Binary packages are currently not provided! Therefore, you need to build the Spirit core library or the desktop user interface yourself.
The Spirit framework is designed to run across different
platforms and uses CMake
for its build process, which will
generate the appropriate build scripts for each platform.
System-wide installation is not actively supported.
While you can call make install
after building,
you may not achieve the desired results.
Requirements
- cmake >= 3.2
- compiler with C++11 support, e.g. gcc >= 5.1
Build
CMake is used to automatically generate makefiles.
# enter the top-level Spirit directory
$ cd spirit
# make a build directory and enter that
$ mkdir build
$ cd build
# Generate makefiles
$ cmake ..
# Build
$ make
Note that you can use the -j
option of make to run the
build in parallel.
To manually specify the build type (default is 'Release'),
call cmake --build . --config Release
instead of make
.
By default, the desktop GUI will try to build. The corresponding
CMake option is SPIRIT_UI_CXX_USE_QT
.
- Qt >= 5.7 (including qt-charts)
- OpenGL drivers >= 3.3
Necessary OpenGL drivers should be available through the regular drivers for any remotely modern graphics card.
The Python package is built by default. The corresponding
CMake option is SPIRIT_BUILD_FOR_PYTHON
.
The package is then located at core/python
. You can then
- make it locatable, e.g. by adding
path/to/spirit/core/python
to yourPYTHONPATH
cd core/python
andpip install -e . --user
to install it
Alternatively, the most recent release version can be
installed from the official package,
e.g. pip install spirit --user
.
The OpenMP backend can be used to speed up calculations by using a multicore CPU.
At least version 4.5 of OpenMP needs to be supported by your compiler.
Build
You need to set the corresponding CMake variable, e.g. by calling
cd build
cmake -DSPIRIT_USE_OPENMP=ON ..
cd ..
The CUDA backend can be used to speed up calculations by using a GPU.
Spirit uses unified memory. At least version 8 of the CUDA toolkit is required and the GPU needs compute capability 3.0 or higher!
If the GUI is used, compute capability 6.0 or higher is required! (see the CUDA programming guide: coherency)
Note that the GUI cannot be used on the CUDA backend on OSX! (see the CUDA programming guide: coherency and requirements)
Note: the precision of the core will be automatically set
to float
in order to avoid the performance cost of double
precision operations on GPUs.
Build
You need to set the corresponding SPIRIT_USE_CUDA
CMake
variable, e.g. by calling
cd build
cmake -DSPIRIT_USE_CUDA=ON ..
cd ..
You may additionally need to
- pass the
CUDA_TOOLKIT_ROOT_DIR
to cmake or edit it in the root CMakeLists.txt - select the appropriate arch for your GPU using the
SPIRIT_CUDA_ARCH
CMake variable
Using emscripten, Spirit can be built as a Web assembly library, meaning that it can be used e.g. from within JavaScript.
Build
The CMake option you need to set to ON
is called
SPIRIT_BUILD_FOR_JS
.
You need to have emscripten available, meaning you might
need to source, e.g. source /usr/local/bin/emsdkvars.sh
.
Then to build, call
cd build
cmake .. -DCMAKE_TOOLCHAIN_FILE=/usr/local/emsdk/emscripten/1.38.29/cmake/Modules/Platform/Emscripten.cmake
make
cd ..
More options than described above are available, allowing for example to deactivate building the Python library or the unit tests.
To list all available build options, call
cd build
cmake -LH ..
The build options of Spirit all start with SPIRIT_
.