Skip to content

Latest commit

 

History

History
184 lines (144 loc) · 7.26 KB

TPL-Manual-Install.md

File metadata and controls

184 lines (144 loc) · 7.26 KB

There are a few externally developed third-party libraries (TPL) that are required to build SEACAS. You can build the libraries manually as detailed below, or you can use the install-tpl.sh script as described in the previous section.

Zoltan

A snapshot of zoltan_distrib_v3.83.tar.gz is provided in seacas/packages/zoltan. This will be built automatically as part of the SEACAS build process.

HDF5

If you are using the netcdf-4 capability in the netcdf library or are using the MatIO library for conversion of exodus to/from matlab format, then you will need the hdf5 library.

WARNING

There are some issues with using HDF5-1.10.0 through HDF5-1.10.2 since it will possibly create files which are unreadable by applications using an earlier version of the library. As of HDF5-1.10.3 and later, the HDF5 team added an option that makes it possible for the library to create files readable by those applications. This flag is currently being used by NetCDF, but not CGNS. Therefore, you should only use hdf5-1.10.3 or later if you are only using NetCDF, or if you do not need compatability with applications using an HDF5-1.8.X version.

The hdf5 library is used for the netcdf4 capability in netcdf which in turn is used by exodus. The netcdf4 capability is typically used for large models (>150 million elements); if you are not planning to create or read models of this size, you do not have to build hdf5.

NetCDF

The most recent released version is recommended. For use with Exodus, some local modifications to the netcdf.h include file are required if using verions prior to 4.5.1. See NetCDF-Mapping.md for an explanation of why these modifications are required (or highly recommended)

  • Download the latest netcdf-c release from https://www.unidata.ucar.edu/downloads/netcdf/index.jsp and put it inside seacas/TPL/netcdf

  • cd TPL/netcdf

  • tar zxvf netcdf-4.6.1.tar.gz

  • If the version is prior to 4.5.1, then you need to modify the following defines in seacas/TPL/netcdf/netcdf-4.6.1/include/netcdf.h. Versions 4.5.1 or later do not check these limits and can be run unmodified.

    #define NC_MAX_DIMS     65536    /* max dimensions per file */
#define NC_MAX_VARS     524288   /* max variables per file */

  • If you did not build HDF5, then you will need to edit the runcmake.sh script and remove all lines mentioning HDF5 and also set ENABLE_NETCDF _4 to OFF

  • cd netcdf-4.6.1 and enter the command:

    mkdir build
cd build
sh ../../runcmake.sh

  • make && make install

Kokkos

The most recent released version is recommended.

  • Download the latest Kokkos release from https://github.com/kokkos/kokkos/releases and put it inside seacas/TPL/kokkos

  • cd TPL/kokkos

  • tar zxvf 2.8.0.tar.gz

  • cd kokkos-2.8.0 and enter the command:

    mkdir build
cd build
sh ../../runcmake.sh

  • The default runcmake.sh will only build the serial device. Modify build options to tailor for your system as documented in the Kokkos documentation.

  • make && make install

MatIO

The MatIO library is used in the exo2mat and mat2exo programs which convert an exodus file to and from a MATLAB binary file. To use this do:

  • Download matio via git:

  • cd TPL/matio

  • git clone https://github.com/tbeu/matio.git

  • cd matio and enter the command:

    ./autogen.sh
sh ../runconfigure.sh

  • make && make install

GNU Parallel

GNU Parallel is a shell tool for executing jobs in parallel using one or more computers. A job is typically a single command or a small script that has to be run for each of the lines in the input. The typical input is a list of files, a list of hosts, a list of users, or a list of tables. In SEACAS, this is only used by epup which runs multiple epu jobs concurrently. To build:

  • Download the most recent version of the library from ftp://ftp.gnu.org/gnu/parallel/parallel-latest.tar.bz2.

    cd TPL/parallel
tar jxvf /path/to/parallel-latest.tar.bz2
cd parallel-20150522
sh ../runconfigure.sh

  • make && make install

CGNS

Support for CGNS in the IOSS library is being added. To use this capability, you will need to download and install the CGNS library:

  • Download CGNS via git:

    cd TPL/cgns
git clone https://github.com/CGNS/CGNS.git

  • Build using CMake.

    • Modify TPL/cgns/runconfigure.sh to meet your environment
    • cd CGNS
    • mkdir build
    • cd build
    • ../../runconfigure.sh
    • make && make install

DataWarehouse

The Data Warehouse is a collection of data management tools that Sandia is currently developing to improve how datasets migrate between memory and storage resources in a distributed system. While the software is currently only available internally to Sandia, there will be an open source release of the tools during Fall of 2017. The Data Warehouse extensions to SEACAS are intended to serve as a placeholder for the upcoming release and are not intended for use by external developers at this time.

The repository data-warehouse-release is a superbuild for the Data Warehouse tools and was created to make the build process as easy as possible. It includes the files INSTALL.md and INSTALL_TPL.md which contain instructions for building Data Warehouse and it's TPLs: Boost, googletest, libfabric, and libhio (optional). These builds are straightforward so a "runconfigure.sh" script is left to the end-user. Note that it's possible to supply your own build of these tools. Following the SEACAS pattern for building TPLs:

    cd TPL
    git clone [email protected]:nessie-dev/data-warehouse-release.git
    follow the instructions in INSTALL_TPL.md and then INSTALL.md
    consider installing data-warehouse-release and it's TPLs to the directory set in  the $ACCESS env. var.

To build SEACAS with an installation of data-warehouse-release and it's TPLs, add the following lines to the list of cmake command arguments as found in the top-level cmake-config file. Where DataWarehouse_PATH in environment variable that contains the path to the top-level install directory for the Data Warehouse, and HAVE_DATA_WAREHOUSE={ON|OFF} is a variable defined in cmake-config.

    -DTPL_ENABLE_DATAWAREHOUSE:BOOL=${HAVE_DATA_WAREHOUSE}           \
    -DDataWarehouse_LIBRARY_DIRS:PATH=${DataWarehouse_PATH}/lib     \
    -DDataWarehouse_INCLUDE_DIRS:PATH=${DataWarehouse_PATH}/include \