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
|
#include "rb_lapack.h"
extern VOID dlasd1_(integer* nl, integer* nr, integer* sqre, doublereal* d, doublereal* alpha, doublereal* beta, doublereal* u, integer* ldu, doublereal* vt, integer* ldvt, integer* idxq, integer* iwork, doublereal* work, integer* info);
static VALUE
rblapack_dlasd1(int argc, VALUE *argv, VALUE self){
VALUE rblapack_nl;
integer nl;
VALUE rblapack_nr;
integer nr;
VALUE rblapack_sqre;
integer sqre;
VALUE rblapack_d;
doublereal *d;
VALUE rblapack_alpha;
doublereal alpha;
VALUE rblapack_beta;
doublereal beta;
VALUE rblapack_u;
doublereal *u;
VALUE rblapack_vt;
doublereal *vt;
VALUE rblapack_idxq;
integer *idxq;
VALUE rblapack_info;
integer info;
VALUE rblapack_d_out__;
doublereal *d_out__;
VALUE rblapack_u_out__;
doublereal *u_out__;
VALUE rblapack_vt_out__;
doublereal *vt_out__;
integer *iwork;
doublereal *work;
integer n;
integer ldu;
integer ldvt;
integer m;
VALUE rblapack_options;
if (argc > 0 && TYPE(argv[argc-1]) == T_HASH) {
argc--;
rblapack_options = argv[argc];
if (rb_hash_aref(rblapack_options, sHelp) == Qtrue) {
printf("%s\n", "USAGE:\n idxq, info, d, alpha, beta, u, vt = NumRu::Lapack.dlasd1( nl, nr, sqre, d, alpha, beta, u, vt, [:usage => usage, :help => help])\n\n\nFORTRAN MANUAL\n SUBROUTINE DLASD1( NL, NR, SQRE, D, ALPHA, BETA, U, LDU, VT, LDVT, IDXQ, IWORK, WORK, INFO )\n\n* Purpose\n* =======\n*\n* DLASD1 computes the SVD of an upper bidiagonal N-by-M matrix B,\n* where N = NL + NR + 1 and M = N + SQRE. DLASD1 is called from DLASD0.\n*\n* A related subroutine DLASD7 handles the case in which the singular\n* values (and the singular vectors in factored form) are desired.\n*\n* DLASD1 computes the SVD as follows:\n*\n* ( D1(in) 0 0 0 )\n* B = U(in) * ( Z1' a Z2' b ) * VT(in)\n* ( 0 0 D2(in) 0 )\n*\n* = U(out) * ( D(out) 0) * VT(out)\n*\n* where Z' = (Z1' a Z2' b) = u' VT', and u is a vector of dimension M\n* with ALPHA and BETA in the NL+1 and NL+2 th entries and zeros\n* elsewhere; and the entry b is empty if SQRE = 0.\n*\n* The left singular vectors of the original matrix are stored in U, and\n* the transpose of the right singular vectors are stored in VT, and the\n* singular values are in D. The algorithm consists of three stages:\n*\n* The first stage consists of deflating the size of the problem\n* when there are multiple singular values or when there are zeros in\n* the Z vector. For each such occurrence the dimension of the\n* secular equation problem is reduced by one. This stage is\n* performed by the routine DLASD2.\n*\n* The second stage consists of calculating the updated\n* singular values. This is done by finding the square roots of the\n* roots of the secular equation via the routine DLASD4 (as called\n* by DLASD3). This routine also calculates the singular vectors of\n* the current problem.\n*\n* The final stage consists of computing the updated singular vectors\n* directly using the updated singular values. The singular vectors\n* for the current problem are multiplied with the singular vectors\n* from the overall problem.\n*\n\n* Arguments\n* =========\n*\n* NL (input) INTEGER\n* The row dimension of the upper block. NL >= 1.\n*\n* NR (input) INTEGER\n* The row dimension of the lower block. NR >= 1.\n*\n* SQRE (input) INTEGER\n* = 0: the lower block is an NR-by-NR square matrix.\n* = 1: the lower block is an NR-by-(NR+1) rectangular matrix.\n*\n* The bidiagonal matrix has row dimension N = NL + NR + 1,\n* and column dimension M = N + SQRE.\n*\n* D (input/output) DOUBLE PRECISION array,\n* dimension (N = NL+NR+1).\n* On entry D(1:NL,1:NL) contains the singular values of the\n* upper block; and D(NL+2:N) contains the singular values of\n* the lower block. On exit D(1:N) contains the singular values\n* of the modified matrix.\n*\n* ALPHA (input/output) DOUBLE PRECISION\n* Contains the diagonal element associated with the added row.\n*\n* BETA (input/output) DOUBLE PRECISION\n* Contains the off-diagonal element associated with the added\n* row.\n*\n* U (input/output) DOUBLE PRECISION array, dimension(LDU,N)\n* On entry U(1:NL, 1:NL) contains the left singular vectors of\n* the upper block; U(NL+2:N, NL+2:N) contains the left singular\n* vectors of the lower block. On exit U contains the left\n* singular vectors of the bidiagonal matrix.\n*\n* LDU (input) INTEGER\n* The leading dimension of the array U. LDU >= max( 1, N ).\n*\n* VT (input/output) DOUBLE PRECISION array, dimension(LDVT,M)\n* where M = N + SQRE.\n* On entry VT(1:NL+1, 1:NL+1)' contains the right singular\n* vectors of the upper block; VT(NL+2:M, NL+2:M)' contains\n* the right singular vectors of the lower block. On exit\n* VT' contains the right singular vectors of the\n* bidiagonal matrix.\n*\n* LDVT (input) INTEGER\n* The leading dimension of the array VT. LDVT >= max( 1, M ).\n*\n* IDXQ (output) INTEGER array, dimension(N)\n* This contains the permutation which will reintegrate the\n* subproblem just solved back into sorted order, i.e.\n* D( IDXQ( I = 1, N ) ) will be in ascending order.\n*\n* IWORK (workspace) INTEGER array, dimension( 4 * N )\n*\n* WORK (workspace) DOUBLE PRECISION array, dimension( 3*M**2 + 2*M )\n*\n* INFO (output) INTEGER\n* = 0: successful exit.\n* < 0: if INFO = -i, the i-th argument had an illegal value.\n* > 0: if INFO = 1, a singular value did not converge\n*\n\n* Further Details\n* ===============\n*\n* Based on contributions by\n* Ming Gu and Huan Ren, Computer Science Division, University of\n* California at Berkeley, USA\n*\n* =====================================================================\n*\n\n");
return Qnil;
}
if (rb_hash_aref(rblapack_options, sUsage) == Qtrue) {
printf("%s\n", "USAGE:\n idxq, info, d, alpha, beta, u, vt = NumRu::Lapack.dlasd1( nl, nr, sqre, d, alpha, beta, u, vt, [:usage => usage, :help => help])\n");
return Qnil;
}
} else
rblapack_options = Qnil;
if (argc != 8 && argc != 8)
rb_raise(rb_eArgError,"wrong number of arguments (%d for 8)", argc);
rblapack_nl = argv[0];
rblapack_nr = argv[1];
rblapack_sqre = argv[2];
rblapack_d = argv[3];
rblapack_alpha = argv[4];
rblapack_beta = argv[5];
rblapack_u = argv[6];
rblapack_vt = argv[7];
if (argc == 8) {
} else if (rblapack_options != Qnil) {
} else {
}
nl = NUM2INT(rblapack_nl);
sqre = NUM2INT(rblapack_sqre);
alpha = NUM2DBL(rblapack_alpha);
nr = NUM2INT(rblapack_nr);
beta = NUM2DBL(rblapack_beta);
n = nl+nr+1;
if (!NA_IsNArray(rblapack_d))
rb_raise(rb_eArgError, "d (4th argument) must be NArray");
if (NA_RANK(rblapack_d) != 1)
rb_raise(rb_eArgError, "rank of d (4th argument) must be %d", 1);
if (NA_SHAPE0(rblapack_d) != n)
rb_raise(rb_eRuntimeError, "shape 0 of d must be nl+nr+1");
if (NA_TYPE(rblapack_d) != NA_DFLOAT)
rblapack_d = na_change_type(rblapack_d, NA_DFLOAT);
d = NA_PTR_TYPE(rblapack_d, doublereal*);
m = n + sqre;
if (!NA_IsNArray(rblapack_u))
rb_raise(rb_eArgError, "u (7th argument) must be NArray");
if (NA_RANK(rblapack_u) != 2)
rb_raise(rb_eArgError, "rank of u (7th argument) must be %d", 2);
ldu = NA_SHAPE0(rblapack_u);
if (NA_SHAPE1(rblapack_u) != n)
rb_raise(rb_eRuntimeError, "shape 1 of u must be nl+nr+1");
if (NA_TYPE(rblapack_u) != NA_DFLOAT)
rblapack_u = na_change_type(rblapack_u, NA_DFLOAT);
u = NA_PTR_TYPE(rblapack_u, doublereal*);
if (!NA_IsNArray(rblapack_vt))
rb_raise(rb_eArgError, "vt (8th argument) must be NArray");
if (NA_RANK(rblapack_vt) != 2)
rb_raise(rb_eArgError, "rank of vt (8th argument) must be %d", 2);
ldvt = NA_SHAPE0(rblapack_vt);
if (NA_SHAPE1(rblapack_vt) != m)
rb_raise(rb_eRuntimeError, "shape 1 of vt must be n + sqre");
if (NA_TYPE(rblapack_vt) != NA_DFLOAT)
rblapack_vt = na_change_type(rblapack_vt, NA_DFLOAT);
vt = NA_PTR_TYPE(rblapack_vt, doublereal*);
{
na_shape_t shape[1];
shape[0] = n;
rblapack_idxq = na_make_object(NA_LINT, 1, shape, cNArray);
}
idxq = NA_PTR_TYPE(rblapack_idxq, integer*);
{
na_shape_t shape[1];
shape[0] = n;
rblapack_d_out__ = na_make_object(NA_DFLOAT, 1, shape, cNArray);
}
d_out__ = NA_PTR_TYPE(rblapack_d_out__, doublereal*);
MEMCPY(d_out__, d, doublereal, NA_TOTAL(rblapack_d));
rblapack_d = rblapack_d_out__;
d = d_out__;
{
na_shape_t shape[2];
shape[0] = ldu;
shape[1] = n;
rblapack_u_out__ = na_make_object(NA_DFLOAT, 2, shape, cNArray);
}
u_out__ = NA_PTR_TYPE(rblapack_u_out__, doublereal*);
MEMCPY(u_out__, u, doublereal, NA_TOTAL(rblapack_u));
rblapack_u = rblapack_u_out__;
u = u_out__;
{
na_shape_t shape[2];
shape[0] = ldvt;
shape[1] = m;
rblapack_vt_out__ = na_make_object(NA_DFLOAT, 2, shape, cNArray);
}
vt_out__ = NA_PTR_TYPE(rblapack_vt_out__, doublereal*);
MEMCPY(vt_out__, vt, doublereal, NA_TOTAL(rblapack_vt));
rblapack_vt = rblapack_vt_out__;
vt = vt_out__;
iwork = ALLOC_N(integer, (4 * n));
work = ALLOC_N(doublereal, (3*pow(m,2) + 2*m));
dlasd1_(&nl, &nr, &sqre, d, &alpha, &beta, u, &ldu, vt, &ldvt, idxq, iwork, work, &info);
free(iwork);
free(work);
rblapack_info = INT2NUM(info);
rblapack_alpha = rb_float_new((double)alpha);
rblapack_beta = rb_float_new((double)beta);
return rb_ary_new3(7, rblapack_idxq, rblapack_info, rblapack_d, rblapack_alpha, rblapack_beta, rblapack_u, rblapack_vt);
}
void
init_lapack_dlasd1(VALUE mLapack, VALUE sH, VALUE sU, VALUE zero){
sHelp = sH;
sUsage = sU;
rblapack_ZERO = zero;
rb_define_module_function(mLapack, "dlasd1", rblapack_dlasd1, -1);
}
|
