/qe-gpu

GPU-accelerated Quantum ESPRESSO using CUDA FORTRAN

Primary LanguageFortranGNU General Public License v2.0GPL-2.0

GPU-accelerated Quantum ESPRESSO (QE-GPU)

This is an open-source custom version of Quantum ESPRESSO with embedded GPU support based on CUDA FORTRAN. This product has been made possible thanks to the effort of the NVIDIA HPC Software and Benchmarks Group. This version is maintained by Filippo Spiga, contributions are encouraged. Partial support was provided by E4 Computer Engineering SpA via the European PRACE Pre-Commercial Procurement project (Phase 3). To contribute please refer to the guidelines in CONTRIBUTING.md

Requirements

The PGI compiler version 17.4 or above is required to use QE-GPU. We suggest the latest community edition 17.10 (freely available from PGI. It containes CUDA SDK and pre-built Open MPI for parallel execution (check the PGI Installation Guide how to install it). No other compilers are supported

You need NVIDIA TESLA Kepler (K20, K40, K80) or Pascal (P100) or Volta (V100). No other cards are supported. NVIDIA TESLA P100 and V100 are strongly recommend for their on-board memory capacity and douple precision performance.

This version of QE-GPU it is based on Quantum ESPRESSO v6.1. It runs exclusively in parallel. For x86 architecture, you also also recent version Intel Math Kernel Library (MKL). For POWER architecture, make sure you can access IBM Engineering and Scientific Subroutine Library (ESSL) for maximum performance.

Installation

To compile QE-GPU there is no automatic procedure. You must copy a make.inc template from "install/" directory into the main directory, edit it based and then run make. Make sure GPU_ARCH and CUDA_RUNTIME are specified correctly and various paths to libraries and header files point into the correct locations.

By invoking make alone a list of acceptable targets will be displayed. Binaries go in "bin/". Read comments in the make.inc templates to customize it further based on your ebvironment and where math libraries are located. The architectures/environments supported are x86-64, POWER and CRAY.

These templates are available:

  • make.inc_x86-64 to compile on a generic x86-64 machine with NVIDIA GPU
  • make.inc_CRAY_PizDaint to compile on Piz Daint at CSCS, CRAY XC30 with P100 GPU (GPU_ARCH=60)
  • make.inc_POWER_DAVIDE* to compile on PRACE "DAVIDE" machine at CINECA, based on POWER8 with GPU (GPU_ARCH=60)
  • make.inc_POWER_SUMMITDEV to compile on early access system SUMMITDEV at ORNL, based on POWER8 with GPU (GPU_ARCH=60)

The QE-GPU package has been reduced in size to the minimum essential. For more information, please refer to the general documentation provided with the full Quantum ESPRESSO suite or visit the official web site http://www.quantum-espresso.org/

Citation

If you use the code for science or any form of scientific and technical dissemination activity, we kindly ask to cite the code using the two following references:

  • Romero, J., Phillips, E. Fatica, M., Spiga, F.: GPU-accelerated Quantum ESPRESSO, Version 1.0 (November 2017), http://doi.org/10.5281/zenodo.823200
  • Romero, J., Phillips, E. Fatica, M., Spiga, F., Giannozzi, P.: A performance study of Quantum ESPRESSO's PWscf code on multi-core and GPU systems, 8th IEEE International Workshop on Performance Modeling, Benchmarking and Simulation of High Performance Computer Systems (PMBS17), Lecture Notes in Computer Science, Springer, Denver (2017)

Benchmarks

Benchmarks will be collected in a separate repository

License

All the material included in this distribution is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

These programs are distributed in the hope that they will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program; if not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.