openseespy-solvers 0.1.0

Ready-to-use, SciPy-style PythonSparse solvers for OpenSeesPy

openseespy-solvers

Sparse linear algebra solvers for OpenSeesPy PythonSparse commands.

Documentation: openseespy-solvers.readthedocs.io

Installation

Requires Python ≥ 3.12, NumPy ≥ 1.26, SciPy ≥ 1.12, and serial OpenSeesPy (parallel/MPI OpenSeesPy builds are not supported — see docs).

Full setup from scratch

Copy-paste overview — see the installation guide for platform notes (UMFPACK, GPU wheels) and the complete verification checklist.

# 1. Environment (conda or venv — all platforms)
conda create -n openseespy-solvers python=3.12 -y
conda activate openseespy-solvers

# 2. Clone (examples and pytest are in the repo, not the pip wheel)
git clone https://github.com/gaaraujo/openseespy-solvers.git
cd openseespy-solvers

# 3. Install
pip install -e ".[dev,opensees]"

# 4–5. Optional backends — UMFPACK, CuPy, nvMath (see installation guide)
# 6. pip check
# 7. Verify
pytest
cd examples && python solvers/scipy_spsolve.py && python solvers/scipy_eigsh.py
python brick_bar.py && python brick_bar_eigen.py

GPU tests in pytest run when CuPy/nvMath are installed; skipped otherwise.

Quick install (library only)

pip install openseespy-solvers
pip install openseespy-solvers[opensees]   # if OpenSeesPy is missing

Recommended solvers

Analysis	GPU (CUDA)	CPU
Linear (PythonSparse, static/transient/…)	nvmath.direct_solver	scipy.spsolve or scipy.umfpack
Eigen	cupy.eigsh	scipy.eigsh

These PythonSparse backends are often faster than native OpenSees sparse/direct/eigen solvers on medium-to-large models because they use optimized library implementations. Details: recommended solvers.

Example

from openseespy_solvers import hybrid
from openseespy_solvers.scipy import spsolve

solver = hybrid(spsolve(), rtol=1e-6, restart=50)
ops.system("PythonSparse", solver.to_openseespy())
ops.analyze(n)

For Newton or transient steps where the tangent changes slowly, hybrid factorizes once then reuses that factorization as a GMRES preconditioner until the system size changes or GMRES fails to converge.

Submodules

Module	Functions
openseespy_solvers.scipy	spsolve, umfpack, cg, gmres, eigsh, lobpcg
openseespy_solvers.scipy.precond	jacobi, ilu, direct
openseespy_solvers.cupy	spsolve, cg, gmres, eigsh, lobpcg (GPU)
openseespy_solvers.cupy.precond	jacobi, ilu, direct
openseespy_solvers.nvmath	direct_solver (GPU)
openseespy_solvers.hybrid	hybrid(direct=...) — frozen factorization + GMRES

Factory signatures match scipy.sparse.linalg and cupyx.scipy.sparse.linalg; A and b are supplied by OpenSees at solve time.

See the tutorial and API reference for details.

GitHub

Repository: github.com/gaaraujo/openseespy-solvers. Report bugs and ask questions via issues; send contributions (pull requests) there as well.

License

BSD 3-Clause — see LICENSE.