{D,Z}BBCSD: stricter zero criterion

The tolerance below which matrix entries are considered zero is
tightened to avoid insufficiently accurate singular vectors.

fixes Reference-LAPACK#965

SRC/dbbcsd.f

@@ -302,7 +302,7 @@
 *  =========================
 *>
 *> \verbatim
-*>  TOLMUL  DOUBLE PRECISION, default = MAX(10,MIN(100,EPS**(-1/8)))
+*>  TOLMUL  DOUBLE PRECISION, default = 10
 *>          TOLMUL controls the convergence criterion of the QR loop.
 *>          Angles THETA(i), PHI(i) are rounded to 0 or PI/2 when they
 *>          are within TOLMUL*EPS of either bound.
@@ -451,7 +451,7 @@ SUBROUTINE DBBCSD( JOBU1, JOBU2, JOBV1T, JOBV2T, TRANS, M, P, Q,
 *
       EPS = DLAMCH( 'Epsilon' )
       UNFL = DLAMCH( 'Safe minimum' )
-      TOLMUL = MAX( TEN, MIN( HUNDRED, EPS**MEIGHTH ) )
+      TOLMUL = TEN
       TOL = TOLMUL*EPS
       THRESH = MAX( TOL, MAXITR*Q*Q*UNFL )
 *

SRC/zbbcsd.f

@@ -302,7 +302,7 @@
 *  =========================
 *>
 *> \verbatim
-*>  TOLMUL  DOUBLE PRECISION, default = MAX(10,MIN(100,EPS**(-1/8)))
+*>  TOLMUL  DOUBLE PRECISION, default = 10
 *>          TOLMUL controls the convergence criterion of the QR loop.
 *>          Angles THETA(i), PHI(i) are rounded to 0 or PI/2 when they
 *>          are within TOLMUL*EPS of either bound.
@@ -450,7 +450,7 @@ SUBROUTINE ZBBCSD( JOBU1, JOBU2, JOBV1T, JOBV2T, TRANS, M, P, Q,
 *
       EPS = DLAMCH( 'Epsilon' )
       UNFL = DLAMCH( 'Safe minimum' )
-      TOLMUL = MAX( TEN, MIN( HUNDRED, EPS**MEIGHTH ) )
+      TOLMUL = TEN
       TOL = TOLMUL*EPS
       THRESH = MAX( TOL, MAXITR*Q*Q*UNFL )
 *