I am trying to install QE with GPU. I used this configure instruction:
sudo ./configure --with-cuda=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/cuda/ --with-cuda-runtime=12.02 --with-cuda-cc=61 CC=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvc CXX=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvc++ FC=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvfortran MPIF90=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/comm_libs/12.3/openmpi4/openmpi-4.1.5/bin/mpif90 --with-cuda=yes
and the results is:
checking build system type... x86_64-pc-linux-gnu
checking ARCH... x86_64
checking setting AR... ... ar
checking setting ARFLAGS... ... ruv
checking whether the Fortran compiler works... yes
checking for Fortran compiler default output file name... a.out
checking for suffix of executables...
checking whether we are cross compiling... no
checking for suffix of object files... o
checking whether the compiler supports GNU Fortran... no
checking whether /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvfortran accepts -g... yes
checking for Fortran flag to compile .f90 files... none
checking for /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/comm_libs/12.3/openmpi4/openmpi-4.1.5/bin/mpif90... no
checking whether the compiler supports GNU Fortran... no
checking whether accepts -g... no
configure: WARNING: MPIF90 not found: using MPIF90 anyway
checking version of /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/comm_libs/12.3/openmpi4/openmpi-4.1.5/bin/mpif90... nvfortran in
setting F90... nvfortran
setting MPIF90... /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/comm_libs/12.3/openmpi4/openmpi-4.1.5/bin/mpif90
checking whether the compiler supports GNU C... yes
checking whether /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvc accepts -g... yes
checking for /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvc option to enable C11 features... none needed
setting CC... /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvc
setting CFLAGS... -O3
using F90... nvfortran
setting FFLAGS... -fast
setting F90FLAGS... -fast -Mcache_align -Mpreprocess -Mlarge_arrays
setting FFLAGS_NOOPT... -O0
setting FFLAGS_NOMAIN... -Mnomain
setting CPP... cpp
setting CPPFLAGS... -P -traditional -Uvector
setting LD... /opt/nvidia/hpc_sdk/Linux_x86_64/24.1/comm_libs/12.3/openmpi4/openmpi-4.1.5/bin/mpif90
setting LDFLAGS...
./install/configure: line 4677: test: in: integer expression expected
./install/configure: line 4677: test: in: integer expression expected
./install/configure: line 4681: test: in: integer expression expected
./install/configure: line 4685: test: in: integer expression expected
./install/configure: line 4686: test: in: integer expression expected
checking whether Fortran compiler accepts -cuda -gpu=cuda12.02... no
configure: error: You do not have the cudafor module. Are you using NVHPC compiler?
___________________________________________________________-
any advise?
Thanks for your help!
are you see nvidia developer?
https://forums.developer.nvidia.com/t/compiling-quantum-espresso-with-gpu-support/222227
Thanks for you suggestion معین پورصادق
I have found the correct way to compile QE with GPU. In my case is:
sudo ./configure --with-cuda=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/cuda/ --with-cuda-runtime=12.3 --with-cuda-cc=61 --disable-parallel F90=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvfortran MPIF90=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvfortran FC=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvfortran CC=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvc CXX=/opt/nvidia/hpc_sdk/Linux_x86_64/24.1/compilers/bin/nvc++
One important thing to said is that I asked in https://gitlab.com/QEF/q-e/-/issues/ and they said that is not convenience to compile with sudo.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude