Gromacs 2021.4 installation, make check fails at regression/complex tests

GROMACS version:2021.4
GROMACS modification: No
Here post your question: make check fails at regression/complex tests

The workstation info:
gputype=NVIDIA 3080
gpus=4
CPU 96, x86_64
Thread(s) per core: 2
Core(s) per socket: 24
Socket(s): 2

Softwares:

  • cmake 3.22.1
  • gcc 7.5.0
  • openmpi 4.0.1
  • cuda 11.5
  • python 2.7.5

cmake parameters:
cmake … -DGMX_BUILD_OWN_FFTW=ON -DREGRESSIONTEST_DOWNLOAD=ON -DGMX_GPU=CUDA -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-11.5

No errors when removing -DGMX_GPU=CUDA -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-11.5 from cmake or conducting GMX_DISABLE_GPU_DETECTION=1 make check.

Same error when using 1 gpu during make check.

Any suggestions?

detailed log:

68/71 Test #68: regressiontests/complex …***Failed 232.26 sec
Re-running aminoacids using only CPU-based non-bonded kernels
Re-running awh_multidim using only CPU-based non-bonded kernels
Re-running butane using only CPU-based non-bonded kernels
Re-running dd121 using only CPU-based non-bonded kernels

Abnormal return value for ’ gmx mdrun -ntmpi 1 -notunepme >mdrun.out 2>&1’ was 1
Retrying mdrun with better settings…
Re-running distance_restraints using only CPU-based non-bonded kernels
Mdrun cannot use the requested (or automatic) number of OpenMP threads, retrying with 8.

Abnormal return value for ’ gmx mdrun -ntmpi 1 -notunepme -nb cpu >mdrun.out 2>&1’ was 1
Retrying mdrun with better settings…
Re-running ethyleenglycol using only CPU-based non-bonded kernels
Re-running field using only CPU-based non-bonded kernels
Re-running nacl using only CPU-based non-bonded kernels
Re-running nbnxn-free-energy using only CPU-based non-bonded kernels
Re-running nbnxn-free-energy-vv using only CPU-based non-bonded kernels
Re-running nbnxn-ljpme-geometric using only CPU-based non-bonded kernels
Re-running nbnxn-ljpme-LB using only CPU-based non-bonded kernels
Re-running nbnxn-ljpme-LB-geometric using only CPU-based non-bonded kernels
Re-running nbnxn-vdw-force-switch using only CPU-based non-bonded kernels
Re-running nbnxn-vdw-potential-switch using only CPU-based non-bonded kernels
Re-running nbnxn-vdw-potential-switch-argon using only CPU-based non-bonded kernels
Re-running nbnxn_pme using only CPU-based non-bonded kernels
Re-running nbnxn_pme_order5 using only CPU-based non-bonded kernels
Re-running nbnxn_pme_order6 using only CPU-based non-bonded kernels
Re-running nbnxn_rf using only CPU-based non-bonded kernels
Re-running nbnxn_rzero using only CPU-based non-bonded kernels
Mdrun was not able to distribute the requested non-bonded tasks to the available GPUs.
Will try again with following task assignment !

Abnormal return value for ’ gmx mdrun -ntmpi 6 -notunepme >mdrun.out 2>&1’ was -1
FAILED. Check mdrun.out, md.log file(s) in nbnxn_vsite for nbnxn_vsite
Re-running nst_mismatch using only CPU-based non-bonded kernels
Re-running octahedron using only CPU-based non-bonded kernels

Abnormal return value for ’ gmx mdrun -ntmpi 1 -notunepme >mdrun.out 2>&1’ was 1
Retrying mdrun with better settings…
Re-running orientation-restraints using only CPU-based non-bonded kernels
Mdrun cannot use the requested (or automatic) number of OpenMP threads, retrying with 8.

Abnormal return value for ’ gmx mdrun -ntmpi 1 -notunepme -nb cpu >mdrun.out 2>&1’ was 1
Retrying mdrun with better settings…
Re-running position-restraints using only CPU-based non-bonded kernels
Re-running pr-vrescale using only CPU-based non-bonded kernels
Re-running pull_constraint using only CPU-based non-bonded kernels
Re-running pull_cylinder using only CPU-based non-bonded kernels
Re-running pull_geometry_angle using only CPU-based non-bonded kernels
Re-running pull_geometry_angle-axis using only CPU-based non-bonded kernels
Re-running pull_geometry_dihedral using only CPU-based non-bonded kernels

Abnormal return value for ’ gmx mdrun -ntmpi 1 -notunepme >mdrun.out 2>&1’ was 1
Retrying mdrun with better settings…
Re-running sw using only CPU-based non-bonded kernels
Mdrun cannot use the requested (or automatic) number of OpenMP threads, retrying with 8.

Abnormal return value for ’ gmx mdrun -ntmpi 1 -notunepme -nb cpu >mdrun.out 2>&1’ was 1
Retrying mdrun with better settings…
Re-running swap_x using only CPU-based non-bonded kernels
Re-running swap_y using only CPU-based non-bonded kernels
Re-running swap_z using only CPU-based non-bonded kernels
Re-running tip4p using only CPU-based non-bonded kernels
Re-running tip4pflex using only CPU-based non-bonded kernels
Re-running urea using only CPU-based non-bonded kernels
Re-running walls using only CPU-based non-bonded kernels
1 out of 87 complex tests FAILED

  Start 69: regressiontests/freeenergy

69/71 Test #69: regressiontests/freeenergy … Passed 74.92 sec
Start 70: regressiontests/rotation
70/71 Test #70: regressiontests/rotation … Passed 20.14 sec
Start 71: regressiontests/essentialdynamics
71/71 Test #71: regressiontests/essentialdynamics … Passed 16.90 sec

99% tests passed, 1 tests failed out of 71