Building and installing HJCFIT:

Compiling DCProgs

A couple of design decisions affect the compilation of DCProgs.

  • c++11 is the new standard for

    the C++ programming languages. It is almost fully implemented by modern (2013) compilers. However, access to c++11 is now always default, and not always straight-forward. However, c++11 introduces a number of features that simplifies programming (e.g. move semantics) greatly. This is a forward looking solution implying some temporary hassle.

  • [GTest](https://code.google.com/p/googletest/) is the c++ unit-test

    framework from google. It is required when running DCProgs’ unit tests only. However, GTest must be compiled by the code it is testing. This means it should be shipped with DCProgs, or it should be downloaded automatically by the compilation tools. This is the option we have chosen. When compiling tests, CMake will automatically download and compile GTest

  • The math is done using Eigen,

    an efficient and widely used C++ numerical library.

Dependencies

  1. Modern C++ compiler:
    • g++-4.6 or greater on any platform,
    • clang with g++-4.2 on Mountain Lion
    • intel 13 with gnu standard libraries from g++ 4.6 (intel provides older gnu standard libraries).

  2. CMake It is available on any platform, either directly through its website or via a packaging system (yum, apt-get for Linux, Homebrew for Mac, chocolatey for windows)

  3. Eigen: Users can either install eigen, or install wget, curl or mercurial and let the build process download eigen.

To compile the python bindings for HJCFIT a few additional dependencies are needed.

  1. A working Python installation.

Multiple different ways of installing python exist. In general we recommend Anaconda but alternatives should work as well. In any case Python along with numpy and scipy should be installed. HJCFIT supports both Python 2.7 and Python 3

  1. SWIG used to generate the wrappings between C++ and Python.
  2. Behave. A behavior driven developement framework for Python.

Compilation on Mac:

Make sure that you have the needed dependencies installed. We recommend using Homebrew to manage dependencies on OSX. Follow the Homebrew documentation to install it. The instructions below assumes that you are running a working python installation with Numpy and `Scipy. Behave is not currently in Anaconda and should be installed using pip in any case.

Installing dependencies using Homebrew:

brew install swig cmake
pip install behave

Then configure and build the code:

cd /path/to/DCProgs
mkdir build && cd build
cmake ..
make
make test

Assuming everything works as expected we can now install HJCFIT

make install

Compilation on Legion:

For any compiler, do:

cd /path/to/DCProgs
mkdir build && cd build

Then, if the preferred compiler is g++, the module compilers/gnu/4.6.3 should first be loaded. A working python version should be loaded. Currently the Legion python modules does not have behave installed so it’s recommended to install it using pip --user. We also need to module load SWIG.

pip install --user behave
module load python3/recommended
module load swig/3.0.7/gnu-4.9.2
module swap compilers compilers/gnu/4.9.2

Then:

cmake ..
make
make test

Assuming everything works as expected we can now install HJCFIT

make install

Compilation on Archer:

See Running HJCFIT on Archer: for more information on running HJCFIT on Archer. Assuming that you have modules setup as in that document building HJCFIT should be as simple as:

git clone -b develop https://github.com/DCPROGS/HJCFIT.git
cd HJCFIT
mkdir build
cd build
cmake ..
make install
make test

Building MPI4Py:

Building mpi4py on Archer for use with the cray MPI and Anaconda is a bit complicated as we need to tweak the config to link against the MPI library used on Archer. This is assuming that we use the anaconda compute modules and gcc 5.1

The following worked for me:

export CC=cc
export CXX=CC
python setup.py build --mpi=mpich3
python setup.py install

where the file mpi.cfg contains:

[mpich3]
mpi_dir = /opt/cray/mpt/7.2.6/gni/mpich-gnu/51
mpicc   = /opt/cray/craype/default/bin/cc
mpicxx  = /opt/cray/craype/default/bin/CC

Compilation on Windows:

Several different ways of building and installing on Windows exist. It should be possible to build the code with both MS Visual Studio and MinGW. Currently we recommend building using MS Visual Studio 2015. The free Community edition of Visual Studio is sufficient to build HJCFIT. Note that older versions of Visual Studio did not ship 64 bit compilers in the free version. This is no longer an issue with the 2015 version. Python 3.5 in normally build with Visual Studio 2015 where as older versions are build with older versions of Visual Studio so to reduce any issues it is recommended to use Visual Studio 2015 and Python 3.5.

Visual Studio 2015:

First ensure Visual Studio is installed. Make sure to select the C++ components during the installation.

You then need to install the dependencies Swig, CMake. You can install curl from Anaconda to enable automatic download of Eigen. It’s recommended to install CMake and Swig from their respective homepages. Make sure that you select add to path when installing CMake. Following this open a command prompt.

To put the relevant Microsoft compilers on Path you should run the relevant bat script. On most systems it should be something like:

"C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\vcvarsall.bat" amd64

You can verify that Visual Studio is correctly loaded by execution cl as

To install Eigen we need wget, curl or mercurial. Curl can be installed directly from conda. To run the Python tests we need to install behave.

conda install curl
pip install behave

We can now build the code and run the tests. Unlike linux and OSX we need to explicitly tell CMake that we want a Release build.

cd /path/to/HJCFIT
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -G "NMake Makefiles"
nmake
nmake test

Assuming everything works as expected we can now install HJCFIT

nmake install