|Routine ID: F011|
|Author(s): G.A. Erskine, H. Lipps||Library: KERNLIB|
|Language: Fortran or Assembler or COMPASS||Revised: 27.11.1984|
These subroutines provide a two-step procedure for solving sets of linear equations
which is faster than the library programs RINV (F010) when (*) must be solved repeatedly for the same matrix A with different sets of right-hand sides. The inverse matrix and the determinant det( A) may also be calculated.
User Entry Names: RFACT, RFEQN, RFINV, DFACT, DFEQN, DFINV, CFACT, CFEQN, CFINV
Internal Entry Names: TMPRNT
Files Referenced: Printer
External References: KERMTR (N001), ABEND (Z035)
For (type REAL), (type DOUBLE PRECISION), (type COMPLEX):
CALL tFACT(N,A,IDIM,IR,IFAIL,DET,JFAIL) CALL tFEQN(N,A,IDIM,IR,K,B) CALL tFINV(N,A,IDIM,IR)
Subroutine tFACT must be called with matrix A in array A prior to any calls to tFEQN and tFINV. On return the situation is as follows:
If A is singular, IFAIL will be set to , in which case any subsequent call to tFEQN or tFINV will give unpredictable results.
If the evaluation of det( A) would probably cause underflow, JFAIL will be set to and DET will be set to zero.
If the evaluation of det( A) would probably cause overflow, JFAIL will be set to and DET will be incorrect.
Execution continues, and subsequent calls to tFEQN and tFINV will give correct results.
Subroutine tFINV may be called only after tFACT has been called, with the contents of A and R unchanged. On return, A will contain the inverse of A. Therefore, once tFINV has been called, it is no longer meaningful to call tFEQN with A as parameter.
Triangular factorization with row interchanges. The inverse matrix is the product, in reverse order, of the in-place inverses of the triangular factors. The array R holds information specifying the row interchanges.
On computers with IBM 370 architecture, inner products are accumulated using double-precision arithmetic internally for arrays of type REAL and COMPLEX.
If or or , a message is printed and program execution is terminated by calling ABEND (Z035).
Assume that the matrix A, the matrix B, and the 10-element vector z are stored according to the Fortran convention in arrays A, B and Z respectively of a program containing the declarations
DIMENSION IR(25) COMPLEX A(25,30),B(25,10),Z(25),DETThen, unless A is singular (which is to cause a jump to statement 100), the following statements will set , replace B by , replace z by , and replace A by :
CALL CFACT (10,A,25,IR,IFAIL,DET,JFAIL) IF(IFAIL .NE. 0) GO TO 100 CALL CFEQN(10,A,25,IR,3,B) CALL CFEQN(10,A,25,IR,1,Z) CALL CFINV(10,A,25,IR)