apxchol 0.1.0

CPU approximate-Cholesky preconditioner for graph-Laplacian / SDDM linear systems

apxchol (Python, CPU)

Approximate-Cholesky preconditioner for graph-Laplacian / SDDM linear systems.

import apxchol

solver = apxchol.solver(A)              # scipy sparse SPD Laplacian or SDDM; factor once
res = solver.solve(b, tol=1e-8, maxiter=500)
res.x, res.iters, res.residual, res.converged

z = solver.apply(r)                     # M^{-1} r
M = solver.aslinearoperator()           # use as M= in scipy.sparse.linalg.cg

res = apxchol.solve(A, b)               # one-shot convenience

Dev install (until a wheel is published)

The extension compiles two large TUs with -O3 -march=native; expect a minute or two.

# Build isolation provisions pybind11 + scikit-build-core automatically;
# editable keeps the source tree in place so the build reaches ../src, ../include.
pip install -e python
pytest python/tests -v

If your environment requires --no-build-isolation, install the build deps into the active env first: pip install pybind11 scikit-build-core (numpy/scipy are pulled in as runtime deps), then pip install -e python --no-build-isolation.

CPU only. GPU, tuning parameters, and 64-bit indices are not exposed in this version.