A2D - a PDE discretization library using almost automatic differentiation

A toolkit for almost automatic differentiation of vector and matrix expressions.

This code relies heavily on the approach for deriving auto-diff expressions by M. B. Giles, "Collected matrix derivative results for forward and reverse mode AD".

Dependencies

A2D is a header-only c++ templated library. The only requirement for using A2D is a C++17 supported compiler.

Install and use

CMake

CMake is preferred to install (copy over) A2D. For basic installation, use the following command:

mkdir build && cd build && cmake .. && make install

This installs A2D (headers and CMake files) into ${HOME}/installs/a2d. Then in the application, add

find_package(A2D REQUIRED PATHS <path-to-sparse-utils-installation>)

to CMakeLists.txt, and use

target_link_libraries(<app-target> A2D::A2D)

in the CMakeLists.txt for the application executables. See examples/CMakeLists.txt for example.

Manual

Alternatively, you can directly include include/a2dcore.h and manually manage the include path for the compiler.

Examples

Use the following to build examples:

cd examples &&
mkdir build &&
cd build &&
cmake .. &&
make -j

Test

Unit tests are implemented using Google Test framework, which is automatically downloaded when building tests. Use the following snippet to build and run unit tests.

mkdir build &&
cd build &&
cmake .. -DA2D_BUILD_TESTS=ON &&
make -j &&
ctest

Code style

clangFormat is used as the auto-formatter, with style --style=Google. If you would like to contribute to the project, please make sure you set up the auto-formatter accordingly.