|
351. SPAMAT: Solutions of Linear Systems of Equations,
Corresponding to Sparse Matrices, Introducing Only Non-
Zero Elements into Memory
by Alain Pellegatti and Philippe François, Laboratoire
de Chimie Théorique, Université de Provence, Place
Victor Hugo, 13331 Marseille Cedex 3, France
The aim of this program is to solve linear systems of
equations corresponding to sparse matrices using
Jordan's method.Only the nonzero elements are
explicitly stored in the memory.
We have to solve AX = B of order N. The matrix A'
considered in the calculations has the dimension (N,
N+1); the (N+1)o column corresponding to the vector B.
The matrix is stored as a one-dimensional array.
The base-row and base-column are a fictitious column
and row useful to indicate the frontiers of the matrix
A'. Therefore, we have N elements in the base-row and
(N+1) elements in the base-column.Since the
operations involving the base-row and the base-column
are completely independent, one can give the same
address by twos to the N first elements in both arrays.
The nonzero elements are then numbered row by row.
Each nonzero matrix element whose address is I is
identified by five more parameters:
LEFT:link to the address of the next nonzero
element to left in the row
IUP: link to the address of the next nonzero
element upward in the column
KI: row number
KJ: column number
VAL: numerical value
For each element of the base row and the base-column,
only two parameters are necessary:
I LEFT KJ=-1 (base-row)
I IUPKI=-1 (base-column)
FORTRAN IV (IBM 360/370)
Lines of Code: 507
Recommended Citation: A. Pellegatti and P. François,
QCPE 11, 351 (1978).
| 
