Simbody

Simbody is a high-performance, open-source toolkit for science- and engineering-quality simulation of articulated mechanisms, including biomechanical structures such as human and animal skeletons, mechanical systems like robots, vehicles, and machines, and anything else that can be described as a set of rigid bodies interconnected by joints, influenced by forces and motions, and restricted by constraints. Simbody includes a multibody dynamics library for modeling motion in generalized/internal coordinates in O(n) time. This is sometimes called a Featherstone-style physics engine.

Simbody provides a C++ API that is used to build domain-specific applications; it is not a standalone application itself. For example, it is used by biomechanists in OpenSim, by roboticists in Gazebo, and for biomolecular research in MacroMoleculeBuilder (MMB). Here's an artful simulation of several RNA molecules containing thousands of bodies, performed with MMB by Samuel Flores:

Read more about Simbody at the Simbody homepage.

Simple example: a double pendulum

Here's some code to simulate and visualize a 2-link chain:

#include "Simbody.h"
using namespace SimTK;
int main() {
    // Define the system.
    MultibodySystem system;
    SimbodyMatterSubsystem matter(system);
    GeneralForceSubsystem forces(system);
    Force::Gravity gravity(forces, matter, -YAxis, 9.8);

    // Describe mass and visualization properties for a generic body.
    Body::Rigid bodyInfo(MassProperties(1.0, Vec3(0), UnitInertia(1)));
    bodyInfo.addDecoration(Transform(), DecorativeSphere(0.1));

    // Create the moving (mobilized) bodies of the pendulum.
    MobilizedBody::Pin pendulum1(matter.Ground(), Transform(Vec3(0)),
            bodyInfo, Transform(Vec3(0, 1, 0)));
    MobilizedBody::Pin pendulum2(pendulum1, Transform(Vec3(0)),
            bodyInfo, Transform(Vec3(0, 1, 0)));

    // Set up visualization.
    Visualizer viz(system);
    system.addEventReporter(new Visualizer::Reporter(viz, 0.01));

    // Initialize the system and state.
    State state = system.realizeTopology();
    pendulum2.setRate(state, 5.0);

    // Simulate for 20 seconds.
    RungeKuttaMersonIntegrator integ(system);
    TimeStepper ts(system, integ);
    ts.initialize(state);
    ts.stepTo(20.0);
}

See Simbody's User Guide for a step-by-step explanation of this example.

Features

• Wide variety of joint, constraint, and force types; easily user-extended.
• Forward, inverse, and mixed dynamics. Motion driven by forces or prescribed motion.
• Contact (Hertz, Hunt and Crossley models).
• Gradient descent, interior point, and global (CMA) optimizers.
• A variety of numerical integrators with error control.
• Visualizer, using OpenGL

You want to...

• install Simbody.
• use Simbody in your own program.
• view API documentation.
• learn the theory behind Simbody.
• extend Simbody.
• get support at the Simbody Forum.
• report a bug or suggest a feature.

---

Dependencies

Simbody depends on the following:

• cross-platform building: CMake 2.8.8 or later (3.1.3 or later for Visual Studio).
• compiler: Visual Studio 2015 (Windows only), gcc 4.8.1 or later (typically on Linux), or Clang 3.4 or later (typically on Mac, possibly through Xcode)
• linear algebra: LAPACK and BLAS
• visualization (optional): FreeGLUT, Xi and Xmu
• API documentation (optional): Doxygen 1.8.6 or later; we recommend at least 1.8.8.

Using Simbody

• Creating your own Simbody-using project with CMake To get started with your own Simbody-using project, check out the cmake/SampleCMakeLists.txt file.

Installing

Simbody works on Windows, Mac, and Linux. For Windows, you must build from source. For Mac and Linux, you can use a package manager or build from source. In this file, we provide instructions for 4 different ways of installing Simbody:

1. Windows: build from source using Microsoft Visual Studio.
2. Linux or Mac (make): build from source using gcc or Clang with make.
3. Mac (Homebrew): automated build/install with Homebrew.
4. Ubuntu/Debian: install pre-built binaries with apt-get.
5. Windows using MinGW: build from source using MinGW.

These are not the only ways to install Simbody, however. For example, on a Mac, you could use CMake and Xcode.

C++11 and gcc/Clang

Simbody 3.6 and later uses C++11 features (the -std=c++11 flag). Simbody 3.3 and earlier use only C++03 features, and Simbody 3.4 and 3.5 can use either C++03 or C++11; see the SIMBODY_STANDARD_11 CMake variable in these versions. Note that if you want to use the C++11 flag in your own project, Simbody must have been compiled with the C++11 flag as well.

Windows using Visual Studio

Get the dependencies

All needed library dependencies are provided with the Simbody installation on Windows, including linear algebra and visualization dependencies.

1. Download and install Microsoft Visual Studio, version 2015. The Community edition is free and sufficient. By default, Visual Studio 2015 does not provide C++ support; when installing, be sure to select Custom, and check Programming Languages > Visual C++ > Common Tools for Visual C++ 2015. If you have already installed Visual Studio without C++ support, simply re-run the installer and select Modify.
2. Download and install CMake, version 3.1.3 or higher.
3. (optional) If you want to build API documentation, download and install Doxygen, version 1.8.8 or higher.

Download the Simbody source code

• Method 1: Download the source code from https://github.com/simbody/simbody/releases. Look for the highest-numbered release, click on the .zip button, and unzip it on your computer. We'll assume you unzipped the source code into C:/Simbody-source.
• Method 2: Clone the git repository.

  1. Get git. There are many options: Git for Windows (most advanced), TortoiseGit (intermediate; good for TortoiseSVN users), GitHub for Windows (easiest).

  2. Clone the github repository into C:/Simbody-source. Run the following in a Git Bash / Git Shell, or find a way to run the equivalent commands in a GUI client:

      $ git clone https://github.com/simbody/simbody.git C:/Simbody-source
      $ git checkout Simbody-3.5.1
     

  3. In the last line above, we assumed you want to build a released version. Feel free to change the version you want to build. If you want to build the latest development version ("bleeding edge") of Simbody off the master branch, you can omit the checkout line.

Configure and generate project files

1. Open CMake.
2. In the field Where is the source code, specify C:/Simbody-source.
3. In the field Where to build the binaries, specify something like C:/Simbody-build, just not inside your source directory. This is not where we will install Simbody; see below.
4. Click the Configure button.
   1. When prompted to select a generator, choose either Visual Studio 14 2015 (to build 32-bit binaries) or Visual Studio 14 2015 Win64 (to build 64-bit binaries).
   2. Click Finish.
5. Where do you want to install Simbody on your computer? Set this by changing the CMAKE_INSTALL_PREFIX variable. We'll assume you set it to C:/Simbody. If you choose a different installation location, make sure to use yours where we use C:/Simbody below.
6. Play around with the other build options:
   • BUILD_EXAMPLES to see what Simbody can do. On by default.
   • BUILD_TESTING to ensure your Simbody works correctly. On by default.
   • BUILD_VISUALIZER to be able to watch your system move about! If building remotely, you could turn this off. On by default.
   • BUILD_STATIC_LIBRARIES builds the three libraries as static libraries, whose names will end with _static. Off by default.
   • BUILD_TESTS_AND_EXAMPLES_STATIC if static libraries, and tests or examples are being built, creates statically-linked tests/examples. Can take a while to build, and it is unlikely you'll use the statically-linked libraries.
   • BUILD_TESTS_AND_EXAMPLES_SHARED if tests or examples are being built, creates dynamically-linked tests/examples. Unless you know what you're doing, leave this one on.
7. Click the Configure button again. Then, click Generate to make Visual Studio project files.

Build and install

1. Open C:/Simbody-build/Simbody.sln in Visual Studio.

2. Select your desired Solution configuration from the drop-down at the top.

   • Debug: debugger symbols; no optimizations (more than 10x slower). Library and visualizer names end with _d.
   • RelWithDebInfo: debugger symbols; optimized. This is the configuration we recommend.
   • Release: no debugger symbols; optimized. Generated libraries and executables are smaller but not faster than RelWithDebInfo.
   • MinSizeRel: minimum size; optimized. May be slower than RelWithDebInfo or Release.

   You at least want optimized libraries (all configurations but Debug are optimized), but you can have Debug libraries coexist with them. To do this, go through the full installation process twice, once for each configuration.

3. Build the project ALL_BUILD by right-clicking it and selecting Build.

4. Run the tests by right-clicking RUN_TESTS and selecting Build. Make sure all tests pass. You can use RUN_TESTS_PARALLEL for a faster test run if you have multiple cores.

5. (Optional) Build the project doxygen to get API documentation generated from your Simbody source. You will get some warnings if your doxygen version is earlier than Doxygen 1.8.8; upgrade if you can.

6. Install Simbody by right-clicking INSTALL and selecting Build.

Play around with examples

Within your build in Visual Studio (not the installation):

1. Make sure your configuration is set to a release configuration (e.g., RelWithDebInfo).
2. Right click on one of the targets whose name begins with Example - and select Select as Startup Project.
3. Type Ctrl-F5 to start the program.

Set environment variables and test the installation

If you are only building Simbody to use it with OpenSim, you can skip this section.

1. Allow executables to find Simbody libraries (.dll's) by adding the Simbody bin/ directory to your PATH environment variable.
   1. In the Start menu (Windows 7) or screen (Windows 8), search environment.
   2. Select Edit the system environment variables.
   3. Click Environment Variables....
   4. Under System variables, click Path, then click Edit.
   5. Add C:/Simbody/bin; to the front of the text field. Don't forget the semicolon!
2. Allow Simbody and other projects (e.g., OpenSim) to find Simbody. In the same Environment Variables window:
   1. Under User variables for..., click New....
   2. For Variable name, type SIMBODY_HOME.
   3. For Variable value, type C:/Simbody.
3. Changes only take effect in newly-opened windows. Close any Windows Explorer or Command Prompt windows.
4. Test your installation by navigating to C:/Simbody/examples/bin and running SimbodyInstallTest.exe or SimbodyInstallTestNoViz.exe.

Note: Example binaries are not installed for Debug configurations. They are present in the build environment, however, so you can run them from there. They will run very slowly!

Layout of installation

How is your Simbody installation organized?

• bin/ the visualizer and shared libraries (.dll's, used at runtime).
• doc/ a few manuals, as well as API docs (SimbodyAPI.html).
• examples/
  • src/ the source code for the examples.
  • bin/ the examples, compiled into executables; run them! (Not installed for Debug builds.)
• include/ the header (.h) files; necessary for projects that use Simbody.
• lib/ "import" libraries, used during linking.
• cmake/ CMake files that are useful for projects that use Simbody.

Linux or Mac using make

These instructions are for building Simbody from source on either a Mac or on Ubuntu.

Get dependencies

On a Mac, the Xcode developer package gives LAPACK and BLAS to you via the Accelerate framework. Mac's come with the visualization dependencies.

On Ubuntu, we need to get the dependencies ourselves. Open a terminal and run the following commands.

1. Get the necessary dependencies: $ sudo apt-get install cmake liblapack-dev. The cmake on Ubuntu 12.04 is not new enough; you could instead download it from cmake.org or use this third party PPA.
2. If you want to use the CMake GUI, install cmake-qt-gui.
3. For visualization (optional): $ sudo apt-get install freeglut3-dev libxi-dev libxmu-dev
4. For API documentation (optional): $ sudo apt-get install doxygen

Get the Simbody source code

There are two ways to get the source code.

• Method 1: Download the source code from https://github.com/simbody/simbody/releases. Look for the highest-numbered release, click on the .zip button, and unzip it on your computer. We'll assume you unzipped the source code into ~/simbody-source.
• Method 2: Clone the git repository.

  1. Get git.

     • Mac: You might have it already, especially if you have Xcode, which is free in the App Store. If not, one method is to install Homebrew and run brew install git in a terminal.
     • Ubuntu: run sudo apt-get install git in a terminal.

  2. Clone the github repository into ~/simbody-source.

      $ git clone https://github.com/simbody/simbody.git ~/simbody-source
      $ git checkout Simbody-3.5.1
     

  3. In the last line above, we assumed you want to build a released version. Feel free to change the version you want to build. If you want to build the latest development version ("bleeding edge") of Simbody off the master branch, you can omit the checkout line.

Configure and generate Makefiles

1. Create a directory in which we'll build Simbody. We'll assume you choose ~/simbody-build. Don't choose a location inside ~/simbody-source.

    $ mkdir ~/simbody-build
    $ cd ~/simbody-build
   

2. Configure your Simbody build with CMake. We'll use the cmake command but you could also use the interactive tools ccmake or cmake-gui. You have a few configuration options to play with here.

   • If you don't want to fuss with any options, run:

       $ cmake ~/simbody-source
     

   • Where do you want to install Simbody? By default, it is installed to /usr/local/. That's a great default option, especially if you think you'll only use one version of Simbody at a time. You can change this via the CMAKE_INSTALL_PREFIX variable. Let's choose ~/simbody:

       $ cmake ~/simbody-source -DCMAKE_INSTALL_PREFIX=~/simbody
     

   • Do you want the libraries to be optimized for speed, or to contain debugger symbols? You can change this via the CMAKE_BUILD_TYPE variable. There are 4 options:

     • Debug: debugger symbols; no optimizations (more than 10x slower). Library and visualizer names end with _d.
     • RelWithDebInfo: debugger symbols; optimized. This is the configuration we recommend.
     • Release: no debugger symbols; optimized. Generated libraries and executables are smaller but not faster than RelWithDebInfo.
     • MinSizeRel: minimum size; optimized. May be slower than RelWithDebInfo or Release.

     You at least want optimized libraries (all configurations but Debug are optimized), but you can have Debug libraries coexist with them. To do this, go through the full installation process twice, once for each configuration. It is typical to use a different build directory for each build type (e.g., ~/simbody-build-debug and ~/simbody-build-release).

   • There are a few other variables you might want to play with:

     • BUILD_EXAMPLES to see what Simbody can do. On by default.
     • BUILD_TESTING to ensure your Simbody works correctly. On by default.
     • BUILD_VISUALIZER to be able to watch your system move about! If building on a cluster, you could turn this off. On by default.
     • BUILD_STATIC_LIBRARIES builds the three libraries as static libraries, whose names will end with _static.
     • BUILD_TESTS_AND_EXAMPLES_STATIC if tests or examples are being built, creates statically-linked tests/examples. Can take a while to build, and it is unlikely you'll use the statically-linked libraries.
     • BUILD_TESTS_AND_EXAMPLES_SHARED if tests or examples are being built, creates dynamically-linked tests/examples. Unless you know what you're doing, leave this one on.

     You can combine all these options. Here's another example:

       $ cmake ~/simbody-source -DCMAKE_INSTALL_PREFIX=~/simbody -DCMAKE_BUILD_TYPE=RelWithDebInfo -DBUILD_VISUALIZER=off
     

Build and install

1. Build the API documentation. This is optional, and you can only do this if you have Doxygen. You will get warnings if your doxygen installation is a version older than Doxygen 1.8.8.

    $ make doxygen
   

2. Compile. Use the -jn flag to build using n processor cores. For example:

    $ make -j8
   

3. Run the tests.

    $ ctest -j8
   

4. Install. If you chose CMAKE_INSTALL_PREFIX to be a location which requires sudo access to write to (like /usr/local/, prepend this command with a sudo .

    $ make -j8 install
   

Just so you know, you can also uninstall (delete all files that CMake placed into CMAKE_INSTALL_PREFIX) if you're in ~/simbody-build.

$ make uninstall

Play around with examples

From your build directory, you can run Simbody's example programs. For instance, try:

    $ ./ExamplePendulum

Set environment variables and test the installation

If you are only building Simbody to use it with OpenSim, you can skip this section.

1. Allow executables to find Simbody libraries (.dylib's or so's) by adding the Simbody lib directory to your linker path.

   • If your CMAKE_INSTALL_PREFIX is /usr/local/, run:

       $ sudo ldconfig
     

   • If your CMAKE_INSTALL_PREFIX is neither /usr/ nor /usr/local/ (e.g., ~/simbody'):

     • Mac:

         $ echo 'export DYLD_LIBRARY_PATH=$DYLD_LIBRARY_PATH:~/simbody/lib' >> ~/.bash_profile
       

     • Ubuntu:

         $ echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:~/simbody/lib/x86_64-linux-gnu' >> ~/.bashrc
       

     These commands add a line to a configuration file that is loaded every time you open a new terminal. If using Ubuntu, you may need to replace x86_64-linux-gnu with the appropriate directory on your computer.

2. Allow Simbody and other projects (e.g., OpenSim) to find Simbody. Make sure to replace ~/simbody with your CMAKE_INSTALL_PREFIX.

   • Mac:

       $ echo 'export SIMBODY_HOME=~/simbody' >> ~/.bash_profile
     

   • Ubuntu:

       $ echo 'export SIMBODY_HOME=~/simbody' >> ~/.bashrc
     

3. Open a new terminal.

4. Test your installation:

    $ cd ~/simbody/share/doc/simbody/examples/bin
    $ ./SimbodyInstallTest # or ./SimbodyInstallTestNoViz
   

Layout of installation

The installation creates the following directories in CMAKE_INSTALL_PREFIX. The directory [x86_64-linux-gnu] only exists if you did NOT install to /usr/local/ and varies by platform. Even in that case, the name of your directory may be different.

• include/simbody/ the header (.h) files; necessary for projects that use Simbody.
• lib/[x86_64-linux-gnu]/ shared libraries (.dylib's or .so's).
  • cmake/simbody/ CMake files that are useful for projects that use Simbody.
  • pkgconfig/ pkg-config files useful for projects that use Simbody.
  • simbody/examples/ the examples, compiled into executables; run them! (Not installed for Debug builds.)
• libexec/simbody/ the simbody-visualizer executable.
• share/doc/simbody/ a few manuals, as well as API docs (SimbodyAPI.html).
  • examples/src source code for the examples.
  • examples/bin symbolic link to the runnable examples.

Mac and Homebrew

If using a Mac and Homebrew, the dependencies are taken care of for you.

Install

1. Install Homebrew.

2. Open a terminal.

3. Add the Open Source Robotics Foundation's list of repositories to Homebrew:

   $ brew tap osrf/simulation
   

4. Install the latest release of Simbody.

   $ brew install simbody
   

   To install from the master branch instead, append --HEAD to the command above.

Where is Simbody installed?

Simbody is now installed to /usr/local/Cellar/simbody/<version>/, where <version> is either the version number (e.g., 3.5.1), or HEAD if you specified --HEAD above.

Some directories are symlinked (symbolically linked) to /usr/local/, which is where your system typically expects to find executables, shared libraries (.dylib's), headers (.h's), etc. The following directories from the Simbody installation are symlinked:

• include/simbody -> /usr/local/include/simbody
• lib -> /usr/local/lib
• share/doc/simbody -> /usr/local/share/doc/simbody

Layout of installation

What's in the /usr/local/Cellar/simbody/<version>/ directory?

• include/simbody/ the header (.h) files; necessary for projects that use Simbody.
• lib/ shared libraries (.dylib's), used at runtime.
  • cmake/simbody/ CMake files that are useful for projects that use Simbody.
  • pkgconfig/ pkg-config files useful for projects that use Simbody.
  • simbody/examples/ the examples, compiled into executables; run them! (Not installed for Debug builds.)
• libexec/simbody/ the simbody-visualizer executable.
• share/doc/simbody/ a few manuals, as well as API docs (SimbodyAPI.html).
  • examples/src source code for the examples.
  • examples/bin symbolic link to executable examples.

Ubuntu and apt-get

Starting with Ubuntu 15.04, Simbody is available in the Ubuntu (and Debian) repositories. You can see a list of all simbody packages for all Ubuntu versions at the Ubuntu Packages website. The latest version of Simbody is usually not available in the Ubuntu repositories; the process for getting a new version of Simbody into the Ubuntu repositories could take up to a year.

Install

1. Open a terminal and run the following command:

    $ sudo apt-get install libsimbody-dev libsimbody-doc
   

Layout of installation

Simbody is installed into the usr/ directory. The directory [x86_64-linux-gnu] varies by platform.

• usr/include/simbody/ the header (.h) files; necessary for projects that use Simbody.
• usr/lib/[x86_64-linux-gnu] shared libraries (.so's).
  • cmake/simbody/ CMake files that are useful for projects that use Simbody.
  • pkgconfig/ pkg-config files useful for projects that use Simbody.
• usr/libexec/simbody/ the simbody-visualizer executable.
• usr/share/doc/simbody/ a few manuals, as well as API docs (SimbodyAPI.html).
  • examples/src source code for the examples.
  • examples/bin symbolic link to executable examples.

Windows using MinGW

Warning: The MinGW generation and build is experimental!

This build is still experimental, because of :

• the various MinGW versions available (Thread model, exception mechanism)
• the compiled libraries Simbody depends on (Blas, Lapack and optionnaly glut).

Below are three sections that gives a list of supported versions, command line instructions, and reasons why is it not so obvious to use MinGW.

Supported MinGW versions

If you do not want to go into details, you need a MinGW version with :

• a Posix thread model and Dwarf exception mechanism on a 32 bit computer
• a Posix thread model and SJLJ exception mechanism on a 64 bit computer

Other versions are supported with additional configurations.

The table below lists the various versions of MinGW versions tested:

| OS | Thread | Exception | Comment | URL | ---|---------|--------|-----------|---------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| 1 | 64 Bits | Posix | SJLJ | All features supported, all binary included (Recommended version) | MinGW64 project GCC 5.2.0 | 2 | 64 Bits | Posix | SEH | Needs to be linked against user's Blas and Lapack | MinGW64 project GCC 5.2.0 | 3 | 32 Bits | Posix | Dwarf | No visualization, all binary included | MinGW64 project GCC 5.2.0 | 4 | 32 Bits | Posix | SJLJ | No visualization, needs to be linked against user's Blas and Lapack | MinGW64 project GCC 5.2.0 |

We recommend to use the first configuration where all features are supported and does not need additional libraries to compile and run. The URL allows to download directly this version. The second version needs to be linked against user's Blas and Lapack (A CLI example is given below). Blas and Lapack sources can be downloaded from netlib. For the 3rd and 4th versions that run that target a 32 bit behaviour, visualization is not possible for the time being. (It is due to a compile and link problem with glut). Moreover for the 4th one, one needs to provide Blas and Lapack libraries.

Please note that only Posix version of MinGW are supported.

If your version is not supported, CMake will detect it while configuring and stops.

Instructions

Below are some examples of command line instructions for various cases. It is assumed you are running commands from a build directory, that can access Simbody source with a command cd ..\simbody.

It is recommended to specify with the installation directory with flag CMAKE_INSTALL_PREFIX (e.g. -DCMAKE_INSTALL_PREFIX="C:\Program Files\Simbody"). If not used, the installation directory will be C:\Program Files (x86)\Simbody on a 64 bit computer. This might be confusing since it is the 32 bit installation location.

Example of instructions where one uses Blas and Lapack libraries provided (to be used in a Windows terminal, where MinGW is in the PATH):

rem CMake configuration
cmake ..\simbody -G "MinGW Makefiles" -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX="C:\Program Files\Simbody"
rem Compilation
mingw32-make
rem Test
mingw32-make test
rem Installation
mingw32-make install

Example of instructions where one uses Blas and Lapack libraries provided (to be used in a Windows terminal, where MinGW is NOT in the PATH):

rem Variable and path definition
set CMAKE="C:\Program Files\CMake\bin\cmake.exe"
set MinGWDir=C:\Program Files\mingw-w64\i686-5.2.0-posix-sjlj-rt_v4-rev0\mingw32
set PATH=%MinGWDir%\bin;%MinGWDir%\i686-w64-mingw32\lib
rem CMake configuration
%CMAKE% ..\simbody -G"MinGW Makefiles" -DCMAKE_BUILD_TYPE=Release ^
 -DCMAKE_INSTALL_PREFIX="C:\Program Files\Simbody" ^
 -DCMAKE_C_COMPILER:PATH="%MinGWDir%\bin\gcc.exe" ^
 -DCMAKE_CXX_COMPILER:PATH="%MinGWDir%\bin\g++.exe" ^
 -DCMAKE_MAKE_PROGRAM:PATH="%MinGWDir%\bin\mingw32-make.exe"
rem Compilation
mingw32-make
rem Test
mingw32-make test
rem Installation
mingw32-make install

Example of instructions where one uses Blas and Lapack libraries provided (to be used in a MSYS terminal with MinGW in the PATH):

# CMake configuration
cmake ../simbody -G "MSYS Makefiles" -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX="C:\Program Files\Simbody"
# Compilation
make
# Test
make test
# Installation
make install

Example of instructions where one provides our own Blas and Lapack libraries (to be used in a MSYS terminal with MinGW in the PATH):

# CMake configuration
cmake ../simbody -G"MSYS Makefiles" -DCMAKE_BUILD_TYPE=Release \
-DCMAKE_INSTALL_PREFIX="C:\Program Files\Simbody" \
-DCMAKE_C_COMPILER:PATH="C:\Program Files\mingw-w64\i686-5.2.0-posix-sjlj-rt_v4-rev0\mingw32\bin\gcc.exe" \
-DCMAKE_CXX_COMPILER:PATH="C:\Program Files\mingw-w64\i686-5.2.0-posix-sjlj-rt_v4-rev0\mingw32\bin\g++.exe" \
-DBUILD_USING_OTHER_LAPACK:PATH="C:\Program Files\lapack-3.5.0\bin\liblapack.dll;C:\Program Files\lapack-3.5.0\bin\libblas.dll"
make
# Test
make test
# Installation
make install

MinGW details

This paragraph explains the reason why one can not use any MinGW version.

MinGW is available with two thread models :

• Win32 thread model
• Posix thread model

One has to use the Posix thread model, since all thread functionalities (e.g. std:mutex) are not implemented.

To ease building on Windows, Simbody provides compiled libraries for Blas and Lapack :

• On Windows 32 Bits, these were compiled with a Dwarf exception mechanism,
• On Windows 64 Bits, these were compiled with a SJLJ exception mechanism.

If one chooses a MinGW compilation, we need to respect this exception mechanism. A program can not rely on both mechanisms. This means that if we want to use the compiled libraries, our MinGW installation should have the same exception mechanism. Otherwise, we need to provide our own Blas and Lapack libraries.

To see which exception mechanism is used, user can look at dlls located in the bin directory of MinGW. The name of mechanism is present in the file libgcc_XXXX.dll, where XXXX can be dw, seh or sjlj. For some MinGW versions, this information is also available by looking at the result of gcc --version.

CMake will check the version of your MinGW, and if the exception mechanism is different, then the configuration stops because of this difference. If one provides Blas and Lapack libraries with the CMake variable BUILD_USING_OTHER_LAPACK, compilation with MinGW is always possible.

Acknowledgments

We are grateful for past and continuing support for Simbody's development in Stanford's Bioengineering department through the following grants:

• NIH U54 GM072970 (Simulation of Biological Structures)
• NIH U54 EB020405 (Mobilize Center)
• NIH R24 HD065690 (Simulation in Rehabilitation Research)
• OSRF subcontract 12-006 to DARPA HR0011-12-C-0111 (Robotics Challenge)

Prof. Scott Delp is the Principal Investigator on these grants and Simbody is used extensively in Scott's Neuromuscular Biomechanics Lab as the basis for the OpenSim biomechanical simulation software application for medical research.