OPS Example Applications

This directory contains a suite of scientific mini-applications and benchmarks ported to the OPS framework. Each application demonstrates how OPS enables single-source code to be automatically parallelized for a range of backends (OpenMP, CUDA, HIP, SYCL, MPI, etc.).

How to Build and Run

  • Each application has a test.sh script in its directory (e.g., OPS/apps/c/CloverLeaf/test.sh) that automates building and running all supported parallel versions.

  • Set up the required environment variables (e.g., OPS_INSTALL_PATH, MPI_INSTALL_PATH, etc.) as described in the main OPS documentation.

  • Run the test script: bash test.sh

  • For more details on generated code, see the OPS-APPS repository.

Main Applications

CloverLeaf (2D & 3D)

  • Location: apps/c/CloverLeaf, apps/c/CloverLeaf_3D, apps/c/CloverLeaf_3D_HDF5

  • Purpose: Solves the compressible Euler equations for hydrodynamics in 2D/3D using explicit, second-order methods. Used for benchmarking and performance studies. See the original CloverLeaf project for more details.

  • Features: Demonstrates all OPS backends, including OpenMP, CUDA, HIP, SYCL, and their combinations with MPI. The HDF5 variant tests file I/O.

Poisson

  • Location: apps/c/poisson

  • Purpose: Solves the Poisson equation using iterative methods (e.g., Jacobi, Gauss-Seidel). Demonstrates OPS support for stencil computations and iterative solvers in C.

  • Features: Supports all major OPS backends and demonstrates iterative solver patterns.

TeaLeaf

  • Location: apps/c/TeaLeaf

  • Purpose: Solves the linear heat conduction equation using iterative solvers (CG, Jacobi, Chebyshev, PPCG). Used for performance and scaling studies. See the original TeaLeaf project for more details.

  • Features: Supports all major OPS backends and demonstrates iterative solver patterns.

ADI (Alternating Direction Implicit)

  • Location: apps/c/adi

  • Purpose: Demonstrates the ADI method for solving parabolic PDEs. Useful for testing OPS with implicit time-stepping and tridiagonal solvers. See the original ADI/tridsolver project for more details.

  • Features: Showcases the tridiagonal solver for available OPS backends (e.g., OpenMP, CUDA, MPI, etc.).

Laplace2D Tutorial (C)

  • Location: apps/c/laplace2d_tutorial

  • Purpose: Simple 2D Laplace equation solver in C. Serves as a tutorial for new OPS users, especially in C.

  • Features: Supports all major OPS backends and demonstrates iterative solver patterns.

Additional C Applications

  • adi_burger / adi_burger_3D: ADI method for Burger’s equation in 2D/3D, demonstrating implicit solvers and tridiagonal systems.

  • compact_scheme: Example of compact finite difference schemes.

  • complex / complex_numbers: Demonstrates OPS support for complex data types and operations.

  • halfprecision: Tests half-precision floating point support in OPS.

  • hdf5_slice: Example of HDF5 file I/O and slicing.

  • mb_shsgc / mb_trid_test / mblock / mblock4D: Multi-block and multi-dimensional domain decomposition and tridiagonal solver tests.

  • mgrid: Multi-grid solver example.

  • multiDim / multiDim3D / multiDim_HDF5: Multi-dimensional stencils and reductions, including HDF5 I/O.

  • ops-lbm: Lattice Boltzmann Method (LBM) fluid dynamics solver.

  • random: Random number generation and usage in OPS.

  • shsgc: Scientific kernel or benchmark (details TBD).

  • tiling_fix: Demonstrates tiling and performance tuning in OPS.

  • tti: Tilted Transverse Isotropy (TTI) wave equation solver.

  • wave_test: Wave equation solver.

Fortran Applications

  • laplace2dtutorial: Simple 2D Laplace equation solver (tutorial).

  • lowdim_test: Low-dimensional test problems for regression and performance.

  • mblock: Multi-block domain decomposition example.

  • multiDim / multiDim3D: Multi-dimensional reduction and stencil operations.

  • poisson: Poisson equation solver (also in C).

  • random: Random number generation in Fortran with OPS.

  • shsgc: Scientific kernel or benchmark (details TBD).

  • mathtest: Mathematical function and operation tests (details TBD).

Laplace2D Tutorial

  • Location: apps/fortran/laplace2dtutorial

  • Purpose: Simple 2D Laplace equation solver. Serves as a tutorial for new OPS users, especially in Fortran.

  • Features: Supports all major OPS backends and demonstrates iterative solver patterns.

Adding New Applications

  • Place your application in the appropriate language subdirectory (c/ or fortran/).

  • Provide a Makefile to support building your application with the OPS build system.

  • Provide a test.sh script to automate building and running all supported backends.

  • Document any special requirements in a README file in your application directory.

For more information, see the main OPS documentation and the OPS-APPS repository for pre-generated code and further examples.