Warning
This documentation is for an old version of Cantera. You can find docs for newer versions here.
This guide contains instructions for compiling Cantera on the following operating systems:
In addition to the above operating systems, Cantera should work on any Unix-like system where the necessary prerequisites are available, but some additional configuration may be required.
For Ubuntu or Debian users, the following packages should be installed using your choice of package manager:
g++ python scons libboost-all-dev libsundials-serial-dev
Building the python module also requires:
cython python-dev python-numpy python-numpy-dev
Checking out the source code from version control requires Git (install git).
The minimum compatible Cython version is 0.19. If your distribution does not contain a suitable version, you may be able to install a more recent version using easy_install or pip.
Building the Fortran interface also requires gfortran or another supported Fortran compiler.
Users of other distributions should install the equivalent packages, which may have slightly different names.
There are a number of requirements for the versions of software to install depending on which interfaces (Python, Matlab) you want to build and what architecture (32-bit or 64-bit) you want to use. See Software used by Cantera for the full list of dependencies.
The build process will produce a Python module compatible with the version of Python used for the compilation. To generate different modules for other versions of Python, you will need to install those versions of Python and recompile.
If you want to build the Matlab toolbox and you have a 64-bit copy of Windows, by default you will be using a 64-bit copy of Matlab, and therefore you need to compile Cantera in 64-bit mode. For simplicity, it is highly recommended that you use a 64-bit version of Python to handle this automatically.
There is no 64-bit installer for SCons under Windows, so you will need to download the ZIP version. After extracting it, start a command prompt in the unzipped folder and run:
python setup.py install
It is generally helpful to have SCons and Python in your PATH. This can usually be accomplished by adding the top-level Python directory (e.g. C:\Python27) to your PATH. This is accessible from:
Control Panel > System and Security > System > Advanced System Settings > Environment Variables
In order to use SCons to install Cantera to a system folder (e.g. C:\Program Files\Cantera) you must run the scons install command in a command prompt that has been launched by selecting the run as administrator option.
Download and install Xcode from the App Store
From a Terminal, run:
sudo xcode-select --install
and agree to the Xcode license agreement
If you don’t have numpy version >= 1.4, you can install a recent version with:
sudo easy_install -U numpy
If you want to build Cantera with Fortran 90 support, download gfortran from:
http://gcc.gnu.org/wiki/GFortranBinaries#MacOS
Download scons-2.x.y.tar.gz from scons.org and extract the contents. Install with either:
sudo python setup.py install
to install for all users, or:
python setup.py install --user
to install to a location in your home directory.
Option 1: Download the most recent source tarball from SourceForge and extract the contents.
Option 2: Check out the code using Git:
git clone https://github.com/Cantera/cantera.git
cd cantera
git checkout 2.2
Check out the code using Git:
git clone https://github.com/Cantera/cantera.git
run scons help to see a list all configuration options for Cantera, or see Configuring Cantera.
Configuration options are specified as additional arguments to the scons command, e.g.:
scons build -j4 blas_lapack_libs=lapack,blas
If the prerequisites are installed in standard locations, the default values should work.
If you installed Sundials to a non-standard location (e.g. the libraries aren’t in /usr/lib), you will need to specify the options:
sundials_include=/path/to/sundials/include
sundials_libdir=/path/to/sundials/lib
If you want to build the Matlab toolbox, you will need to specify the path to the Matlab installation, e.g.:
matlab_path=/opt/MATLAB/R2011a
matlab_path="C:\Program Files\MATLAB\R2011a"
matlab_path=/Applications/MATLAB_R2011a.app
The above paths are typical defaults on Linux, Windows, and OS X, respectively.
SCons saves configuration options specified on the command line in the file cantera.conf in the root directory of the source tree, so generally it is not necessary to respecify configuration options when rebuilding Cantera. To unset a previously set configuration option, either remove the corresponding line from cantera.conf or use the syntax:
option_name=
Sometimes, changes in your environment can cause SCons’s configuration tests (e.g. checking for libraries or compiler capabilities) to unexpectedly fail. To force SCons to re-run these tests rather than trusting the cached results, run scons with the option --config=force.
Cantera 2.1 introduces a new Python module implemented using Cython. This new module provides support for both Python 2.x and Python 3.x. It also features a redesigned API that simplifies many operations and aims to provide a more “Pythonic” interface to Cantera.
Building the new Python module requires the Cython package for Python.
The Cython module is compatible with the following Python versions: 2.6, 2.7, 3.1, 3.2, and 3.3. Support for Python 2.6 and Python 3.1 requires the scipy and unittest2 packages to be installed as well (see Software used by Cantera) to provide certain features that are included in the standard library in more recent versions.
By default, SCons will attempt to build the Cython-based Python module for Python 2, if both Numpy and Cython are installed.
If SCons detects a Python 3 interpreter installed in a default location (i.e. python3 is on the path), it will try to build the new Python module for Python 3. The following SCons options control how the Python 3 module is built:
python3_package=[y|n]
python3_cmd=/path/to/python3/interpreter
python3_array_home=/path/to/numpy
python3_prefix=/path/to/cantera/module
Note that even when building the Python 3 Cantera module, you should still use Python 2 with SCons, as SCons does not currently support Python 3.
Note
The cantera.conf file uses the backslash character \ as an escape character. When modifying this file, backslashes in paths need to be escaped like this: boost_inc_dir = 'C:\\Program Files (x86)\\boost\\include' This does not apply to paths specified on the command line. Alternatively, you can use forward slashes in paths.
To compile with MinGW, use the SCons command line option:
toolchain=mingw
The version of MinGW from http://www.mingw.org is 32-bit only, and therefore cannot be used to build a 64-bit Python module. Versions of MinGW that provide a 64-bit compiler are available from http://mingw-w64.sourceforge.net/ .
Before compiling Cantera, you may need to set up the appropriate environment variables for the Intel compiler suite, e.g.:
source /opt/intel/bin/compilervars.sh intel64
For the Intel compiler to work with SCons, these environment variables need to be passed through SCons by using the command line option:
env_vars=all
If you want to use the Intel MKL versions of BLAS and LAPACK, you will need to provide additional options. The following are typically correct on 64-bit Linux systems:
blas_lapack_libs=mkl_rt blas_lapack_dir=$(MKLROOT)/lib/intel64
Your final SCons call might then look something like:
scons build env_vars=all CC=icc CXX=icpc F90=ifort F77=ifort blas_lapack_libs=mkl_rt blas_lapack_dir=$(MKLROOT)/lib/intel64
When installing Cantera after building with the Intel compiler, the normal method of using sudo to install Cantera will not work because sudo does not pass the environment variables needed by the Intel compiler. Instead, you will need to do something like:
scons build ...
sudo -s
source /path/to/compilervars.sh intel64
scons install
exit
Run scons with the list of desired configuration options, e.g.:
scons build optimize=n blas_lapack_libs=blas,lapack prefix=/opt/cantera
If Cantera compiles successfully, you should see a message that looks like:
*******************************************************
Compilation completed successfully.
- To run the test suite, type 'scons test'.
- To install, type '[sudo] scons install'.
*******************************************************
If you do not see this message, check the output for errors to see what went wrong.
Cantera has a series of tests that can be run with the command:
scons test
When the tests finish, you should see a summary indicating the number of tests that passed and failed.
If you have tests that fail, try looking at the following to determine the source of the error:
- Messages printed to the console while running scons test
- Output files generated by the tests
To build the Cantera HTML documentation, run the commands:
scons doxygen
scons sphinx
or append the options sphinx_docs=y and doxygen_docs=y to the build command, e.g.:
scons build doxygen_docs=y sphinx_docs=y
If you get a compiler error while compiling some of the “f2c” code, then your version of MinGW has a problem with the order of its internal include paths, such that it sees the GCC float.h before its own special version. To fix this problem edit the GCC float.h located at (roughly):
c:\MinGW\lib\gcc\mingw32\4.6.1\include\float.h
and add the following just before the end (before the final #endif)
#ifndef _MINGW_FLOAT_H_
#include_next <float.h>
#endif
This section lists the versions of third-party software that are required to build and use Cantera.
You must have one of the following C++ compilers installed on your system. A Fortran compiler is required only if you plan to use Cantera from a Fortran program.
Numpy
Unittest2
Matlab
Sundials
Required to enable some features such as sensitivity analysis.
Strongly recommended if using reactor network or 1D simulation capabilities.
https://computation.llnl.gov/casc/sundials/download/download.html
Known to work with versions 2.4, 2.5 and 2.6.
To use Sundials with Cantera on a Linux/Unix system, it must be compiled with the -fPIC flag. You can specify this flag when configuring Sundials (2.4 or 2.5):
configure --with-cflags=-fPIC
or Sundials 2.6:
cmake -DCMAKE_C_FLAGS=-fPIC <other command-line options>
Note
If you are compiling Sundials 2.5.0 on Windows using CMake, you need to edit the CMakeLists.txt file first and change the lines:
SET(PACKAGE_STRING "SUNDIALS 2.4.0")
SET(PACKAGE_VERSION "2.4.0")
to read:
SET(PACKAGE_STRING "SUNDIALS 2.5.0")
SET(PACKAGE_VERSION "2.5.0")
instead, so that Cantera can correctly identify the version of Sundials.
Windows Installer XML (WiX) toolset
Distribute (Python)
Packages required for building Sphinx documentation