1DLARRR(1)           LAPACK auxiliary routine (version 3.2)           DLARRR(1)
2
3
4

NAME

6       DLARRR  -  tests  to  decide whether the symmetric tridiagonal matrix T
7       warrants expensive computations which guarantee high relative  accuracy
8       in the eigenvalues
9

SYNOPSIS

11       SUBROUTINE DLARRR( N, D, E, INFO )
12
13           INTEGER        N, INFO
14
15           DOUBLE         PRECISION D( * ), E( * )
16

PURPOSE

18       Perform tests to decide whether the symmetric tridiagonal matrix T war‐
19       rants expensive computations which guarantee high relative accuracy  in
20       the eigenvalues.
21

ARGUMENTS

23       N       (input) INTEGER
24               The order of the matrix. N > 0.
25
26       D       (input) DOUBLE PRECISION array, dimension (N)
27               The N diagonal elements of the tridiagonal matrix T.
28
29       E       (input/output) DOUBLE PRECISION array, dimension (N)
30               On  entry, the first (N-1) entries contain the subdiagonal ele‐
31               ments of the tridiagonal matrix T; E(N) is set to ZERO.
32
33       INFO    (output) INTEGER
34               INFO = 0(default) : the matrix warrants computations preserving
35               relative  accuracy.   INFO  =  1          : the matrix warrants
36               computations guaranteeing only absolute accuracy.
37

FURTHER DETAILS

39       Based on contributions by
40          Beresford Parlett, University of California, Berkeley, USA
41          Jim Demmel, University of California, Berkeley, USA
42          Inderjit Dhillon, University of Texas, Austin, USA
43          Osni Marques, LBNL/NERSC, USA
44          Christof Voemel, University of California, Berkeley, USA
45
46
47
48 LAPACK auxiliary routine (versionNo3v.e2m)ber 2008                       DLARRR(1)