Skip to article frontmatterSkip to article content

Installation Guide

Quickstart

In the root library folder execute:

$ mkdir build
$ cd build
$ cmake ..
$ make check # (optional, runs unit tests)
$ make install

Important Installation Notes

  1. GTSAM requires the following libraries to be installed on your system:

    • BOOST version 1.65 or greater (install through Linux repositories or MacPorts). Please see Boost Notes for version recommendations based on your compiler.

    • Cmake version 3.0 or higher

    • Support for XCode 4.3 command line tools on Mac requires CMake 2.8.8 or higher

    Optional dependent libraries:

    • If TBB is installed and detectable by CMake GTSAM will use it automatically. Ensure that CMake prints “Use Intel TBB : Yes”. To disable the use of TBB, disable the CMake flag GTSAM_WITH_TBB (enabled by default) by providing the argument -DGTSAM_WITH_TBB=OFF to cmake. On Ubuntu, TBB may be installed from the Ubuntu repositories, and for other platforms it may be downloaded from https://www.threadingbuildingblocks.org/
    • GTSAM may be configured to use MKL by toggling GTSAM_WITH_EIGEN_MKL and GTSAM_WITH_EIGEN_MKL_OPENMP to ON; however, best performance is usually achieved with MKL disabled. We therefore advise you to benchmark your problem before using MKL.

    Tested compilers:

    • GCC 4.2-7.3
    • OS X Clang 2.9-10.0
    • OS X GCC 4.2
    • MSVC 2010, 2012, 2017

    Tested systems:

    • Ubuntu 16.04 - 18.04
    • MacOS 10.6 - 10.14
    • Windows 7, 8, 8.1, 10

  2. GTSAM makes extensive use of debug assertions, and we highly recommend you work in Debug mode while developing (enabled by default). Likewise, it is imperative that you switch to release mode when running finished code and for timing. GTSAM will run up to 10x faster in Release mode! See the end of this document for additional debugging tips.

  3. GTSAM has Doxygen documentation. To generate, run ‘make doc’ from your build directory after setting the GTSAM_BUILD_DOCS and GTSAM_BUILD_[HTML|LATEX] cmake flags.

  4. The instructions below install the library to the default system install path and build all components. From a terminal, starting in the root library folder, execute commands as follows for an out-of-source build:

$ mkdir build
$ cd build
$ cmake ..
$ make check (optional, runs unit tests)
$ make install

This will build the library and unit tests, run all of the unit tests, and then install the library itself.

Boost Notes

Versions of Boost prior to 1.65 have a known bug that prevents proper “deep” serialization of objects, which means that objects encapsulated inside other objects don’t get serialized. This is particularly seen when using clang as the C++ compiler.

For this reason we recommend Boost>=1.65, and recommend installing it through alternative channels when it is not available through your operating system’s primary package manager.

Known Issues

Windows Installation

This section details how to build a GTSAM .sln file using Visual Studio.

Prerequisites

Steps

  1. Open Visual Studio.
  2. Select Open a local folder and select the GTSAM source directory.
  3. Go to Project -> CMake Settings.
  1. Saving the CMake settings should automatically generate the cache. Otherwise, click on Project -> Generate Cache. This will generate the CMake build files (as seen in the Output window).
  1. The last step will generate a GTSAM.sln file in the build directory. At this point, GTSAM can be used as a regular Visual Studio project.

Python Installation

To install the Python bindings on Windows:

Install pyparsing(>=2.4.2), pybind-stubgen>=2.5.1 and numpy(>=1.11.0) with the Python environment you wish to develop in. These can all be installed as follows:

pip install -r <gtsam_folder>/python/dev_requirements.txt
  1. Follow the above steps for GTSAM general installation.
  1. Build the project (Build > Build All).
  1. At this point, gtsam in build/<your build>/python is available to be used as a Python package. You can use pip install . in that directory to install the package.

CMake Configuration Options and Details

GTSAM has a number of options that can be configured, which is best done with one of the following:

Important Options:

CMAKE_BUILD_TYPE

We support several build configurations for GTSAM (case insensitive)

cmake -DCMAKE_BUILD_TYPE=[Option] ..

CMAKE_INSTALL_PREFIX

The install folder. The default is typically /usr/local/. To configure to install to your home directory, you could execute:

cmake -DCMAKE_INSTALL_PREFIX:PATH=$HOME ..

GTSAM_TOOLBOX_INSTALL_PATH

The Matlab toolbox will be installed in a subdirectory of this folder, called ‘gtsam’.

cmake -DGTSAM_TOOLBOX_INSTALL_PATH:PATH=$HOME/toolbox ..

GTSAM_BUILD_CONVENIENCE_LIBRARIES

This is a build option to allow for tests in subfolders to be linked against convenience libraries rather than the full libgtsam. Set with the command line as follows:

cmake -DGTSAM_BUILD_CONVENIENCE_LIBRARIES:OPTION=ON ..

GTSAM_BUILD_UNSTABLE

Enable build and install for libgtsam_unstable library. Set with the command line as follows:

cmake -DGTSAM_BUILD_UNSTABLE:OPTION=ON ..

ON: When enabled, libgtsam_unstable will be built and installed with the same options as libgtsam. In addition, if tests are enabled, the unit tests will be built as well. The Matlab toolbox will also be generated if the matlab toolbox is enabled, installing into a folder called gtsam_unstable. OFF (Default) If disabled, no gtsam_unstable code will be included in build or install.

Convenience Options:

GTSAM_BUILD_EXAMPLES_ALWAYS

Whether or not to force building examples, can be true or false.

GTSAM_BUILD_TESTS

Whether or not to build tests, can be true or false.

Check

make check will build and run all of the tests. Note that the tests will only be built when using the “check” targets, to prevent make install from building the tests unnecessarily. You can also run make timing to build all of the timing scripts. To run check on a particular module only, run make check.[subfolder], so to run just the geometry tests, run make check.geometry. Individual tests can be run by appending .run to the name of the test, for example, to run testMatrix, run make testMatrix.run.

MEX_COMMAND: Path to the mex compiler. Defaults to assume the path is included in your shell’s PATH environment variable. mex is installed with matlab at $MATLABROOT/bin/mex

$MATLABROOT can be found by executing the command matlabroot in MATLAB

Performance

Here are some tips to get the best possible performance out of GTSAM.

  1. Build in Release mode. GTSAM will run up to 10x faster compared to Debug mode.
  2. Enable TBB. On modern processors with multiple cores, this can easily speed up optimization by 30-50%. Please note that this may not be true for very small problems where the overhead of dispatching work to multiple threads outweighs the benefit. We recommend that you benchmark your problem with/without TBB.
  3. Use GTSAM_BUILD_WITH_MARCH_NATIVE. A performance gain of 25-30% can be expected on modern processors. Note that this affects the portability of your executable. It may not run when copied to another system with older/different processor architecture. Also note that all dependent projects must be compiled with the same flag, or seg-faults and other undefined behavior may result.
  4. Possibly enable MKL. Please note that our benchmarks have shown that this helps only in very limited cases, and actually hurts performance in the usual case. We therefore recommend that you do not enable MKL, unless you have benchmarked it on your problem and have verified that it improves performance.

Debugging tips

Another useful debugging symbol is _GLIBCXX_DEBUG, which enables debug checks and safe containers in the standard C++ library and makes problems much easier to find.

NOTE: The native Snow Leopard g++ compiler/library contains a bug that makes it impossible to use _GLIBCXX_DEBUG. MacPorts g++ compilers do work with it though.

NOTE: If _GLIBCXX_DEBUG is used to compile gtsam, anything that links against gtsam will need to be compiled with _GLIBCXX_DEBUG as well, due to the use of header-only Eigen.

Installing MKL on Linux

Intel has a guide for installing MKL on Linux through APT repositories at https://software.intel.com/en-us/articles/installing-intel-free-libs-and-python-apt-repo.

After following the instructions, add the following to your ~/.bashrc (and afterwards, open a new terminal before compiling GTSAM): LD_PRELOAD need only be set if you are building the python wrapper to use GTSAM from python.

source /opt/intel/mkl/bin/mklvars.sh intel64
export LD_PRELOAD="$LD_PRELOAD:/opt/intel/mkl/lib/intel64/libmkl_core.so:/opt/intel/mkl/lib/intel64/libmkl_sequential.so"

To use MKL in GTSAM pass the flag -DGTSAM_WITH_EIGEN_MKL=ON to cmake.

The LD_PRELOAD fix seems to be related to a well known problem with MKL which leads to lots of undefined symbol errors, for example:

Failing to specify LD_PRELOAD may lead to errors such as: ImportError: /opt/intel/mkl/lib/intel64/libmkl_vml_avx2.so: undefined symbol: mkl_serv_getenv or Intel MKL FATAL ERROR: Cannot load libmkl_avx2.so or libmkl_def.so. when importing GTSAM using the python wrapper.

GTSAM Docs
GTSAM Docs
GTSAM Docs
User Guide