Mosaic is designed to grow with community-contributed solver backends. Each domain has room for additional solvers that would broaden the benchmark's coverage.
Candidate solvers (non-exhaustive):
- NS-grid: Taichi (GPU-accelerated, Python)
- Structural-mesh: MFEM (C++, widely used in HPC)
- Thermal-mesh: scikit-fem (pure Python, lightweight FEM)
- All: Tensor-Mesh (JAX-based FEM)
What's involved:
Three files under mosaic/tesseracts/<domain>/<solver-name>/: tesseract_api.py, tesseract_config.yaml, and a Dockerfile (or build_pip_venv.sh). The Add a Solver tutorial walks through the full process using a working example.
Quick check: mosaic run -p <domain> --suites forward -s <your-solver> --debug should complete without errors.
Mosaic is designed to grow with community-contributed solver backends. Each domain has room for additional solvers that would broaden the benchmark's coverage.
Candidate solvers (non-exhaustive):
What's involved:
Three files under
mosaic/tesseracts/<domain>/<solver-name>/:tesseract_api.py,tesseract_config.yaml, and aDockerfile(orbuild_pip_venv.sh). The Add a Solver tutorial walks through the full process using a working example.Quick check:
mosaic run -p <domain> --suites forward -s <your-solver> --debugshould complete without errors.