Skip to content

README_GPU.md

Latest commit

 

History

History
71 lines (57 loc) · 2.61 KB

README_GPU.md

File metadata and controls

71 lines (57 loc) · 2.61 KB

Quantum ESPRESSO GPU

License: GPL v2

This repository also contains the GPU-accelerated version of Quantum ESPRESSO.

Installation

This version requires the nvfortran (previously PGI) compiler from the NVidia HPC SDK, v.21.7 or later (freely downloadable from NVidia). Earlier versions may or may not work and are no longer supported. You are advised to use the most recent version of NVidia software you can find.

For compilation using CMake, see GitLab.com/QEF/q-e/-/wikis/Developers/CMake-build-system. For compilation using configure, see the User Guide in Doc/. The configure script checks for the presence of the nvfortran compiler and of a few cuda libraries. The NVidia SDK (standard development kit) must be properly installed under $NVHPC_CUDA_HOME and the path pointing to the cuda libraries must be present in LD_LIBRARY_PATH. A template for the configure command is:

./configure --with-gpu=cuda --with-cuda-runtime=YY --with-cuda-cc=ZZ --enable-openmp [ --with-scalapack=no ][ --with-cuda-mpi=yes ]

where YY is the version of the cuda toolkit and ZZ is the compute capability of the card. You can get those numbers from command nvaccelinfo, if you have a properly configured HPC SDK:

$ nvaccelinfo | grep -e 'Target' -e 'Driver'
CUDA Driver Version:           11000
Default Target:                cc70
...

The version is returned as (1000 major + 10 minor). For example, CUDA 11.0 is represented by 11000. For the above case, configure QE with:

./configure --with-cuda=$CUDA_HOME --with-cuda-cc=70 --with-cuda-runtime=11.0

One can also use command nvidia-smi: for two GPUs with cc70,

$ nvidia-smi --query-gpu=compute_cap --format=csv
7.0
7.0

Enabling faster communications between GPUs, via NVlink or Infiniband RDMA, is essential for optimal performance. If your MPI library is built to be CUDA-aware, then enable --with-cuda-mpi=yes (default: no).

Serial (no MPI) compilation is also supported: use --disable-parallel.

Options --with-cuda=XX and --with-openacc are no longer honored (OpenACC is always needed). It is generally a good idea to disable Scalapack when running small test cases since the serial GPU eigensolver outperforms the parallel CPU eigensolver in many circumstances.

Execution

By default, GPU support is active. The following message will appear at the beginning of the output

     GPU acceleration is ACTIVE.

The current GPU version passes all tests with both parallel and serial compilation.