NESC9656 ITMETH.

1. NAME OR DESIGNATION OF PROGRAM:  ITMETH.

3. DESCRIPTION OF PROGRAM OR FUNCTION

ITMETH is a collection of iterative routines for solving large, sparse linear systems.

4. METHOD OF SOLUTION

ITMETH solves general linear systems of the form AX=B using a variety of methods: Jacobi iteration; Gauss-Seidel iteration; incomplete LU decomposition or matrix splitting with iterative refinement; diagonal scaling, matrix splitting, or incomplete LU decomposition with the conjugate gradient method for the problem AA'Y=B, X=A'Y; biconjugate gradient method wih diagonal  scaling, matrix splitting, or incomplete LU decomposition; and orthomin method with diagonal scaling, matrix splitting, or incomplete LU decomposition. ITMETH also solves symmetric positive definite linear systems AX=B using the conjugate gradient method with diagonal scaling or matrix splitting, or the incomplete Cholesky conjugate gradient method.

5. RESTRICTIONS ON THE COMPLEXITY OF THE PROBLEM:

6. TYPICAL RUNNING TIME:

7. UNUSUAL FEATURES OF THE PROGRAM:

8. RELATED AND AUXILIARY PROGRAMS

ITMETH contains the iterative routines from SLAP but without the dynamic memory allocation facility. The ITMETH source includes several of the BLAS (Basic Linear Algebra Subprograms) from LINPACK (NESC 800).

10. REFERENCES:

11. MACHINE REQUIREMENTS:

13. OPERATING SYSTEM UNDER WHICH PROGRAM IS EXECUTED:

14. OTHER PROGRAMMING OR OPERATING INFORMATION OR RESTRICTIONS

ITMETH  calls two user-supplied routines, MSOLVE and MTSOLV. MSOLVE solves a linear system with the splitting matrix; MTSOLV solves a linear system with the transpose of the splitting matrix. These routines are not included.

15. NAME AND ESTABLISHMENT OF AUTHORS

          A. Greenbaum
          Lawrence Livermore National Laboratory

File name	File description	Records
NESC9656_01.001	Information file	31
NESC9656_01.002	ITMETH FORTRAN source	3057
NESC9656_01.003	BLAS (Basic Linear Algebra Subprogram)	307