![]() |
LAPACK
3.7.0
LAPACK: Linear Algebra PACKage
|
| subroutine dget31 | ( | double precision | RMAX, |
| integer | LMAX, | ||
| integer, dimension( 2 ) | NINFO, | ||
| integer | KNT | ||
| ) |
DGET31
DGET31 tests DLALN2, a routine for solving
(ca A - w D)X = sB
where A is an NA by NA matrix (NA=1 or 2 only), w is a real (NW=1) or
complex (NW=2) constant, ca is a real constant, D is an NA by NA real
diagonal matrix, and B is an NA by NW matrix (when NW=2 the second
column of B contains the imaginary part of the solution). The code
returns X and s, where s is a scale factor, less than or equal to 1,
which is chosen to avoid overflow in X.
If any singular values of ca A-w D are less than another input
parameter SMIN, they are perturbed up to SMIN.
The test condition is that the scaled residual
norm( (ca A-w D)*X - s*B ) /
( max( ulp*norm(ca A-w D), SMIN )*norm(X) )
should be on the order of 1. Here, ulp is the machine precision.
Also, it is verified that SCALE is less than or equal to 1, and that
XNORM = infinity-norm(X). | [out] | RMAX | RMAX is DOUBLE PRECISION
Value of the largest test ratio. |
| [out] | LMAX | LMAX is INTEGER
Example number where largest test ratio achieved. |
| [out] | NINFO | NINFO is INTEGER array, dimension (3)
NINFO(1) = number of examples with INFO less than 0
NINFO(2) = number of examples with INFO greater than 0 |
| [out] | KNT | KNT is INTEGER
Total number of examples tested. |