Skip to content

Latest commit

 

History

History
278 lines (224 loc) · 12.4 KB

INSTALL.rst

File metadata and controls

278 lines (224 loc) · 12.4 KB
Raw
.. seealso::

   You will need to understand how to use `the terminal <http://www.ks.uiuc.edu/Training/Tutorials/Reference/unixprimer.html>`_, what are `environment variables <https://unix.stackexchange.com/questions/44990/what-is-the-difference-between-path-and-ld-library-path/45106#45106>`_ and please read our :ref:`compiling introduction <install-source>`.

Note

If you are a scientific user at a supercomputing facility we might have already prepared a software setup for you. See the :ref:`following chapter <install-profile>` if you can skip this step fully or in part by loading existing modules on those systems.

Dependencies

.. sectionauthor:: Axel Huebl, Klaus Steiniger, Sergei Bastrakov

Overview

overview of PIConGPU library dependencies

Overview of inter-library dependencies for parallel execution of PIConGPU on a typical HPC system. Due to common binary incompatibilities between compilers, MPI and boost versions, we recommend to organize software with a version-aware package manager such as spack and to deploy a hierarchical module system such as lmod. An Lmod example setup can be found here.

Requirements

Mandatory

Compiler
  • C++17 supporting compiler, e.g. GCC 7+ or Clang 6+
  • if you want to build for Nvidia GPUs, check the CUDA supported compilers page
  • note: be sure to build all libraries/dependencies with the same compiler version
  • Debian/Ubuntu:
    • sudo apt-get install gcc-8 g++-8 build-essential
    • sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-8 60 --slave /usr/bin/g++ g++ /usr/bin/g++-8
  • Arch Linux:
  • Spack:
    • spack install gcc@8.5.0
    • make it the default in your packages.yaml or suffix all following spack install commands with a space and %gcc@8.5.0
CMake
  • 3.18.0 or higher
  • Debian/Ubuntu: sudo apt-get install cmake file cmake-curses-gui
  • Arch Linux: sudo pacman --sync cmake
  • Spack: spack install cmake
MPI 2.3+
  • OpenMPI 1.7+ / MVAPICH2 1.8+ or similar
  • for running on Nvidia GPUs, perform a GPU aware MPI install after installing CUDA
  • Debian/Ubuntu: sudo apt-get install libopenmpi-dev
  • Arch Linux: sudo pacman --sync openmpi
  • Spack:
    • GPU support: spack install openmpi+cuda
    • CPU only: spack install openmpi
  • environment:
    • export MPI_ROOT=<MPI_INSTALL>
    • as long as CUDA awareness (openmpi+cuda) is missing: export OMPI_MCA_mpi_leave_pinned=0
Boost
  • 1.74.0+ (program_options, filesystem, system, math, serialization and header-only libs)
  • Debian/Ubuntu: sudo apt-get install libboost-program-options-dev libboost-filesystem-dev libboost-system-dev libboost-math-dev libboost-serialization-dev
  • Arch Linux: sudo pacman --sync boost
  • Spack: spack install boost
  • from source:
    • mkdir -p ~/src ~/lib
    • cd ~/src
    • curl -Lo boost_1_74_0.tar.gz https://boostorg.jfrog.io/artifactory/main/release/1.74.0/source/boost_1_74_0.tar.gz
    • tar -xzf boost_1_74_0.tar.gz
    • cd boost_1_74_0
    • ./bootstrap.sh --with-libraries=filesystem,math,program_options,serialization,system --prefix=$HOME/lib/boost
    • ./b2 cxxflags="-std=c++17" -j4 && ./b2 install
  • environment: (assumes install from source in $HOME/lib/boost)
    • export BOOST_ROOT=$HOME/lib/boost
    • export LD_LIBRARY_PATH=$BOOST_ROOT/lib:$LD_LIBRARY_PATH
git
  • not required for the code, but for our workflows
  • 1.7.9.5 or higher
  • Debian/Ubuntu: sudo apt-get install git
  • Arch Linux: sudo pacman --sync git
  • Spack: spack install git
rsync
  • not required for the code, but for our workflows
  • Debian/Ubuntu: sudo apt-get install rsync
  • Arch Linux: sudo pacman --sync rsync
  • Spack: spack install rsync
alpaka 1.0.X-dev
  • alpaka is included in the PIConGPU source code
cupla 0.4.X-dev
  • cupla is included in the PIConGPU source code
mallocMC 2.6.0crp-dev
  • only required for CUDA and HIP backends
  • mallocMC is included in the PIConGPU source code

PIConGPU Source Code

  • git clone https://github.com/ComputationalRadiationPhysics/picongpu.git $HOME/src/picongpu
    • optional: update the source code with cd $HOME/src/picongpu && git fetch && git pull
    • optional: change to a different branch with git branch (show) and git checkout <BranchName> (switch)
  • environment:
    • export PICSRC=$HOME/src/picongpu
    • export PIC_EXAMPLES=$PICSRC/share/picongpu/examples
    • export PATH=$PATH:$PICSRC
    • export PATH=$PATH:$PICSRC/bin
    • export PATH=$PATH:$PICSRC/src/tools/bin
    • export PYTHONPATH=$PICSRC/lib/python:$PYTHONPATH

Optional Libraries

CUDA
  • 11.0.0+
  • required if you want to run on Nvidia GPUs
  • Debian/Ubuntu: sudo apt-get install nvidia-cuda-toolkit
  • Arch Linux: sudo pacman --sync cuda
  • Spack: spack install cuda
  • at least one CUDA capable GPU
  • compute capability: sm_60 or higher
  • full list of CUDA GPUs and their compute capability
  • More is always better. Especially, if we are talking GPUs :-)
  • environment:
    • export CUDA_ROOT=<CUDA_INSTALL>

If you do not install the following libraries, you will not have the full amount of PIConGPU plugins. We recommend to install at least pngwriter and openPMD.

libpng
  • 1.2.9+ (requires zlib)
  • Debian/Ubuntu dependencies: sudo apt-get install libpng-dev
  • Arch Linux dependencies: sudo pacman --sync libpng
  • Spack: spack install libpng
  • from source:
    • mkdir -p ~/src ~/lib
    • cd ~/src
    • curl -Lo libpng-1.6.34.tar.gz ftp://ftp-osl.osuosl.org/pub/libpng/src/libpng16/libpng-1.6.34.tar.gz
    • tar -xf libpng-1.6.34.tar.gz
    • cd libpng-1.6.34
    • CPPFLAGS=-I$HOME/lib/zlib/include LDFLAGS=-L$HOME/lib/zlib/lib ./configure --enable-static --enable-shared --prefix=$HOME/lib/libpng
    • make
    • make install
  • environment: (assumes install from source in $HOME/lib/libpng)
    • export PNG_ROOT=$HOME/lib/libpng
    • export CMAKE_PREFIX_PATH=$PNG_ROOT:$CMAKE_PREFIX_PATH
    • export LD_LIBRARY_PATH=$PNG_ROOT/lib:$LD_LIBRARY_PATH
pngwriter
  • 0.7.0+ (requires libpng, zlib, and optional freetype)
  • Spack: spack install pngwriter
  • from source:
    • mkdir -p ~/src ~/lib
    • git clone -b 0.7.0 https://github.com/pngwriter/pngwriter.git ~/src/pngwriter/
    • cd ~/src/pngwriter
    • mkdir build && cd build
    • cmake -DCMAKE_INSTALL_PREFIX=$HOME/lib/pngwriter ..
    • make install
  • environment: (assumes install from source in $HOME/lib/pngwriter)
    • export CMAKE_PREFIX_PATH=$HOME/lib/pngwriter:$CMAKE_PREFIX_PATH
    • export LD_LIBRARY_PATH=$HOME/lib/pngwriter/lib:$LD_LIBRARY_PATH
openPMD API
  • optional, but strongly recommended as most PIConGPU output requires it
  • 0.14.3+
  • Spack: spack install openpmd-api
  • For usage in PIConGPU, the openPMD API must have been built either with support for ADIOS2 or HDF5 (or both). When building the openPMD API from source (described below), these dependencies must be built and installed first.
    • For ADIOS2, CMake build instructions can be found in the official documentation. Besides compression, the default configuration should generally be sufficient, the CMAKE_INSTALL_PREFIX should be set to a fitting location. Compression with c-blosc is described below.
    • For HDF5, CMake build instructions can be found in the official documentation. The parameters -DHDF5_BUILD_CPP_LIB=OFF -DHDF5_ENABLE_PARALLEL=ON are required, the CMAKE_INSTALL_PREFIX should be set to a fitting location.
  • from source:
    • mkdir -p ~/src ~/lib
    • git clone -b 0.14.4 https://github.com/openPMD/openPMD-api.git ~/src/openPMD-api
    • cd ~/src/openPMD-api
    • mkdir build && cd build
    • cmake .. -DopenPMD_USE_MPI=ON -DCMAKE_INSTALL_PREFIX=~/lib/openPMD-api Optionally, specify the parameters -DopenPMD_USE_ADIOS2=ON -DopenPMD_USE_HDF5=ON. Otherwise, these parameters are set to ON automatically if CMake detects the dependencies on your system.
    • make -j $(nproc) install
  • environment:* (assumes install from source in $HOME/lib/openPMD-api)
    • export CMAKE_PREFIX_PATH="$HOME/lib/openPMD-api:$CMAKE_PREFIX_PATH"
    • export LD_LIBRARY_PATH="$HOME/lib/openPMD-api/lib64:$HOME/lib/openPMD-api/lib:$LD_LIBRARY_PATH"
  • If PIConGPU is built with openPMD output enabled, the JSON library nlohmann_json will automatically be used, found in the thirdParty/ directory. By setting the CMake parameter PIC_nlohmann_json_PROVIDER=extern, CMake can be instructed to search for an installation of nlohmann_json externally. Refer to LICENSE.md for further information.
c-blosc for openPMD API with ADIOS2
  • not a direct dependency of PIConGPU, but an optional dependency for openPMD API with ADIOS2; installation is described here since it is lacking in documentation elsewhere
  • general purpose compressor, used in ADIOS2 for in situ data reduction
  • Debian/Ubuntu: sudo apt-get install libblosc-dev
  • Arch Linux: sudo pacman --sync blosc
  • Spack: spack install c-blosc
  • from source:
    • mkdir -p ~/src ~/lib
    • git clone -b v1.21.1 https://github.com/Blosc/c-blosc.git ~/src/c-blosc/
    • cd ~/src/c-blosc
    • mkdir build && cd build
    • cmake -DCMAKE_INSTALL_PREFIX=$HOME/lib/c-blosc -DPREFER_EXTERNAL_ZLIB=ON ..
    • make install
  • environment: (assumes install from source in $HOME/lib/c-blosc)
    • export BLOSC_ROOT=$HOME/lib/c-blosc
    • export CMAKE_PREFIX_PATH=$BLOSC_ROOT:$CMAKE_PREFIX_PATH
    • export LD_LIBRARY_PATH=$BLOSC_ROOT/lib:$LD_LIBRARY_PATH
ISAAC
  • 1.4.0+
  • requires boost (header only), IceT, Jansson, libjpeg (preferably libjpeg-turbo), libwebsockets (only for the ISAAC server, but not the plugin itself)
  • enables live in situ visualization, see more here Plugin description
  • Spack: spack install isaac
  • from source: build the in situ library and its dependencies as described in ISAAC's INSTALL.md
  • environment: set environment variable CMAKE_PREFIX_PATH for each dependency and the ISAAC in situ library
VampirTrace
  • for developers: performance tracing support
  • download 5.14.4 or higher, e.g. from www.tu-dresden.de
  • from source:
    • mkdir -p ~/src ~/build ~/lib
    • cd ~/src
    • curl -Lo VampirTrace-5.14.4.tar.gz "http://wwwpub.zih.tu-dresden.de/~mlieber/dcount/dcount.php?package=vampirtrace&get=VampirTrace-5.14.4.tar.gz"
    • tar -xzf VampirTrace-5.14.4.tar.gz
    • cd VampirTrace-5.14.4
    • ./configure --prefix=$HOME/lib/vampirtrace --with-cuda-dir=<CUDA_ROOT>
    • make all -j
    • make install
  • environment: (assumes install from source in $HOME/lib/vampirtrace)
    • export VT_ROOT=$HOME/lib/vampirtrace
    • export PATH=$VT_ROOT/bin:$PATH