d3/d0a/dlasdt_8f_source.html

00001       SUBROUTINE DLASDT( N, LVL, ND, INODE, NDIML, NDIMR, MSUB )
00002 *
00003 *  -- LAPACK auxiliary routine (version 3.2.2) --
00004 *  -- LAPACK is a software package provided by Univ. of Tennessee,    --
00005 *  -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..--
00006 *     June 2010
00007 *
00008 *     .. Scalar Arguments ..
00009       INTEGER            LVL, MSUB, N, ND
00010 *     ..
00011 *     .. Array Arguments ..
00012       INTEGER            INODE( * ), NDIML( * ), NDIMR( * )
00013 *     ..
00014 *
00015 *  Purpose
00016 *  =======
00017 *
00018 *  DLASDT creates a tree of subproblems for bidiagonal divide and
00019 *  conquer.
00020 *
00021 *  Arguments
00022 *  =========
00023 *
00024 *   N      (input) INTEGER
00025 *          On entry, the number of diagonal elements of the
00026 *          bidiagonal matrix.
00027 *
00028 *   LVL    (output) INTEGER
00029 *          On exit, the number of levels on the computation tree.
00030 *
00031 *   ND     (output) INTEGER
00032 *          On exit, the number of nodes on the tree.
00033 *
00034 *   INODE  (output) INTEGER array, dimension ( N )
00035 *          On exit, centers of subproblems.
00036 *
00037 *   NDIML  (output) INTEGER array, dimension ( N )
00038 *          On exit, row dimensions of left children.
00039 *
00040 *   NDIMR  (output) INTEGER array, dimension ( N )
00041 *          On exit, row dimensions of right children.
00042 *
00043 *   MSUB   (input) INTEGER
00044 *          On entry, the maximum row dimension each subproblem at the
00045 *          bottom of the tree can be of.
00046 *
00047 *  Further Details
00048 *  ===============
00049 *
00050 *  Based on contributions by
00051 *     Ming Gu and Huan Ren, Computer Science Division, University of
00052 *     California at Berkeley, USA
00053 *
00054 *  =====================================================================
00055 *
00056 *     .. Parameters ..
00057       DOUBLE PRECISION   TWO
00058       PARAMETER          ( TWO = 2.0D+0 )
00059 *     ..
00060 *     .. Local Scalars ..
00061       INTEGER            I, IL, IR, LLST, MAXN, NCRNT, NLVL
00062       DOUBLE PRECISION   TEMP
00063 *     ..
00064 *     .. Intrinsic Functions ..
00065       INTRINSIC          DBLE, INT, LOG, MAX
00066 *     ..
00067 *     .. Executable Statements ..
00068 *
00069 *     Find the number of levels on the tree.
00070 *
00071       MAXN = MAX( 1, N )
00072       TEMP = LOG( DBLE( MAXN ) / DBLE( MSUB+1 ) ) / LOG( TWO )
00073       LVL = INT( TEMP ) + 1
00074 *
00075       I = N / 2
00076       INODE( 1 ) = I + 1
00077       NDIML( 1 ) = I
00078       NDIMR( 1 ) = N - I - 1
00079       IL = 0
00080       IR = 1
00081       LLST = 1
00082       DO 20 NLVL = 1, LVL - 1
00083 *
00084 *        Constructing the tree at (NLVL+1)-st level. The number of
00085 *        nodes created on this level is LLST * 2.
00086 *
00087          DO 10 I = 0, LLST - 1
00088             IL = IL + 2
00089             IR = IR + 2
00090             NCRNT = LLST + I
00091             NDIML( IL ) = NDIML( NCRNT ) / 2
00092             NDIMR( IL ) = NDIML( NCRNT ) - NDIML( IL ) - 1
00093             INODE( IL ) = INODE( NCRNT ) - NDIMR( IL ) - 1
00094             NDIML( IR ) = NDIMR( NCRNT ) / 2
00095             NDIMR( IR ) = NDIMR( NCRNT ) - NDIML( IR ) - 1
00096             INODE( IR ) = INODE( NCRNT ) + NDIML( IR ) + 1
00097    10    CONTINUE
00098          LLST = LLST*2
00099    20 CONTINUE
00100       ND = LLST*2 - 1
00101 *
00102       RETURN
00103 *
00104 *     End of DLASDT
00105 *
00106       END