File: iparmq

package info (click to toggle)
ruby-lapack 1.5-1
  • links: PTS, VCS
  • area: main
  • in suites: wheezy
  • size: 28,336 kB
  • sloc: ansic: 190,568; ruby: 3,837; makefile: 4
file content (177 lines) | stat: -rw-r--r-- 8,068 bytes parent folder | download | duplicates (5)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
--- 
:name: iparmq
:md5sum: 73ec37559601ded71e0b8df251efd10e
:category: :function
:type: integer
:arguments: 
- ispec: 
    :type: integer
    :intent: input
- name: 
    :type: char
    :intent: input
- opts: 
    :type: char
    :intent: input
- n: 
    :type: integer
    :intent: input
- ilo: 
    :type: integer
    :intent: input
- ihi: 
    :type: integer
    :intent: input
- lwork: 
    :type: integer
    :intent: input
:substitutions: {}

:fortran_help: "      INTEGER FUNCTION IPARMQ( ISPEC, NAME, OPTS, N, ILO, IHI, LWORK )\n\n\
  *  Purpose\n\
  *  =======\n\
  *\n\
  *       This program sets problem and machine dependent parameters\n\
  *       useful for xHSEQR and its subroutines. It is called whenever \n\
  *       ILAENV is called with 12 <= ISPEC <= 16\n\
  *\n\n\
  *  Arguments\n\
  *  =========\n\
  *\n\
  *       ISPEC  (input) integer scalar\n\
  *              ISPEC specifies which tunable parameter IPARMQ should\n\
  *              return.\n\
  *\n\
  *              ISPEC=12: (INMIN)  Matrices of order nmin or less\n\
  *                        are sent directly to xLAHQR, the implicit\n\
  *                        double shift QR algorithm.  NMIN must be\n\
  *                        at least 11.\n\
  *\n\
  *              ISPEC=13: (INWIN)  Size of the deflation window.\n\
  *                        This is best set greater than or equal to\n\
  *                        the number of simultaneous shifts NS.\n\
  *                        Larger matrices benefit from larger deflation\n\
  *                        windows.\n\
  *\n\
  *              ISPEC=14: (INIBL) Determines when to stop nibbling and\n\
  *                        invest in an (expensive) multi-shift QR sweep.\n\
  *                        If the aggressive early deflation subroutine\n\
  *                        finds LD converged eigenvalues from an order\n\
  *                        NW deflation window and LD.GT.(NW*NIBBLE)/100,\n\
  *                        then the next QR sweep is skipped and early\n\
  *                        deflation is applied immediately to the\n\
  *                        remaining active diagonal block.  Setting\n\
  *                        IPARMQ(ISPEC=14) = 0 causes TTQRE to skip a\n\
  *                        multi-shift QR sweep whenever early deflation\n\
  *                        finds a converged eigenvalue.  Setting\n\
  *                        IPARMQ(ISPEC=14) greater than or equal to 100\n\
  *                        prevents TTQRE from skipping a multi-shift\n\
  *                        QR sweep.\n\
  *\n\
  *              ISPEC=15: (NSHFTS) The number of simultaneous shifts in\n\
  *                        a multi-shift QR iteration.\n\
  *\n\
  *              ISPEC=16: (IACC22) IPARMQ is set to 0, 1 or 2 with the\n\
  *                        following meanings.\n\
  *                        0:  During the multi-shift QR sweep,\n\
  *                            xLAQR5 does not accumulate reflections and\n\
  *                            does not use matrix-matrix multiply to\n\
  *                            update the far-from-diagonal matrix\n\
  *                            entries.\n\
  *                        1:  During the multi-shift QR sweep,\n\
  *                            xLAQR5 and/or xLAQRaccumulates reflections and uses\n\
  *                            matrix-matrix multiply to update the\n\
  *                            far-from-diagonal matrix entries.\n\
  *                        2:  During the multi-shift QR sweep.\n\
  *                            xLAQR5 accumulates reflections and takes\n\
  *                            advantage of 2-by-2 block structure during\n\
  *                            matrix-matrix multiplies.\n\
  *                        (If xTRMM is slower than xGEMM, then\n\
  *                        IPARMQ(ISPEC=16)=1 may be more efficient than\n\
  *                        IPARMQ(ISPEC=16)=2 despite the greater level of\n\
  *                        arithmetic work implied by the latter choice.)\n\
  *\n\
  *       NAME    (input) character string\n\
  *               Name of the calling subroutine\n\
  *\n\
  *       OPTS    (input) character string\n\
  *               This is a concatenation of the string arguments to\n\
  *               TTQRE.\n\
  *\n\
  *       N       (input) integer scalar\n\
  *               N is the order of the Hessenberg matrix H.\n\
  *\n\
  *       ILO     (input) INTEGER\n\
  *       IHI     (input) INTEGER\n\
  *               It is assumed that H is already upper triangular\n\
  *               in rows and columns 1:ILO-1 and IHI+1:N.\n\
  *\n\
  *       LWORK   (input) integer scalar\n\
  *               The amount of workspace available.\n\
  *\n\n\
  *  Further Details\n\
  *  ===============\n\
  *\n\
  *       Little is known about how best to choose these parameters.\n\
  *       It is possible to use different values of the parameters\n\
  *       for each of CHSEQR, DHSEQR, SHSEQR and ZHSEQR.\n\
  *\n\
  *       It is probably best to choose different parameters for\n\
  *       different matrices and different parameters at different\n\
  *       times during the iteration, but this has not been\n\
  *       implemented --- yet.\n\
  *\n\
  *\n\
  *       The best choices of most of the parameters depend\n\
  *       in an ill-understood way on the relative execution\n\
  *       rate of xLAQR3 and xLAQR5 and on the nature of each\n\
  *       particular eigenvalue problem.  Experiment may be the\n\
  *       only practical way to determine which choices are most\n\
  *       effective.\n\
  *\n\
  *       Following is a list of default values supplied by IPARMQ.\n\
  *       These defaults may be adjusted in order to attain better\n\
  *       performance in any particular computational environment.\n\
  *\n\
  *       IPARMQ(ISPEC=12) The xLAHQR vs xLAQR0 crossover point.\n\
  *                        Default: 75. (Must be at least 11.)\n\
  *\n\
  *       IPARMQ(ISPEC=13) Recommended deflation window size.\n\
  *                        This depends on ILO, IHI and NS, the\n\
  *                        number of simultaneous shifts returned\n\
  *                        by IPARMQ(ISPEC=15).  The default for\n\
  *                        (IHI-ILO+1).LE.500 is NS.  The default\n\
  *                        for (IHI-ILO+1).GT.500 is 3*NS/2.\n\
  *\n\
  *       IPARMQ(ISPEC=14) Nibble crossover point.  Default: 14.\n\
  *\n\
  *       IPARMQ(ISPEC=15) Number of simultaneous shifts, NS.\n\
  *                        a multi-shift QR iteration.\n\
  *\n\
  *                        If IHI-ILO+1 is ...\n\
  *\n\
  *                        greater than      ...but less    ... the\n\
  *                        or equal to ...      than        default is\n\
  *\n\
  *                                0               30       NS =   2+\n\
  *                               30               60       NS =   4+\n\
  *                               60              150       NS =  10\n\
  *                              150              590       NS =  **\n\
  *                              590             3000       NS =  64\n\
  *                             3000             6000       NS = 128\n\
  *                             6000             infinity   NS = 256\n\
  *\n\
  *                    (+)  By default matrices of this order are\n\
  *                         passed to the implicit double shift routine\n\
  *                         xLAHQR.  See IPARMQ(ISPEC=12) above.   These\n\
  *                         values of NS are used only in case of a rare\n\
  *                         xLAHQR failure.\n\
  *\n\
  *                    (**) The asterisks (**) indicate an ad-hoc\n\
  *                         function increasing from 10 to 64.\n\
  *\n\
  *       IPARMQ(ISPEC=16) Select structured matrix multiply.\n\
  *                        (See ISPEC=16 above for details.)\n\
  *                        Default: 3.\n\
  *\n\
  *     ================================================================\n"