Return to dlabad.f CVS log

Up to [local] / rpl / lapack / lapack

Annotation of rpl/lapack/lapack/dlabad.f, revision 1.5

1.1       bertrand    1:       SUBROUTINE DLABAD( SMALL, LARGE )
                      2: *
                      3: *  -- LAPACK auxiliary routine (version 3.2) --
                      4: *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
                      5: *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
                      6: *     November 2006
                      7: *
                      8: *     .. Scalar Arguments ..
                      9:       DOUBLE PRECISION   LARGE, SMALL
                     10: *     ..
                     11: *
                     12: *  Purpose
                     13: *  =======
                     14: *
                     15: *  DLABAD takes as input the values computed by DLAMCH for underflow and
                     16: *  overflow, and returns the square root of each of these values if the
                     17: *  log of LARGE is sufficiently large.  This subroutine is intended to
                     18: *  identify machines with a large exponent range, such as the Crays, and
                     19: *  redefine the underflow and overflow limits to be the square roots of
                     20: *  the values computed by DLAMCH.  This subroutine is needed because
                     21: *  DLAMCH does not compensate for poor arithmetic in the upper half of
                     22: *  the exponent range, as is found on a Cray.
                     23: *
                     24: *  Arguments
                     25: *  =========
                     26: *
                     27: *  SMALL   (input/output) DOUBLE PRECISION
                     28: *          On entry, the underflow threshold as computed by DLAMCH.
                     29: *          On exit, if LOG10(LARGE) is sufficiently large, the square
                     30: *          root of SMALL, otherwise unchanged.
                     31: *
                     32: *  LARGE   (input/output) DOUBLE PRECISION
                     33: *          On entry, the overflow threshold as computed by DLAMCH.
                     34: *          On exit, if LOG10(LARGE) is sufficiently large, the square
                     35: *          root of LARGE, otherwise unchanged.
                     36: *
                     37: *  =====================================================================
                     38: *
                     39: *     .. Intrinsic Functions ..
                     40:       INTRINSIC          LOG10, SQRT
                     41: *     ..
                     42: *     .. Executable Statements ..
                     43: *
                     44: *     If it looks like we're on a Cray, take the square root of
                     45: *     SMALL and LARGE to avoid overflow and underflow problems.
                     46: *
                     47:       IF( LOG10( LARGE ).GT.2000.D0 ) THEN
                     48:          SMALL = SQRT( SMALL )
                     49:          LARGE = SQRT( LARGE )
                     50:       END IF
                     51: *
                     52:       RETURN
                     53: *
                     54: *     End of DLABAD
                     55: *
                     56:       END