*** WARNING in fm/cp_fm_elpa.F:522 :: Setting real_kernel for ELPA failed ***

[jona...@hotmail.com]

I have built Cp2k with the following toolchain command:" 

./install_cp2k_toolchain.sh --install-all --with-tblite=no --with-ace=no --with-gcc=system --math-mode=mkl --with-mkl=/opt/intel/oneapi/mkl/2025.2 --mpi-mode=openmpi --enable-cuda=yes --gpu-ver=A100 --target-cpu=znver2 -j 32  " 

And built Gromacs with the following cmake command:" 

cmake .. \
         -DGMX_BUILD_OWN_FFTW=ON \
         -DREGRESSIONTEST_DOWNLOAD=OFF \
         -DCMAKE_INSTALL_PREFIX=~/gromacs \
         -DBUILD_SHARED_LIBS=OFF \
         -DGMX_PREFER_STATIC_LIBS=ON \
         -DGMXAPI=OFF \
         -DGMX_INSTALL_NBLIB_API=OFF \
         -DGMX_DOUBLE=OFF \
         -DGMX_SIMD=AVX2_256 \
         -DGMX_MPI=ON \
         -DGMX_GPU=CUDA \  
         -DGMX_CUDA_TARGET_SM="80" \
         -DGMX_CP2K=ON \
         -DCMAKE_CUDA_COMPILER=/usr/local/cuda-12.6//bin/nvcc \
         -DCMAKE_C_COMPILER=/home/users/jonasbaa/WaterPhase/cp2k/tools/toolchain/install/openmpi-5.0.8/bin/mpicc \
         -DCMAKE_CXX_COMPILER=/home/users/jonasbaa/WaterPhase/cp2k/tools/toolchain/install/openmpi-5.0.8/bin/mpicxx \
         -DCMAKE_Fortran_COMPILER=/home/users/jonasbaa/WaterPhase/cp2k/tools/toolchain/install/openmpi-5.0.8/bin/mpifort \
         -DCP2K_DIR=/home/users/jonasbaa/WaterPhase/cp2k/lib/Linux-gnu-x86_64/psmp \
         -DCP2K_LINKER_FLAGS=' To many flags to mention here '.

But when I run my energy minimization/simulation with the simple PBE functional:" 

; CP2K section
qmmm-cp2k-active    = true      ; Activate QMMM MdModule
qmmm-cp2k-qmgroup   = QMatoms   ; Index group of QM atoms
qmmm-cp2k-qmmethod  = PBE       ; Can be amended to INPUT and then use the -qmi flag for a custom cp2k.inp file in the gmx_mpi_d grompp command.
qmmm-cp2k-qmcharge  = 0         ; Charge of QM system

qmmm-cp2k-qmmultiplicity = 1    ; Multiplicity of QM system",

I get the following warning:" *** WARNING in fm/cp_fm_elpa.F:522 :: Setting real_kernel for ELPA failed *** ", and as a result my GPU are not active during the run. I would appreciate all help I can get on how to make sure that the QM part of my QM/MM simulation can be done on the GPU. I assume it can be done by solving this warning.

Best regards,

Jonas