Describe the bug
I perform two scfs with the same input files for Ti3Nb and then compare their output "data-dHRx-sparse_SPIN0.csr". I find the tensors in these two files are different, and it seems unlikely to be caused by machine errors. ABACUS (3.9.0.2) was compiled through "toolchain_gnu.sh". Thank your!

Expected behavior
No response
To Reproduce
No response
Environment
No response
Additional Context
No response
Task list for Issue attackers (only for developers)
Describe the bug
I perform two scfs with the same input files for Ti3Nb and then compare their output "data-dHRx-sparse_SPIN0.csr". I find the tensors in these two files are different, and it seems unlikely to be caused by machine errors. ABACUS (3.9.0.2) was compiled through "toolchain_gnu.sh". Thank your!
Expected behavior
No response
To Reproduce
No response
Environment
No response
Additional Context
No response
Task list for Issue attackers (only for developers)