SimGrid should work out of the box on Linux, macOS, FreeBSD, and Windows (under Windows, you need to install the Windows Subsystem Linux to get more than the Java bindings).
Binaries for Linux¶
To get all of SimGrid on Debian or Ubuntu, simply type one of the following lines, or several lines if you need several languages.
$ apt install libsimgrid-dev # if you want to develop in C or C++ $ apt install simgrid-java # if you want to develop in Java $ apt install python3-simgrid # if you want to develop in Python
If you build pre-compiled packages for other distributions, drop us an email.
Stable Java Package¶
The jar file can be retrieved from the Release page. This file is self-contained, including the native components for Linux, macOS and Windows. Copy it to your project’s classpath and you’re set.
Nightly built Java Package¶
For non-Windows systems (Linux, macOS, or FreeBSD), head to Jenkins. In the build history, pick the last green (or at least yellow) build that is not blinking (i.e., not currently under build). In the list, pick a system that is close to yours, and click on the ball in the Debug row. The build artifact will appear at the top of the resulting page.
For Windows, head to AppVeyor. Click on the artifact link on the right, and grab your file. If the latest build failed, there will be no artifact. Then you will need to first click on “History” at the top and look for the last successful build.
Binary Java Troubleshooting¶
Here are some error messages that you may get when trying to use the binary Java package.
- Your architecture is not supported by this jarfile
If your system is not supported, you should compile your own jarfile by compiling SimGrid from the source.
- Library not found: boost-context
You should obviously install the
boost-contextlibrary on your machine, for example with
Version numbering and deprecation¶
SimGrid tries to be both a research instrument that you can trust, and a vivid project targeting the future issues. We have 4 stable versions per year, numbered 3.24 or 3.25. Backward compatibility is ensured for one year: Code compiling without warning on 3.24 will still compile with 3.28, but maybe with some deprecation warnings. You should update your SimGrid installation at least once a year and fix those deprecation warnings: the compatibility wrappers are usually removed after 4 versions. Another approach is to never update your SimGrid installation, but we don’t provide any support to old versions.
Interim versions (also called pre-versions) may be released between stable releases. They are numbered 3.X.Y, with even Y (for example, 3.23.2 was released on July 8. 2019 as a pre-version of 3.24). These versions should be as usable as regular stable releases, even if they may be somewhat less tested and documented. They play no role in our deprecation handling, and they are not really announced to not spam our users.
Version numbered 3.X.Y with odd Y are git versions. They often work, but no guarantee is given whatsoever (all releases are given “as is”, but that’s even more so for these unreleased versions).
Installing from the Source¶
Getting the Dependencies¶
- C++ compiler (either g++, clang, or icc).
We use the C++14 standard, and older compilers tend to fail on us. It seems that g++ 5.0 or higher is required nowadays (because of boost). SimGrid compiles well with clang or icc too.
- Python 3.
SimGrid should build without Python. That is only needed by our regression test suite.
- cmake (v3.5).
ccmakeprovides a nicer graphical interface compared to
ccmakeif you need to see absolutely all configuration options (e.g., if your Python installation is not standard).
- boost (at least v1.48, v1.59 recommended)
On Debian / Ubuntu:
apt install libboost-dev libboost-context-dev
On macOS with homebrew:
brew install boost
- Java (optional):
Debian / Ubuntu:
apt install default-jdk libgcj18-dev(or any version of libgcj)
macOS or Windows: Grab a full JDK
- Lua (optional – must be v5.3)
SimGrid won’t work with any other version of Lua.
Debian / Ubuntu:
apt install liblua5.3-dev lua5.3
choco install lua53
- From the source
You need to patch the sources to build dynamic libraries. First download lua 5.3
Open the archive:
tar xvfz lua-5.3.*.tar.gz
Enter the directory:
Patch the sources:
patch -p1 < /path/to/simgrid/...../tools/lualib.patch
Build and install lua:
make linux && sudo make install
For platform-specific details, please see below.
Getting the Sources¶
Grab the last stable release from FramaGit, and compile it as follows:
$ tar xf simgrid-3-XX.tar.gz $ cd simgrid-* $ cmake -DCMAKE_INSTALL_PREFIX=/opt/simgrid . $ make $ make install
If you want to stay on the bleeding edge, get the current git version, and recompile it as with stable archives. You may need some extra dependencies.
$ git clone https://framagit.org/simgrid/simgrid.git $ cd simgrid $ cmake -DCMAKE_INSTALL_PREFIX=/opt/simgrid . $ make $ make install
This section is about compile-time options, which are very different from run-time options. Compile-time options fall into two categories. SimGrid-specific options define which part of the framework to compile while Generic options are provided by cmake itself.
Our build system often gets mixed up if you change something on
your machine after the build configuration. For example, if
SimGrid fails to detect your fortran compiler, it is not enough to
install a fortran compiler. You also need to delete all Cmake
files, such as
CMakeCache.txt. Since Cmake also generates some
files in the tree, you may need to wipe out your complete tree and
start with a fresh one when you install new dependencies.
Another (better) solution is to build out of the source tree.
Generic build-time options¶
These options specify, for example, the path to various system elements (Python
path, compiler to use, etc). In most case, CMake automatically discovers the
right value for these elements, but you can set them manually as needed.
Notably, such variables include
CXX, defining the paths to the C
and C++ compilers;
CXXFLAGS specifying extra options to pass
to the C and C++ compilers; and
PYTHON_EXECUTABLE specifying the path to the
The best way to discover the exact name of the option that you need to
change is to press
t in the
ccmake graphical interface, as all
options are shown (and documented) in the advanced mode.
Once you know their name, there are several ways to change the values of
build-time options. You can naturally use the ccmake graphical
interface for that, or you can use environment variables, or you can
-D flag of
For example, you can change the compilers by issuing these commands to set some environment variables before launching cmake:
$ export CC=gcc-5.1 $ export CXX=g++-5.1
The same can be done by passing
-D parameters to cmake, as follows.
Note that the dot at the end is mandatory (see Out of Tree Compilation).
$ cmake -DCC=clang -DCXX=clang++ .
SimGrid compilation options¶
Here is the list of all SimGrid-specific compile-time options (the default choice is in upper case).
- CMAKE_INSTALL_PREFIX (path)
Where to install SimGrid (/opt/simgrid, /usr/local, or elsewhere).
- enable_compile_optimizations (ON/off)
Ask the compiler to produce efficient code. You probably want to leave this option activated, unless you plan to modify SimGrid itself: efficient code takes more time to compile, and appears mangled to some debuggers.
- enable_compile_warnings (on/OFF)
Ask the compiler to issue error messages whenever the source code is not perfectly clean. If you are a SimGrid developer, you have to activate this option to enforce the code quality. As a regular user, this option is of little use.
- enable_debug (ON/off)
Disabling this option discards all log messages of severity debug or below at compile time (see Textual logging). The resulting code is marginaly faster than if you discard these messages at runtime, but it obviously becomes impossible to get any debug info from SimGrid when things go wrong.
- enable_documentation (on/OFF)
Generates the documentation pages. Building the documentation is not as easy as it used to be, and you should probably use the online version for now.
- enable_java (on/OFF)
Generates the java bindings of SimGrid. You must also enable MSG for this to work.
- enable_jedule (on/OFF)
Produces execution traces from SimDag simulations, which can then be visualized with the Jedule external tool.
- enable_lua (on/OFF)
Generate the lua bindings to the SimGrid internals (requires lua-5.3).
- enable_lib_in_jar (ON/off)
Embeds the native java bindings into the produced jar file.
- enable_lto (ON/off)
Enables the Link Time Optimization in the C++ compiler. This feature really speeds up the code produced, but it is fragile with older gcc versions.
- enable_maintainer_mode (on/OFF)
(dev only) Regenerates the XML parsers whenever the DTD is modified (requires flex and flexml).
- enable_mallocators (ON/off)
Activates our internal memory caching mechanism. This produces faster code, but it may fool the debuggers.
- enable_model-checking (on/OFF)
Activates the formal verification mode. This will hinder simulation speed even when the model checker is not activated at run time.
- enable_msg (on/OFF)
Activates the MSG legacy interface.
- enable_ns3 (on/OFF)
Activates the ns-3 bindings. See section ns-3 as a SimGrid model.
- enable_smpi (ON/off)
Allows one to run MPI code on top of SimGrid.
- enable_smpi_ISP_testsuite (on/OFF)
Adds many extra tests for the model checker module.
- enable_smpi_MPICH3_testsuite (on/OFF)
Adds many extra tests for the MPI module.
- minimal-bindings (on/OFF)
Take as few optional dependencies as possible, to get minimal library bindings in Java and Python.
- SMPI_C_FLAGS, SMPI_CXX_FLAGS, SMPI_Fortran_FLAGS (string)
Default compiler options to use in smpicc, smpicxx, or smpiff. This can be useful to set options like “-m32” or “-m64”.
Reset the build configuration¶
To empty the CMake cache (either when you add a new library or when
things go seriously wrong), simply delete your
may also want to directly edit this file in some circumstances.
Out of Tree Compilation¶
By default, the files produced during the compilation are placed in the source directory. It is however often better to put them all in a separate directory: cleaning the tree becomes as easy as removing this directory, and you can have several such directories to test several parameter sets or architectures.
For that, go to the directory where the files should be produced, and invoke cmake (or ccmake) with the full path to the SimGrid source as last argument.
$ mkdir build $ cd build $ cmake [options] .. $ make
Existing Compilation Targets¶
In most cases, compiling and installing SimGrid is enough:
$ make $ make install # try "sudo make install" if you don't have the permission to write
In addition, several compilation targets are provided in SimGrid. If
your system is well configured, the full list of targets is available
for completion when using the
Tab key. Note that some of the
existing targets are not really for public consumption so don’t worry
if some do not work for you.
make: Build the core of SimGrid that gets installed, but not any example.
make tests: Build the tests and examples.
make simgrid: Build only the SimGrid library. Not any example nor the helper tools.
make s4u-comm-pingpong: Build only this example (works for any example)
make java-all: Build all Java examples and their dependencies
make clean: Clean the results of a previous compilation
make install: Install the project (doc/ bin/ lib/ include/)
make uninstall: Uninstall the project (doc/ bin/ lib/ include/)
make dist: Build a distribution archive (tar.gz)
make distcheck: Check the dist (make + make dist + tests on the distribution)
make documentation: Create SimGrid documentation
If you want to see what is really happening, try adding
your compilation requests:
$ make VERBOSE=1
Testing your build¶
Once everything is built, you may want to test the result. SimGrid
comes with an extensive set of regression tests (as described in the
@ref inside_tests “insider manual”). The tests are not built by
default, so you first have to build them with
make tests. You can
then run them with
ctest, that comes with CMake. We run them
every commit and the results are on our Jenkins.
$ make tests # Build the tests $ ctest # Launch all tests $ ctest -R s4u # Launch only the tests whose names match the string "s4u" $ ctest -j4 # Launch all tests in parallel, at most 4 concurrent jobs $ ctest --verbose # Display all details on what's going on $ ctest --output-on-failure # Only get verbose for the tests that fail $ ctest -R s4u -j4 --output-on-failure # You changed S4U and want to check that you \ # didn't break anything, huh? \ # That's fine, I do so all the time myself.
SimGrid compiles like a charm with clang (version 3.0 or higher) on macOS:
$ cmake -DCMAKE_C_COMPILER=/path/to/clang -DCMAKE_CXX_COMPILER=/path/to/clang++ . $ make
Troubleshooting your macOS build.
- CMake Error: Parse error in cache file build_dir/CMakeCache.txt. Offending entry: /SDKs/MacOSX10.8.sdk
This was reported with the XCode version of clang 4.1. The work around is to edit the
CMakeCache.txtfile directly, to change the following entry:
You can safely ignore the warning about “-pthread” not being used, if it appears.
- /usr/include does not seem to exist
This directory does not exist by default on modern macOS versions, and you may need to create it with
The best solution to get SimGrid working on windows is to install the Ubuntu subsystem of Windows 10. All of SimGrid (but the model checker) works in this setting.
Native builds not very well supported. Have a look to our appveypor configuration file to see how we manage to use mingw-64 to build the DLL that the Java file needs.
The drawback of MinGW-64 is that the produced DLL are not compatible with MS Visual C. Some clang-based tools seem promising to fix this, but this is of rather low priority for us. It it’s important for you and if you get it working, please @ref community_contact “tell us”.
Once you have the Python development headers installed as well as a recent version of the pybind11 <https://pybind11.readthedocs.io/en/stable/> module (version at least 2.4), recompiling the Python bindings from the source should be as easy as:
# cd simgrid-source-tree $ python setup.py build install
Starting with SimGrid 3.13, it should even be possible to install simgrid without downloading the source with pip:
$ pip install simgrid
Once you have the full JDK installed, things should be as simple as:
$ cmake -Denable_java=ON -Dminimal-bindings=ON . $ make simgrid-java_jar # Only build the jarfile
After the compilation, the file
`simgrid.jar` is produced in the
Troubleshooting Java Builds
Sometimes, the build system fails to find the JNI headers. First locate them as follows:
$ locate jni.h /usr/lib/jvm/java-8-openjdk-amd64/include/jni.h /usr/lib/jvm/java-9-openjdk-amd64/include/jni.h /usr/lib/jvm/java-10-openjdk-amd64/include/jni.h
Then, set the JAVA_INCLUDE_PATH environment variable to the right
path, and relaunch cmake. If you have several versions of JNI installed
(as above), pick the one corresponding to the report of
$ export JAVA_INCLUDE_PATH=/usr/lib/jvm/java-8-openjdk-amd64/include/ $ cmake -Denable_java=ON . $ make
Note that the filename
`jni.h` was removed from the path.
Linux Multi-Arch specific instructions¶
On a multiarch x86_64 Linux, it should be possible to compile a 32-bit version of SimGrid with something like:
$ CFLAGS=-m32 \ CXXFLAGS=-m32 \ FFLAGS=-m32 \ PKG_CONFIG_LIBDIR=/usr/lib/i386-linux-gnu/pkgconfig/ \ cmake . \ -DCMAKE_SYSTEM_PROCESSOR=i386 \ -DCMAKE_Fortran_COMPILER=/some/path/to/i686-linux-gnu-gfortran \ -DGFORTRAN_EXE=/some/path/to/i686-linux-gnu-gfortran \ -DSMPI_C_FLAGS=-m32 \ -DSMPI_CXX_FLAGS=-m32 \ -DSMPI_Fortran_FLAGS=-m32
If needed, implement
i686-linux-gnu-gfortran as a script:
#!/usr/bin/env sh exec gfortran -m32 "$@"