Python wrapper for STRUPACK. It currently supports

• single/double precision real/complex input matrix
• 32bit integer for matirx indices
• all C-interface routines for solver configuration

• 2020/12/17: Strumpack 5.0.0 Support
• 2019/07/23: C-interface SWIG wrapper works in serial mode (tested with Python3.7)

mpi compilers STRUMPACK installed as a shared library

   $ python setup.py instal

If include/library files are not in the standard path, you may need to define enviromental variable. See more detail in Install.txt

  $ cd example
  $ python example.py
  or
  $ mpirun -np 2 python example_MPI_DIST.py  

This solves a linear system correspoinding to a diffusion equation in 2D, and plot the following image

STRUMPACK.<PREFIX>StrumpackSolver where "S", "D", "C", "Z" corresponds to precision of float32, float64, complex64, complex128, respectively.

   from scipy.sparse import lil_matrix
   from STRUMPACK import *
     
   size = 10
   dtype = np.float64
   A = lil_matrix((10, 10),  dtype=dtype)
     
   # fill matrix in somewhy....
   for k in range(size):
       A[k, k] = 2
       if k > 0: A[k, k-1] = 1
   A = A.tocsr()
   b = np.ones(N, dtype=dtype)
   x = np.zeros(N, dtype=dtype)

   spss = SStrumpackSolver()
   # Note: command line argument can be passed as a list
   #       it calls set_from_options internally.
   #    spss = SStrumpackSolver(["--sp_enable_gpu", ])
   #    spss.set_from_options()
   spss.set_csr_matrix(A)
   spss.solve(b, x, 0)

STRUMPACK.<PREFIX>StrumpackSolver has methods to access functions defined in STRUMPACK C interface (StrumpackSparseSolver.hpp). For example, the following line shows how to set verbose mode and reordering method.

    from STRUMPACK import *
    spss = SStrumpackSolver()
    spss.set_verbose(1)
    spss.set_reordering_method(STRUMPACK_GEOMETRIC)