Python wrapper for QUDA written in Cython.

This project aims to benifit from the optimized linear algebra library CuPy in Python based on CUDA. CuPy and QUDA will allow us to do most operation in lattice QCD research with high performance.

This project is based on the latest QUDA release v1.1.0.

Use pip install . to build and install the wrapper library, or use python3 setup.py build_ext --inplace to build the library in the repo folder and not install it. You need to build libquda.so and move it to the repo folder as a prerequisite.

This is an example to build QUDA for single GPU. The default install directory of QUDA should be /path/to/quda/build/usqcd.

git clone https://github.com/CLQCD/quda.git
pushd quda
git checkout b47950dd
mkdir build
pushd build
cmake .. -DQUDA_DIRAC_DOMAIN_WALL=OFF -DQUDA_CLOVER_DYNAMIC=OFF -DQUDA_CLOVER_RECONSTRUCT=OFF -DQUDA_DIRAC_NDEG_TWISTED_CLOVER=OFF -DQUDA_DIRAC_NDEG_TWISTED_MASS=OFF -DQUDA_DIRAC_TWISTED_CLOVER=OFF -DQUDA_DIRAC_TWISTED_MASS=OFF -DQUDA_INTERFACE_MILC=OFF -DQUDA_LAPLACE=ON -DQUDA_MULTIGRID=ON -DQUDA_LAPLACE=ON
cmake --build . -j12
cmake --install .
popd
popd

Install PyQuda with your Python interpreter. Here we copy libquda.so built previously to the root directory of PyQuda to simplify the environ.

export LD_LIBRARY_PATH=.:$LD_LIBRARY_PATH
git clone https://github.com/IHEP-LQCD/PyQuda.git
pushd PyQuda
python3 -m pip install -r requirements.txt
cp ../quda/build/lib/libquda.so ./
python3 -m pip install .
popd

chroma is needed here for generating the reference files used by test scripts.

export LD_LIBRARY_PATH=.:$LD_LIBRARY_PATH
pushd PyQuda
chroma -i tests/test.clover.ini.xml
python3 tests/test.clover.py
popd

I recommand you to build PyQuda using inplace mode instead of installing PyQuda.

export LD_LIBRARY_PATH=.:$LD_LIBRARY_PATH
git clone https://github.com/IHEP-LQCD/PyQuda.git
pushd PyQuda
python3 -m pip install -r requirements.txt
cp ../quda/build/lib/libquda.so ./
python3 setup.py build_ext --inplace
popd

Now you can modify Python files in the project and immediately get the result by running scripts. Adding the root directory to sys.path list is needed if you are running scripts from other directories.

first you should compile quda, and get libquda.so

then, compile mrhs_qcu and get libqcu.so

Put these libs into PyQuda/lib directory, then source env.sh, then install.

There are some test cases in PyQuda/tests

TODO