OP2

OP2 is a high-level embedded domain specific language for writing unstructured mesh algorithms with automatic parallelisation on multi-core and many-core architectures. The API is embedded in both C/C++ and Fortran.

This repository contains the implementation of the code translation tools and run-time support libraries, and is structured as follows:

• op2: The C/C++ OP2 run-time libraries and Fortran bindings.
• translator: The Python code translators for both C/C++ and Fortran.
• apps: Example applications that demonstrate use of the API.
• makefiles: Shared infrastructure of the GNU Make based build-system.
• doc: LaTeX documentation source.

Documentation

Documentation is available on Read the Docs.

Quick-start

Firstly, OP2 has a varienty of toolchain dependencies that you will likely be able to obtain from your package manager or programming environment:

• GNU Make > 4.2
• A C/C++17 compatible compiler: Currently supported compilers are GCC, Clang, Cray, Intel, IBM XL and NVHPC.
• (Optional) A Fortran compiler: Currently supported compilers are GFortran, Cray, Intel, IBM XL and NVHPC.
• (Optional) An MPI implementation: Any implementation with the mpicc, mpicxx, and mpif90 wrappers is supported.
• (Optional) NVIDIA CUDA > 11.8

In addition there are a few optional library dependencies that you will likely have to build manually, although some package managers or programming environments may be able to provide appropriate versions:

• (Optional) (PT-)Scotch: Used for MPI mesh partitioning. Build both the sequential Scotch and parallel PT-Scotch.
• (Optional) ParMETIS: Used for MPI mesh partitioning. Build with 32-bit indicies (-DIDXSIZE32) and without -DSCOTCH_PTHREAD.
• (Optional) HDF5: Used for HDF5 I/O. You may build with and without --enable-parallel (depending on if you need MPI), and then specify both builds via the environment variables listed below.

Finally, to build OP2 and any of the apps:

1. Set either OP2_COMPILER={gnu, cray, intel, xl, nvhpc}, or OP2_{C, C_CUDA, F}_COMPILER={...} depending on your compiler setup. Alternatively if there is a profile specific to the cluster you are building on in makefiles/profiles you may use e.g. OP2_PROFILE=cirrus-intel.
2. (Optional) Set PTSCOTCH_INSTALL_PATH, PARMETIS_INSTALL_PATH, and HDF5_{SEQ, PAR}_INSTALL_PATH to the locations of the respective dependency builds containing include and lib folders. Certain build environments such as Spack, Nix and certain environment module implementations may already provided the required library directories through the environent or a compiler wrapper; if this is the case you do not need to set these environment variables.
3. (Optional) Set CUDA_INSTALL_PATH to the location of the installed CUDA toolkit.
4. (Optional) Set NV_ARCH to a comma separated list of NVIDIA GPU architectures (Fermi, Kepler, ..., Ampere).
5. Run make config in the op2 directory and verify that the compilers, libraries and compilation flags are as you intend.
6. Run make -j$(nproc) in the op2 directory to build the run-time libraries.
7. Run make -j$(nproc) in any of the app directories to build the respective apps.

Citing

To cite OP2, please reference the following paper:

G. R. Mudalige, M. B. Giles, I. Reguly, C. Bertolli and P. H. J. Kelly, "OP2: An active library framework for solving unstructured mesh-based applications on multi-core and many-core architectures," 2012 Innovative Parallel Computing (InPar), 2012, pp. 1-12, doi: 10.1109/InPar.2012.6339594.

@INPROCEEDINGS{6339594,
  author={Mudalige, G.R. and Giles, M.B. and Reguly, I. and Bertolli, C. and Kelly, P.H.J},
  booktitle={2012 Innovative Parallel Computing (InPar)},
  title={OP2: An active library framework for solving unstructured mesh-based applications on multi-core and many-core architectures},
  year={2012},
  volume={},
  number={},
  pages={1-12},
  doi={10.1109/InPar.2012.6339594}}