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
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
|
---
:name: sggevx
:md5sum: 7044c66a49c1acd580038604dab023c1
:category: :subroutine
:arguments:
- balanc:
:type: char
:intent: input
- jobvl:
:type: char
:intent: input
- jobvr:
:type: char
:intent: input
- sense:
:type: char
:intent: input
- n:
:type: integer
:intent: input
- a:
:type: real
:intent: input/output
:dims:
- lda
- n
- lda:
:type: integer
:intent: input
- b:
:type: real
:intent: input/output
:dims:
- ldb
- n
- ldb:
:type: integer
:intent: input
- alphar:
:type: real
:intent: output
:dims:
- n
- alphai:
:type: real
:intent: output
:dims:
- n
- beta:
:type: real
:intent: output
:dims:
- n
- vl:
:type: real
:intent: output
:dims:
- ldvl
- n
- ldvl:
:type: integer
:intent: input
- vr:
:type: real
:intent: output
:dims:
- ldvr
- n
- ldvr:
:type: integer
:intent: input
- ilo:
:type: integer
:intent: output
- ihi:
:type: integer
:intent: output
- lscale:
:type: real
:intent: output
:dims:
- n
- rscale:
:type: real
:intent: output
:dims:
- n
- abnrm:
:type: real
:intent: output
- bbnrm:
:type: real
:intent: output
- rconde:
:type: real
:intent: output
:dims:
- n
- rcondv:
:type: real
:intent: output
:dims:
- n
- work:
:type: real
:intent: output
:dims:
- MAX(1,lwork)
- lwork:
:type: integer
:intent: input
:option: true
:default: "(lsame_(&balanc,\"S\")||lsame_(&balanc,\"B\")||lsame_(&jobvl,\"V\")||lsame_(&jobvr,\"V\")) ? 6*n : lsame_(&sense,\"E\") ? 10*n : (lsame_(&sense,\"V\")||lsame_(&sense,\"B\")) ? 2*n*n+8*n+16 : 2*n"
- iwork:
:type: integer
:intent: workspace
:dims:
- "lsame_(&sense,\"E\") ? 0 : n+6"
- bwork:
:type: logical
:intent: workspace
:dims:
- "lsame_(&sense,\"N\") ? 0 : n"
- info:
:type: integer
:intent: output
:substitutions:
ldvr: "lsame_(&jobvr,\"V\") ? n : 1"
ldvl: "lsame_(&jobvl,\"V\") ? n : 1"
:fortran_help: " SUBROUTINE SGGEVX( BALANC, JOBVL, JOBVR, SENSE, N, A, LDA, B, LDB, ALPHAR, ALPHAI, BETA, VL, LDVL, VR, LDVR, ILO, IHI, LSCALE, RSCALE, ABNRM, BBNRM, RCONDE, RCONDV, WORK, LWORK, IWORK, BWORK, INFO )\n\n\
* Purpose\n\
* =======\n\
*\n\
* SGGEVX computes for a pair of N-by-N real nonsymmetric matrices (A,B)\n\
* the generalized eigenvalues, and optionally, the left and/or right\n\
* generalized eigenvectors.\n\
*\n\
* Optionally also, it computes a balancing transformation to improve\n\
* the conditioning of the eigenvalues and eigenvectors (ILO, IHI,\n\
* LSCALE, RSCALE, ABNRM, and BBNRM), reciprocal condition numbers for\n\
* the eigenvalues (RCONDE), and reciprocal condition numbers for the\n\
* right eigenvectors (RCONDV).\n\
*\n\
* A generalized eigenvalue for a pair of matrices (A,B) is a scalar\n\
* lambda or a ratio alpha/beta = lambda, such that A - lambda*B is\n\
* singular. It is usually represented as the pair (alpha,beta), as\n\
* there is a reasonable interpretation for beta=0, and even for both\n\
* being zero.\n\
*\n\
* The right eigenvector v(j) corresponding to the eigenvalue lambda(j)\n\
* of (A,B) satisfies\n\
*\n\
* A * v(j) = lambda(j) * B * v(j) .\n\
*\n\
* The left eigenvector u(j) corresponding to the eigenvalue lambda(j)\n\
* of (A,B) satisfies\n\
*\n\
* u(j)**H * A = lambda(j) * u(j)**H * B.\n\
*\n\
* where u(j)**H is the conjugate-transpose of u(j).\n\
*\n\
*\n\n\
* Arguments\n\
* =========\n\
*\n\
* BALANC (input) CHARACTER*1\n\
* Specifies the balance option to be performed.\n\
* = 'N': do not diagonally scale or permute;\n\
* = 'P': permute only;\n\
* = 'S': scale only;\n\
* = 'B': both permute and scale.\n\
* Computed reciprocal condition numbers will be for the\n\
* matrices after permuting and/or balancing. Permuting does\n\
* not change condition numbers (in exact arithmetic), but\n\
* balancing does.\n\
*\n\
* JOBVL (input) CHARACTER*1\n\
* = 'N': do not compute the left generalized eigenvectors;\n\
* = 'V': compute the left generalized eigenvectors.\n\
*\n\
* JOBVR (input) CHARACTER*1\n\
* = 'N': do not compute the right generalized eigenvectors;\n\
* = 'V': compute the right generalized eigenvectors.\n\
*\n\
* SENSE (input) CHARACTER*1\n\
* Determines which reciprocal condition numbers are computed.\n\
* = 'N': none are computed;\n\
* = 'E': computed for eigenvalues only;\n\
* = 'V': computed for eigenvectors only;\n\
* = 'B': computed for eigenvalues and eigenvectors.\n\
*\n\
* N (input) INTEGER\n\
* The order of the matrices A, B, VL, and VR. N >= 0.\n\
*\n\
* A (input/output) REAL array, dimension (LDA, N)\n\
* On entry, the matrix A in the pair (A,B).\n\
* On exit, A has been overwritten. If JOBVL='V' or JOBVR='V'\n\
* or both, then A contains the first part of the real Schur\n\
* form of the \"balanced\" versions of the input A and B.\n\
*\n\
* LDA (input) INTEGER\n\
* The leading dimension of A. LDA >= max(1,N).\n\
*\n\
* B (input/output) REAL array, dimension (LDB, N)\n\
* On entry, the matrix B in the pair (A,B).\n\
* On exit, B has been overwritten. If JOBVL='V' or JOBVR='V'\n\
* or both, then B contains the second part of the real Schur\n\
* form of the \"balanced\" versions of the input A and B.\n\
*\n\
* LDB (input) INTEGER\n\
* The leading dimension of B. LDB >= max(1,N).\n\
*\n\
* ALPHAR (output) REAL array, dimension (N)\n\
* ALPHAI (output) REAL array, dimension (N)\n\
* BETA (output) REAL array, dimension (N)\n\
* On exit, (ALPHAR(j) + ALPHAI(j)*i)/BETA(j), j=1,...,N, will\n\
* be the generalized eigenvalues. If ALPHAI(j) is zero, then\n\
* the j-th eigenvalue is real; if positive, then the j-th and\n\
* (j+1)-st eigenvalues are a complex conjugate pair, with\n\
* ALPHAI(j+1) negative.\n\
*\n\
* Note: the quotients ALPHAR(j)/BETA(j) and ALPHAI(j)/BETA(j)\n\
* may easily over- or underflow, and BETA(j) may even be zero.\n\
* Thus, the user should avoid naively computing the ratio\n\
* ALPHA/BETA. However, ALPHAR and ALPHAI will be always less\n\
* than and usually comparable with norm(A) in magnitude, and\n\
* BETA always less than and usually comparable with norm(B).\n\
*\n\
* VL (output) REAL array, dimension (LDVL,N)\n\
* If JOBVL = 'V', the left eigenvectors u(j) are stored one\n\
* after another in the columns of VL, in the same order as\n\
* their eigenvalues. If the j-th eigenvalue is real, then\n\
* u(j) = VL(:,j), the j-th column of VL. If the j-th and\n\
* (j+1)-th eigenvalues form a complex conjugate pair, then\n\
* u(j) = VL(:,j)+i*VL(:,j+1) and u(j+1) = VL(:,j)-i*VL(:,j+1).\n\
* Each eigenvector will be scaled so the largest component have\n\
* abs(real part) + abs(imag. part) = 1.\n\
* Not referenced if JOBVL = 'N'.\n\
*\n\
* LDVL (input) INTEGER\n\
* The leading dimension of the matrix VL. LDVL >= 1, and\n\
* if JOBVL = 'V', LDVL >= N.\n\
*\n\
* VR (output) REAL array, dimension (LDVR,N)\n\
* If JOBVR = 'V', the right eigenvectors v(j) are stored one\n\
* after another in the columns of VR, in the same order as\n\
* their eigenvalues. If the j-th eigenvalue is real, then\n\
* v(j) = VR(:,j), the j-th column of VR. If the j-th and\n\
* (j+1)-th eigenvalues form a complex conjugate pair, then\n\
* v(j) = VR(:,j)+i*VR(:,j+1) and v(j+1) = VR(:,j)-i*VR(:,j+1).\n\
* Each eigenvector will be scaled so the largest component have\n\
* abs(real part) + abs(imag. part) = 1.\n\
* Not referenced if JOBVR = 'N'.\n\
*\n\
* LDVR (input) INTEGER\n\
* The leading dimension of the matrix VR. LDVR >= 1, and\n\
* if JOBVR = 'V', LDVR >= N.\n\
*\n\
* ILO (output) INTEGER\n\
* IHI (output) INTEGER\n\
* ILO and IHI are integer values such that on exit\n\
* A(i,j) = 0 and B(i,j) = 0 if i > j and\n\
* j = 1,...,ILO-1 or i = IHI+1,...,N.\n\
* If BALANC = 'N' or 'S', ILO = 1 and IHI = N.\n\
*\n\
* LSCALE (output) REAL array, dimension (N)\n\
* Details of the permutations and scaling factors applied\n\
* to the left side of A and B. If PL(j) is the index of the\n\
* row interchanged with row j, and DL(j) is the scaling\n\
* factor applied to row j, then\n\
* LSCALE(j) = PL(j) for j = 1,...,ILO-1\n\
* = DL(j) for j = ILO,...,IHI\n\
* = PL(j) for j = IHI+1,...,N.\n\
* The order in which the interchanges are made is N to IHI+1,\n\
* then 1 to ILO-1.\n\
*\n\
* RSCALE (output) REAL array, dimension (N)\n\
* Details of the permutations and scaling factors applied\n\
* to the right side of A and B. If PR(j) is the index of the\n\
* column interchanged with column j, and DR(j) is the scaling\n\
* factor applied to column j, then\n\
* RSCALE(j) = PR(j) for j = 1,...,ILO-1\n\
* = DR(j) for j = ILO,...,IHI\n\
* = PR(j) for j = IHI+1,...,N\n\
* The order in which the interchanges are made is N to IHI+1,\n\
* then 1 to ILO-1.\n\
*\n\
* ABNRM (output) REAL\n\
* The one-norm of the balanced matrix A.\n\
*\n\
* BBNRM (output) REAL\n\
* The one-norm of the balanced matrix B.\n\
*\n\
* RCONDE (output) REAL array, dimension (N)\n\
* If SENSE = 'E' or 'B', the reciprocal condition numbers of\n\
* the eigenvalues, stored in consecutive elements of the array.\n\
* For a complex conjugate pair of eigenvalues two consecutive\n\
* elements of RCONDE are set to the same value. Thus RCONDE(j),\n\
* RCONDV(j), and the j-th columns of VL and VR all correspond\n\
* to the j-th eigenpair.\n\
* If SENSE = 'N' or 'V', RCONDE is not referenced.\n\
*\n\
* RCONDV (output) REAL array, dimension (N)\n\
* If SENSE = 'V' or 'B', the estimated reciprocal condition\n\
* numbers of the eigenvectors, stored in consecutive elements\n\
* of the array. For a complex eigenvector two consecutive\n\
* elements of RCONDV are set to the same value. If the\n\
* eigenvalues cannot be reordered to compute RCONDV(j),\n\
* RCONDV(j) is set to 0; this can only occur when the true\n\
* value would be very small anyway.\n\
* If SENSE = 'N' or 'E', RCONDV is not referenced.\n\
*\n\
* WORK (workspace/output) REAL array, dimension (MAX(1,LWORK))\n\
* On exit, if INFO = 0, WORK(1) returns the optimal LWORK.\n\
*\n\
* LWORK (input) INTEGER\n\
* The dimension of the array WORK. LWORK >= max(1,2*N).\n\
* If BALANC = 'S' or 'B', or JOBVL = 'V', or JOBVR = 'V',\n\
* LWORK >= max(1,6*N).\n\
* If SENSE = 'E', LWORK >= max(1,10*N).\n\
* If SENSE = 'V' or 'B', LWORK >= 2*N*N+8*N+16.\n\
*\n\
* If LWORK = -1, then a workspace query is assumed; the routine\n\
* only calculates the optimal size of the WORK array, returns\n\
* this value as the first entry of the WORK array, and no error\n\
* message related to LWORK is issued by XERBLA.\n\
*\n\
* IWORK (workspace) INTEGER array, dimension (N+6)\n\
* If SENSE = 'E', IWORK is not referenced.\n\
*\n\
* BWORK (workspace) LOGICAL array, dimension (N)\n\
* If SENSE = 'N', BWORK is not referenced.\n\
*\n\
* INFO (output) INTEGER\n\
* = 0: successful exit\n\
* < 0: if INFO = -i, the i-th argument had an illegal value.\n\
* = 1,...,N:\n\
* The QZ iteration failed. No eigenvectors have been\n\
* calculated, but ALPHAR(j), ALPHAI(j), and BETA(j)\n\
* should be correct for j=INFO+1,...,N.\n\
* > N: =N+1: other than QZ iteration failed in SHGEQZ.\n\
* =N+2: error return from STGEVC.\n\
*\n\n\
* Further Details\n\
* ===============\n\
*\n\
* Balancing a matrix pair (A,B) includes, first, permuting rows and\n\
* columns to isolate eigenvalues, second, applying diagonal similarity\n\
* transformation to the rows and columns to make the rows and columns\n\
* as close in norm as possible. The computed reciprocal condition\n\
* numbers correspond to the balanced matrix. Permuting rows and columns\n\
* will not change the condition numbers (in exact arithmetic) but\n\
* diagonal scaling will. For further explanation of balancing, see\n\
* section 4.11.1.2 of LAPACK Users' Guide.\n\
*\n\
* An approximate error bound on the chordal distance between the i-th\n\
* computed generalized eigenvalue w and the corresponding exact\n\
* eigenvalue lambda is\n\
*\n\
* chord(w, lambda) <= EPS * norm(ABNRM, BBNRM) / RCONDE(I)\n\
*\n\
* An approximate error bound for the angle between the i-th computed\n\
* eigenvector VL(i) or VR(i) is given by\n\
*\n\
* EPS * norm(ABNRM, BBNRM) / DIF(i).\n\
*\n\
* For further explanation of the reciprocal condition numbers RCONDE\n\
* and RCONDV, see section 4.11 of LAPACK User's Guide.\n\
*\n\
* =====================================================================\n\
*\n"
|
