Bug: `init_wfc nao` doesn't support gpu

[dzzz2001]

Describe the bug

When I run the test case “tests/integrate/102_pw_pint_uks” using the command OMP_NUM_THREADS=1 mpirun -n 1 abacus, I noticed that if I set both init_wfc nao and device gpu, the program crashes. To further investigate, I tested it again using gdb, and the output is as follows:
I'm not sure if this is expected behavior. Is init_wfc nao not supported on GPU, or is there a bug?

Expected behavior

No response

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]

It is a bug.

[kirk0830]

Hi, this is because the psi_initializer does not support GPU yet. There is a data sending from CPU to GPU procedure lacking, so when GPU call the function like diag_subspace_init() will refer to a null memory space.

The function in your screenshot, the proj_ao_onkG, will calculate and assign values on a memory space managed by a smart pointer owned by psi_initializer instance. In ESolver, it will share its access out, to do the wavefunction initialization.

I will find some time to support GPU, during which I will discuss with GPU-programming developers, find the best time to synchronize the memory.

Thanks for your feedback.