# DDαAMG
----------------------

NOTE : This version of DDαAMG contains some critical parts ported to CUDA (MPI, OpenMP and SSE vectorization are still present throughout). For more information on these improvements, contact : g.ramirez@math.uni-wuppertal.de

RECOMMENDED USAGE (as used on JUWELS @ JSC):

* modify compile_modules.sh, commenting/uncommenting as needed
* modify MakeSettings.mk as needed
* if not done yet, to into dependencies/ and run install_dependencies_01.sh
* >$ . setup_env_and_compile.sh
* take sample_E250_xPU.ini and submit_E250_xPU.job as the default example, in combination with sample_devel.ini
* >$ sbatch submit_E250_xPU.job

To use this version, run the script (note the dot at the beginning):

>$ . setup_env_and_compile.sh

----------------------

The DDalphaAMG solver library is an inverter for Wilson-Clover fermions from lattice QCD.

## Install

The root directory contains a `Makefile` and `MakeSettings.mk.template`.
Create a copy of `MakeSettings.mk.template` and name it `MakeSettings.mk`
and adjust compiler and library paths in the newly created file.
`MakeSettings.mk` contains all user-adjustable settings and should be
self-documenting. `Makefile` is not intended to be changed by the
user/distributor.

```make -j8 dd_alpha_amg``` compiles the standalone dd_alpha_amg executable.
The Makefile contains additional rules "library" and
"documentation" for compiling the code as a library (not yet supported for
the CUDA version) and for compiling the user documentation. The library
interface can be found in the "include" folder after compiling the library.

## Documentation

After having compiled the user documentation via
`make documentation` please consult the compiled PDF in the directory doc
and the doxygen documentation in doc/doxygen.