NAME

zpoequ - compute row and column scalings intended to equilibrate a Hermitian positive definite matrix A and reduce its condition number (with respect to the two-norm)

SYNOPSIS

SUBROUTINE ZPOEQU(
N, A, LDA, S, SCOND, AMAX, INFO )

void zpoequ(long int n, doublecomplex ∗za, long int lda,
double ∗s, double ∗scond, double ∗amax, long int ∗info)

INTEGER INFO, LDA, N

DOUBLE PRECISION AMAX, SCOND

DOUBLE PRECISION S( ∗ )

COMPLEX∗16 A( LDA, ∗ )

PURPOSE

ZPOEQU computes row and column scalings intended to equilibrate a Hermitian positive definite matrix A and reduce its condition number (with respect to the two-norm).  S contains the scale factors, S(i) = 1/sqrt(A(i,i)), chosen so that the scaled matrix B with elements B(i,j) = S(i)∗A(i,j)∗S(j) has ones on the diagonal.  This choice of S puts the condition number of B within a factor N of the smallest possible condition number over all possible diagonal scalings. 
 

ARGUMENTS

N       (input) INTEGER
The order of the matrix A.  N >= 0.

A       (input) COMPLEX∗16 array, dimension (LDA,N)
The N-by-N Hermitian positive definite matrix whose scaling factors are to be computed.  Only the diagonal elements of A are referenced.

LDA     (input) INTEGER
The leading dimension of the array A.  LDA >= max(1,N).

S       (output) DOUBLE PRECISION array, dimension (N)
If INFO = 0, S contains the scale factors for A.

SCOND   (output) DOUBLE PRECISION
If INFO = 0, S contains the ratio of the smallest S(i) to the largest S(i).  If SCOND >= 0.1 and AMAX is neither too large nor too small, it is not worth scaling by S.

AMAX    (output) DOUBLE PRECISION
Absolute value of largest matrix element.  If AMAX is very close to overflow or very close to underflow, the matrix should be scaled.

INFO    (output) INTEGER
= 0:  successful exit
< 0:  if INFO = -i, the i-th argument had an illegal value
> 0:  if INFO = i, the i-th diagonal element is nonpositive.

Sun, Inc.  —  Last change: 20 Sep 1996