mumfpack_sparse's Introduction

mUMFPACK Sparse Matrix Solver Example (Fortran 2003)

The mUMFPACK module provides a Fortran-2003 implementation of the full UMFPACK interface.

There are three parts in this program.

Create Sparse Triplet (ST) Matrix File, using ST_IO
Read the ST file and Converse it to Compressed Column (CC) format, using ST_TO_CC
mUMFPACK provides a high-level operator .umfpack., just one line can solve the sparse matrix instead of using umf4sym, umf4num so that saving memory from creating arrays.

Quickstart

Dependancies

To get started, you will need to install suitesparse. You can use homebrew to install it in your Mac. You also need to build ST_IO and ST_TO_CC library in your search path.

brew install suitesparse

Compile

gfortran umfpack.f90 sparse.f90 -lst_io -lst_to_cc -lcc_io -lumfpack -lamd -lcholmod -lcolamd -lsuitesparseconfig -lblas -o sparse

Run

$ ./sparse 
Sparse Triplet (ST) header information:

Minimum row index I_MIN =        1
Maximum row index I_MAX =        5
Minimum col index J_MIN =        1
Maximum col index J_MAX =        5
Number of rows        M =        5
Number of columns     N =        5
Number of nonzeros  NST =       12

Number of CC values =   12

The CC matrix:
  #     I     J         A
----  ----  ----  ----------------

  0     0     0     2.0000000
  1     1     0     3.0000000
  2     0     1     3.0000000
  3     2     1    -1.0000000
  4     4     1     4.0000000
  5     1     2     4.0000000
  6     2     2    -3.0000000
  7     3     2     1.0000000
  8     4     2     2.0000000
  9     2     3     2.0000000
 10     1     4     6.0000000
 11     4     4     1.0000000

High-level Fortran operator (.umfpack.), CSC allocatable operands
x (1) = 1.0000000000
x (2) = 2.0000000000
x (3) = 3.0000000000
x (4) = 4.0000000000
x (5) = 5.0000000000

Recommend Projects