WATHEN
Assemble, Factor, Solve a Finite Element System

WATHEN is a FORTRAN90 library which compares storage schemes (full, banded, sparse triplet, sparse) and solution strategies (A\x, Linpack, conjugate gradient (CG)) for linear systems involving the Wathen matrix, which can arise when solving a problem using the finite element method (FEM).

The Wathen matrix is a typical example of a matrix that arises during finite element computations. The parameters NX and NY specify how many elements are to be set up in the X and Y directions. The number of variables N is then

        N = 3 NX NY + 2 NX + 2 NY + 1
      

However, the matrix is sparse, and a banded or sparse storage scheme can be used to save storage. However, even if storage is saved, a revised program may eat up too much time because MATLAB's sparse storage scheme is not efficiently used by inserting nonzero elements one at a time. Moreover, if banded storage is employed, the user must provide a suitable fast solver. Simply "translating" a banded solver from another language will probably not provide an efficient routine.

This library looks at how the complexity of the problem grows with increasing NX and NY; how the computing time increases; how the various full, banded and sparse approaches perform.

Licensing:

The computer code and data files made available on this web page are distributed under the GNU LGPL license.

Languages:

WATHEN is available in a C version and a C++ version and a FORTRAN90 version and a MATLAB version and a Python version.

Related Data and Programs:

CG, a FORTRAN90 library which implements a simple version of the conjugate gradient (CG) method for solving a system of linear equations of the form A*x=b, suitable for situations in which the matrix A is positive definite (only real, positive eigenvalues) and symmetric.

LINPACK_D, a FORTRAN90 library which factors and solves linear systems using double precision real arithmetic, by Jack Dongarra, Jim Bunch, Cleve Moler, Pete Stewart.

SPARSE_DISPLAY, is a FORTRAN90 library which can read information defining a matrix of numbers and display the sparsity pattern or location of the nonzero elements using gnuplot. This operation is already available in the built-in MATLAB "spy" command.

TEST_MAT, a FORTRAN90 library which defines test matrices for which some of the determinant, eigenvalues, inverse, null vectors, P*L*U factorization or linear system solution are already known, including the Vandermonde and Wathen matrix.

Reference:

1. Nicholas Higham,
   Algorithm 694: A Collection of Test Matrices in MATLAB,
   ACM Transactions on Mathematical Software,
   Volume 17, Number 3, September 1991, pages 289-305.
2. Andrew Wathen,
   Realistic eigenvalue bounds for the Galerkin mass matrix,
   IMA Journal of Numerical Analysis,
   Volume 7, Number 4, October 1987, pages 449-457.

Source Code:

• wathen.f90, the source code.

Examples and Tests:

• wathen_test.f90, a sample calling program.
• wathen_test.txt, the output file.

List of Routines:

• BANDWIDTH returns the bandwidth of a matrix.
• CG_FULL uses the conjugate gradient method for a full storage matrix.
• CG_ST uses the conjugate gradient method for a sparse triplet storage matrix.
• DAXPY computes constant times a vector plus a vector.
• DDOT forms the dot product of two vectors.
• DGBFA factors a real band matrix by elimination.
• DGBSL solves a real banded system factored by DGBCO or DGBFA.
• DGEFA factors a real general matrix.
• DGESL solves a real general linear system A * X = B.
• DSCAL scales a vector by a constant.
• IDAMAX indexes the array element of maximum absolute value.
• NONZEROS counts the nonzeros in a matrix.
• R8_UNIFORM_01 returns a unit pseudorandom R8.
• R8GE_CG uses the conjugate gradient method on an R8GE system.
• R8GE_MXV multiplies an R8GE matrix by an R8VEC.
• R8MAT_UNIFORM_01 fills an R8MAT with unit pseudorandom numbers.
• R8VEC_PRINT prints an R8VEC.
• R8VEC_UNIFORM_01 returns a unit pseudorandom R8VEC.
• ST_MV multiplies a sparse triple matrix times a vector.
• TIMESTAMP prints the current YMDHMS date as a time stamp.
• WATHEN_BANDWIDTH returns the bandwidth of the WATHEN matrix.
• WATHEN_BANDED returns the Wathen matrix, using banded storage.
• WATHEN_FULL returns the Wathen matrix, using full storage.
• WATHEN_ORDER returns the order of the WATHEN matrix.
• WATHEN_ST: Wathen matrix stored in sparse triplet format.
• WATHEN_ST_SIZE: Size of Wathen matrix stored in sparse triplet format.

You can go up one level to the FORTRAN90 source codes.