pypardiso 0.4.7

Python interface to the Intel MKL Pardiso library to solve large sparse linear systems of equations

PyPardiso

PyPardiso is a python package to solve large sparse linear systems of equations with the Intel oneAPI Math Kernel Library PARDISO solver, a shared-memory multiprocessing parallel direct sparse solver.

PyPardiso provides the same functionality as SciPy's scipy.sparse.linalg.spsolve for solving the sparse linear system Ax=b. However in many cases it is significantly faster than SciPy's built-in single-threaded SuperLU solver.

PyPardiso is not a python interface to the PARDISO Solver from the PARDISO 7.2 Solver Project and it also doesn't currently support complex numbers. Check out JuliaSparse/Pardiso.jl for these more advanced use cases. For macOS users we recommend scikit-umfpack as an alternative fast solver, since MKL is not available on Apple silicon.

Installation

PyPardiso runs on Linux and Windows. It can be installed with conda or pip. It is recommended to install PyPardiso using a virtual environment.

conda-forge	PyPI
conda install -c conda-forge pypardiso	pip install pypardiso

Basic usage

How to solve the sparse linear system Ax=b for x, where A is a square, sparse matrix in CSR (or CSC) format and b is a vector (or matrix):

In [1]: import pypardiso

In [2]: import numpy as np

In [3]: import scipy.sparse as sp

In [4]: A = sp.rand(10, 10, density=0.5, format='csr')

In [5]: A
Out[5]:
<10x10 sparse matrix of type '<class 'numpy.float64'>'
	with 50 stored elements in Compressed Sparse Row format>

In [6]: b = np.random.rand(10)

In [7]: x = pypardiso.spsolve(A, b)

In [8]: x
Out[8]:
array([ 0.02918389,  0.59629935,  0.33407289, -0.48788966,  3.44508841,
        0.52565687, -0.48420646,  0.22136413, -0.95464127,  0.58297397])