abacus crushed with spin+HSE calculation

[zhang-changwei]

Describe the bug

LiF 3x3x3 supercell. I'm running cell optimiztion. When I open nspin=2 option, abacus crushs at the HSE update step. Part of the output is showed as follows:

......
 Updating EXX and rerun SCF 
 GE1    3.18e-14  6.39e-08  -9.155630e+04  0.000000e+00   4.950e-05  3.946e+01  
 GE2    5.12e-14  5.21e-08  -9.155630e+04  -1.743483e-05  2.838e-05  3.796e+01  
 GE3    4.03e-14  4.86e-08  -9.155630e+04  -4.485836e-06  2.748e-06  3.800e+01  
 GE4    1.42e-14  5.57e-08  -9.155630e+04  4.927450e-08   3.775e-08  3.834e+01
Assertion failed in file ../../src/mpid/ch4/shm/posix/eager/include/intel_transport_recv.h at line 1013: cma_read_nbytes == size
/public1/soft/oneAPI/2022.1/mpi/2021.5.0/lib/release/libmpi.so.12(MPL_backtrace_show+0x1c) [0x2b7b4fab4bcc]
/public1/soft/oneAPI/2022.1/mpi/2021.5.0/lib/release/libmpi.so.12(MPIR_Assert_fail+0x21) [0x2b7b4f48edf1]
.....

The full input files are contains in the attachment. LiF.zip

Expected behavior

No response

To Reproduce

No response

Environment

abacus 3.2.0

Additional Context

No response

[mohanchen]

For further discussions, we will move to https://github.com/deepmodeling/abacus-develop