Next: 3 The domain decomposition method Up: 2 Iterative methods Prev: 2.1 The modules Contents

2.2 The algorithms

It could be interesting to use the algorithms directly, in particular to write an efficient non-linear procedure. Two programs correspond to each module: one in single precision and one in double precision. In principle, the names of the two programs differ only in the last letter: R in single precision (library RESR), D in double (library RESD).

2.2.1 List of programs

• Matrix assembly 
  • ASGC1D assembly of a symmetric or non-symmetric matrix in double precision,
  • ASGC1R assembly of a symmetric or non-symmetric matrix in single precision.
• Assembly of the right-hand-side 
  • ASSBPD assembly of the RHS in double precision (structure B in secondary memory),
  • ASSEBD assembly of the RHS in double precision (structure B in main memory),
  • ASSMBD vectorial assembly of the RHS in double precision,
  • ASSBPR assembly of the RHS in single precision (structure B in secondary memory),
  • ASSEBR assembly of the RHS in single precision (structure B in main memory),
  • ASSMBR vectorial assembly of the RHS in double precision.
• Impose the boundary conditions 
  • CLGC1D boundary conditions of type = V in double precision,
  • CLGC2D boundary conditions i.t.o. linear relations in double precision,
  • CLGC1R boundary conditions of type = V in single precision,
  • CLGC2R boundary conditions i.t.o. linear relations in single precision.
• Solution by iterative method 
  • Incomplete factorization  (computation of the preconditioning matrix)
    • CDLL1D incomplete Cholesky factorization in double precision,
    • CDLL2D incomplete Crout factorization in double precision,
    • CDLU1D incomplete Gauss factorization in double precision,
    • CDLL1R incomplete Cholesky factorization in single precision,
    • CDLL2R incomplete Crout factorization in single precision,
    • CDLU1R incomplete Gauss factorization in single precision.
  • Solution conjugate gradient iterations
    • DGRA1D accelerated double conjugate gradient with incomplete Gauss preconditioning in double precision,
    • ICHR1D conjugate gradient with incomplete Cholesky/Crout preconditioning in double precision,
    • GCDIAD conjugate gradient with diagonal preconditioning in double precision,
    • SIMGCD conjugate gradient without preconditioning in double precision,
    • SSOR1D conjugate gradient with SSOR preconditioning in double precision,
    • DGRA1R accelerated double conjugate gradient with incomplete Gauss preconditioning in single precision,
    • ICHR1R conjugate gradient with Cholesky/Crout preconditioning in single precision,
    • GCDIAR conjugate gradient with diagonal preconditioning   in single precision,
    • SIMGCR conjugate gradient without preconditioning in single precision,
    • SSOR1R conjugate gradient with SSOR preconditioning   in single precision.
  • Solution of a preconditioned linear system
    • DRCHID incomplete Cholesky preconditioning in double precision,
    • DRCRID incomplete Crout preconditioning in double precision,
    • DGRA1D incomplete Gauss preconditioning in double precision,
    • SSOR2D preconditioning by matrix SSOR in double precision,
    • DRCHIR incomplete Cholesky preconditioning in single precision,
    • DRCRIR incomplete Crout preconditioning in single precision,
    • DGRA1R incomplete Gauss preconditioning in single precision,
    • SSOR2R preconditioning by matrix SSOR in single precision.