prerequisites

C++ compiler with support for the C++17 standard, or a more recent standard. This includes the following compilers:
- GNU C++, version 7.0 or higher
- Clang, version 5 or higher
- Apple Clang, version 9.3 or higher
- Intel C++ compiler, version 19 or higher
CMake, version 3.10 or higher; if CUDA support is needed, CMake 3.17 or higher is required.
Git 1.8 or later (required to obtain TiledArray and MADNESS source code from GitHub)
Boost version 1.66 or later; installed, if not found
PaRSEC or MADNESS runtimes (see below); installed, if not found

build

cmake <cmake args>
cmake --build . --target <test or example you want (see below)>

useful cmake command-line arguments:

CMAKE_CXX_COMPILER, e.g. -DCMAKE_CXX_COMPILER=clang++ to use clang
BOOST_ROOT, e.g. -DBOOST_ROOT=path (to ensure that path is correct look for path/boost/version.hpp exists)

MADNESS runtime notes

The following tests/examples are supported with MADNESS

test-mad
t9-mad
serialization

PaRSEC runtime notes

currently PaRSEC support is being retrofitted to be able to use the latest PaRSEC.

spmm-* notes

To compile (block-)sparse SUMMA must:

obtain and install Eigen (header-only) library and pass the path to Eigen source to CMake as -DEIGEN3_INCLUDE_DIR=<path to Eigen>
block-sparse only obtain and install BTAS header-only tensor library from https://github.com/BTAS/BTAS ; clone and pass the path to BTAS source to CMake as -DBTAS_INSTALL_DIR=<path to BTAS>
to build element-sparse SUMMA example: make spmm-mad
to build block-sparse SUMMA example: make bspmm-mad