NRPy

NRPy is a Python/SymPy-based symbolic code generation toolkit for numerical relativity and relativistic astrophysics. It uses symbolic equation modules to generate standalone C/CUDA applications, Einstein Toolkit thorns, Charm++ projects, and JAX/Python workflows. In its current public form, NRPy provides the core infrastructure for single-patch BlackHoles@Home (BH@H) numerical relativity simulations in Cartesian, curvilinear, and singular coordinate systems.

At a high level:

• nrpy/examples contains runnable project generators.
• nrpy/equations, nrpy/infrastructures, and nrpy/helpers contain the symbolic physics, backend-specific code generation, and shared support code that those generators use.

NRPy is the product of many scientists' work. Please cite the relevant NRPy papers in your publications; these citations go a long way toward proper acknowledgement of these efforts and support many junior scientists' careers.

For end users who want to generate projects from the published package:

python -m pip install nrpy

For contributors working from a clone of this repository:

git clone https://github.com/nrpy/nrpy.git
cd nrpy
python -m pip install -U -r requirements-dev.txt
python -m pip install -e .

After an editable install, you can run generators from the repository root with python -m nrpy.examples.<example_name>.

If you run Python examples directly from this directory without relying on an editable install, append . to PYTHONPATH first:

export PYTHONPATH="${PYTHONPATH:+${PYTHONPATH}:}."

python -m pip install nrpy installs the Python package, but many generated projects also need external tools:

• Standalone BHaH examples: a C compiler and make; some also link against GSL
• Waveform and geodesic examples: a C compiler, make, and GSL
• Einstein Toolkit / Carpet / CarpetX generators: the Python package install is enough to generate thorns; an existing Einstein Toolkit environment is needed to build and run the generated thorns
• superB examples: a Charm++ toolchain; some also link against GSL
• JAX example generation: Python tooling for the generated project, plus JAX in that generated environment as needed

If you are new to NRPy, start with one of the standalone BHaH examples below that does not require GSL. Those examples are the closest path to the current public single-patch BH@H workflow.

This is the safest first workflow because it only requires Python, a C compiler, and make.

python -m nrpy.examples.wave_equation_cartesian

This writes a buildable project to project/wave_equation_cartesian/.

cd project/wave_equation_cartesian
make

./wave_equation_cartesian

The generator also creates a parameter file named wave_equation_cartesian.par. Many standalone BHaH examples accept command-line overrides as well; for this example, convergence_factor is one supported input.

Expect generated diagnostics and data products in the project directory. The exact filenames vary by example, but the important milestone is:

• generation succeeds
• make succeeds
• the executable runs without further manual project editing

For this example, one simple command-line override is convergence_factor. For example:

./wave_equation_cartesian 2.0

This reruns the example at higher resolution and writes output files such as out0d-conv_factor2.00.txt.

If you want the closest public example of the single-patch BH@H numerical relativity workflow, use:

python -m nrpy.examples.two_blackholes_collide

This example:

• evolves Brill-Lindquist binary black hole initial data
• uses a single-patch curvilinear spherical-coordinate setup
• serves as a compact public example of the single-patch BH@H workflow
• is intentionally lightweight: it is cheap enough to run in seconds even on very modest hardware

Once the first standalone workflow works, these are reasonable next steps:

• Single-patch BH@H NR example: python -m nrpy.examples.two_blackholes_collide
• Lightweight elliptic initial-data workflow: python -m nrpy.examples.nrpyelliptic_conformally_flat
• Waveform generation, requires GSL: python -m nrpy.examples.seobnrv5_aligned_spin_inspiral -seobnrv5_bob
• Einstein Toolkit / CarpetX thorn generation; building the generated thorns requires an ET environment: python -m nrpy.examples.carpetx_wavetoy_thorns

BHaH is NRPy's standalone application infrastructure. In this public repository, it provides the core capabilities for single-patch numerical relativity evolutions in Cartesian, curvilinear, and singular coordinate systems, and these generators typically produce buildable C or CUDA-ready projects under project/<name>/.

• Wave equation solvers: python -m nrpy.examples.wave_equation_cartesian, python -m nrpy.examples.wave_equation_curvilinear, python -m nrpy.examples.wave_equation_multicoordinates
• Black hole evolution and diagnostics: python -m nrpy.examples.two_blackholes_collide, python -m nrpy.examples.blackhole_spectroscopy (requires GSL), python -m nrpy.examples.spinning_blackhole
• Elliptic / initial-data workflows: python -m nrpy.examples.nrpyelliptic_conformally_flat

These generators target the Einstein Toolkit rather than producing standalone executables. ETLegacy refers to the classic Einstein Toolkit / Carpet-style target; CarpetX targets the newer CarpetX driver stack.

• ETLegacy / Carpet workflows: python -m nrpy.examples.carpet_wavetoy_thorns, python -m nrpy.examples.carpet_baikal_thorns
• CarpetX workflows: python -m nrpy.examples.carpetx_wavetoy_thorns, python -m nrpy.examples.carpetx_baikal_thorns

These generators write thorn directories under project/<name>/. You only need an Einstein Toolkit checkout when you want to compile or run those generated thorns.

These generators target superB, NRPy's Charm++-based infrastructure for distributed-memory workflows and a scalable extension of single-patch BH@H-generated code.

• GR evolution and spectroscopy: python -m nrpy.examples.superB_two_blackholes_collide, python -m nrpy.examples.superB_blackhole_spectroscopy
• Elliptic initial data: python -m nrpy.examples.superB_nrpyelliptic_conformally_flat

These generators focus on compact-binary dynamics, inspiral, and waveform modeling.

• SEOBNRv5 / BOB waveform projects: python -m nrpy.examples.seobnrv5_aligned_spin_inspiral -seobnrv5_bob, python -m nrpy.examples.seobnrv5_aligned_spin_inspiral -seobnrv5_nrpy
• SEBOBv2 standalone project, requires GSL: python -m nrpy.examples.sebobv2
• JAX project generation: python -m nrpy.examples.sebobv1_jax
• Post-Newtonian utility: python -m nrpy.examples.nrpypn_quasicircular_momenta

These generators cover additional relativistic physics applications beyond the main PDE and thorn-generation workflows.

• Neutron-star and matter-related workflows: python -m nrpy.examples.tovola_neutron_star (requires GSL), python -m nrpy.examples.hydro_without_hydro (requires GSL)

These generators focus on test-particle dynamics in curved spacetimes. They produce standalone C or CUDA projects that evolve trajectories and include companion Python scripts for visualization (e.g., trajectory plots and lensed image rendering).

• Single massive particle integration, requires GSL: python -m nrpy.examples.mass_geodesic_integrator
• Single photon integration: python -m nrpy.examples.photon_geodesic_integrator
• Batch photon ray-tracing (supports CUDA or OpenMP): python -m nrpy.examples.photon_geodesic_batch_integrator

• Apparent horizon library generation: python -m nrpy.examples.bhahaha

• Standalone BHaH examples usually generate a project directory with C source, headers, a Makefile, a parameter file, and a runnable executable target for single-patch Cartesian or curvilinear-coordinate workflows.
• Waveform and geodesic generators also produce standalone projects, but they typically require GSL at build time.
• Some additional standalone physics examples, including blackhole_spectroscopy, hydro_without_hydro, tovola_neutron_star, and sebobv2, also require GSL at build time.
• Einstein Toolkit and CarpetX generators produce thorn directories that you copy or link into an Einstein Toolkit checkout; an ET environment is needed for the subsequent build/run step, not for thorn generation itself.
• superB generators produce Charm++ projects rather than plain standalone executables, and some of them also require GSL.
• sebobv1_jax generates a Python/JAX project instead of a C executable.

Most generators print project-specific next steps when they finish, but the exact instructions vary by example.

• nrpy/examples: runnable generators and project entry points
• nrpy/infrastructures: backend-specific code generation layers, including BHaH, ETLegacy, CarpetX, superB, and JAX
• nrpy/equations: symbolic physics and equation modules
• nrpy/helpers: shared code generation and utility helpers
• nrpy/tests: reference-metric validation modules
• bin/nrpyinline.py: advanced utility for extracting and running embedded NRPy snippets from text or source files

The recommended local workflow is:

python -m pip install -U -r requirements-dev.txt
python -m pip install -e .

You do not need to modify PYTHONPATH if you install the repository in editable mode. If you run Python examples directly from this directory without using the editable install, append . to PYTHONPATH first:

export PYTHONPATH="${PYTHONPATH:+${PYTHONPATH}:}."

Useful local sanity checks include:

python -m nrpy.examples.wave_equation_cartesian
python -m nrpy.examples.sebobv1_jax

For broader validation, see .github/workflows/main.yml, which exercises static analysis, standalone project generation/builds, Einstein Toolkit validation, Charm++ / superB workflows, and SEOB consistency checks.

Older notebook-centered NRPy+ (NRPy 1.0) material lives in the separate nrpytutorial repository. This repository focuses on the current package and generator-based workflow.