Quick Start

Get OUxSBLI running in under 10 minutes. This guide assumes you already have the NVIDIA HPC SDK installed. If not, start with Installation.

Prerequisites

Requirement	Minimum version
NVIDIA HPC SDK (`mpif90`)	24.x or 25.x
CMake	3.18
fypp	any (`pip install fypp`)
GPU	NVIDIA Ampere or newer
MPI ranks available	2

1. Clone the repository

git clone https://github.com/htymjun/OUxSBLI.git
cd OUxSBLI

2. Install fypp

CMake uses fypp to preprocess the Fortran templates at build time:

pip install fypp

3. Build the NS Taylor-Green vortex case

3D_solver/NSTGV is the recommended starting point — it is a well-validated 3D case with no special boundary conditions.

cd 3D_solver/NSTGV
cmake -B build && cmake --build build -j

CMake automatically runs fypp on all .f90.fypp templates and compiles the result. The executable a.out is placed in build/.

4. Run the simulation

cd build
mpirun -n 2 ./a.out

VTK output files (Q00000.vtr, Q00001.vtr, …) appear in data/ inside the build directory. The default configuration runs for several hundred timesteps; you can reduce nt in mod_globals.f90 for a quick smoke-test.

5. Visualise in ParaView

Open ParaView and choose File → Open.
Navigate to 3D_solver/NSTGV/build/data/ and select Q00000.vtr (ParaView will offer to open the whole time series).
Click Apply, then choose a field (e.g. density or velocity magnitude) from the variable dropdown.

6. Next steps

Goal	Where to look
Change the convective scheme or spatial order	Configuration
Run a different test case	Installation → Available Cases
Use the Python API for parametric runs	Python API
Understand the numerical methods	Theory
Set up on a fresh machine	Installation