Compilation and Installation

Procedure

1. Use PuTTY to log in to the server as the root user.
2. Go to the main program installation directory.

   cd /path/to/qe

3. Decompress the installation package.

   tar -zxvf q-e-qe-6.8.tar

4. Go to the directory generated after the decompression.

   cd q-e-qe-6.8/

5. Perform the compilation.

   ./configure --with-cuda=yes --with-cuda-runtime=11.4 --with-cuda-cc=80  --enable-openmp --with-scalapack=no

   

   Modify --with-cuda-cc=80 based on the graphics card architecture of the current host. In this example, the graphics card is V100.

   make -j32 pwall

   • If the compilation host is not connected to the Internet, download the devicexlib-master.tar.gz package and upload it to the compilation host. After the compilation is complete, run the make pp -j32 command.

     mv devicexlib-master.tar.gz /path/to/qe-gpu-6.7/external devicexlib.tar.gz
     tar xzf devicexlib.tar.gz --strip-components=1
     export F90FLAGS="-FAST -mCACHE_align -Mpreprocess -Mlarge_arrays"
     ./configure FC=pgf90 CC=pgcc --with-cuda=no --with-cuda-cc= --with-cuda-runtime= --disable-parallel --enable-cuda-env-check=no
     make pp -j32

   • If error information similar to the following is displayed, the link library configuration is incorrect. You need to modify the make.inc file.

     

     1. Open the make.inc file.

        vi make.inc

     2. Press i to enter the edit mode and add LD_LIBS = -lcurand to the following location:

        

     3. Press Esc, type :wq!, and press Enter to save the settings and exit.
6. Check whether an executable file is generated.

   ll /path/to/qe-gpu-6.7/bin/pw.x

   If the command output contains the pw.x file information, the executable file has been generated.

7. Set the environment variables.

   export PATH=/path/to/qe/bin/:$PATH