Getting Started

This page shows how to build, run, and visualize a small demo locally.

Prerequisites

  • C++17 compiler and CMake ≥ 3.21
  • MPI (OpenMPI or MPICH)
  • NetCDF C library (libnetcdf-dev)
  • Python 3.11+ for visualization: pip install -r requirements.txt

Build 

cmake -S . -B build -G Ninja -DMPI_TEST_PROCS=4
cmake --build build --parallel

Run a demo 

mpirun --oversubscribe -np 4 ./build/src/climate_sim       --config=configs/dev.yaml       --output.format=netcdf

Suggested dev.yaml parameters for a compelling animation (already in repo):

grid:    { nx: 512, ny: 512, dx: 1.0, dy: 1.0 }
physics: { D: 0.5, vx: 0.5, vy: 0.0 }
time:    { dt: 0.1, steps: 800, out_every: 10 }
bc:      periodic
output:  { prefix: "dev" }
ic:
  preset: gaussian_hotspot
  params: { A: 1.0, sigma_frac: 0.06, xc_frac: 0.5, yc_frac: 0.5 }

CSV vs NetCDF outputs

  • CSV: human-readable, one file per rank, useful for debugging.
  • NetCDF: structured, one file per snapshot (recommended for production & visualization).

Visualize

Single frame (snapshot) 

python -m visualization.cli show --dir outputs --step 400 --save demo_frame.png --overlay-minmax

Animation 

# MP4 (requires ffmpeg)
python -m visualization.cli animate --dir outputs --fps 24 --save demo.mp4 --overlay-minmax

# GIF (using Pillow)
python -m visualization.cli animate --dir outputs --fps 12 --save demo.gif --writer pillow

Tips

  • Use --cmap plasma or another Matplotlib colormap for different visuals.
  • Combine overlays (--overlay-minmax --overlay-rankgrid) for richer diagnostics.
  • For large runs, prefer NetCDF output for efficiency.

Install

You can install climate-sim using these commands

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build --parallel
sudo cmake --install build --prefix /usr/local # Or specify a different folder
which climate_sim # Verify installation

← Back to Home