem.

Emmm，but as you can see in another issue, I run into the same issue using 4 V100 GPU without OMP but mpi.

However, OMP is vital since for cpus with hyperthread, by setting OMP to 2 will accelerate the process. Meanwhile, I have some server troubling with mpi, only Openmp can be used.

[OutisLi]

I run this example on c52_m384_cpu machine with OMP_NUM_THREADS=1 mpirun -np 48 abacus | tee abacus.log and it worked well. While OMP_NUM_THREADS=2 mpirun -np 26 abacus | tee abacus2.log will report the same error after start of STEP OF RELAXATION : 6 seems to be a parallel problem.

Is there any progress on this issue?

[OutisLi]

Can this bug fix be merged to the LTS branch?

[Cstandardlib]

Seems that the PR that introduces this bug(#6428) has not been merged to LTS?

[OutisLi]

Seems that the PR that introduces this bug(#6428) has not been merged to LTS?

But this bug was found using LTS image