You can not select more than 25 topics Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.

sgglse.c 20 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658
  1. #include <math.h>
  2. #include <stdlib.h>
  3. #include <string.h>
  4. #include <stdio.h>
  5. #include <complex.h>
  6. #ifdef complex
  7. #undef complex
  8. #endif
  9. #ifdef I
  10. #undef I
  11. #endif
  12. #if defined(_WIN64)
  13. typedef long long BLASLONG;
  14. typedef unsigned long long BLASULONG;
  15. #else
  16. typedef long BLASLONG;
  17. typedef unsigned long BLASULONG;
  18. #endif
  19. #ifdef LAPACK_ILP64
  20. typedef BLASLONG blasint;
  21. #if defined(_WIN64)
  22. #define blasabs(x) llabs(x)
  23. #else
  24. #define blasabs(x) labs(x)
  25. #endif
  26. #else
  27. typedef int blasint;
  28. #define blasabs(x) abs(x)
  29. #endif
  30. typedef blasint integer;
  31. typedef unsigned int uinteger;
  32. typedef char *address;
  33. typedef short int shortint;
  34. typedef float real;
  35. typedef double doublereal;
  36. typedef struct { real r, i; } complex;
  37. typedef struct { doublereal r, i; } doublecomplex;
  38. #ifdef _MSC_VER
  39. static inline _Fcomplex Cf(complex *z) {_Fcomplex zz={z->r , z->i}; return zz;}
  40. static inline _Dcomplex Cd(doublecomplex *z) {_Dcomplex zz={z->r , z->i};return zz;}
  41. static inline _Fcomplex * _pCf(complex *z) {return (_Fcomplex*)z;}
  42. static inline _Dcomplex * _pCd(doublecomplex *z) {return (_Dcomplex*)z;}
  43. #else
  44. static inline _Complex float Cf(complex *z) {return z->r + z->i*_Complex_I;}
  45. static inline _Complex double Cd(doublecomplex *z) {return z->r + z->i*_Complex_I;}
  46. static inline _Complex float * _pCf(complex *z) {return (_Complex float*)z;}
  47. static inline _Complex double * _pCd(doublecomplex *z) {return (_Complex double*)z;}
  48. #endif
  49. #define pCf(z) (*_pCf(z))
  50. #define pCd(z) (*_pCd(z))
  51. typedef blasint logical;
  52. typedef char logical1;
  53. typedef char integer1;
  54. #define TRUE_ (1)
  55. #define FALSE_ (0)
  56. /* Extern is for use with -E */
  57. #ifndef Extern
  58. #define Extern extern
  59. #endif
  60. /* I/O stuff */
  61. typedef int flag;
  62. typedef int ftnlen;
  63. typedef int ftnint;
  64. /*external read, write*/
  65. typedef struct
  66. { flag cierr;
  67. ftnint ciunit;
  68. flag ciend;
  69. char *cifmt;
  70. ftnint cirec;
  71. } cilist;
  72. /*internal read, write*/
  73. typedef struct
  74. { flag icierr;
  75. char *iciunit;
  76. flag iciend;
  77. char *icifmt;
  78. ftnint icirlen;
  79. ftnint icirnum;
  80. } icilist;
  81. /*open*/
  82. typedef struct
  83. { flag oerr;
  84. ftnint ounit;
  85. char *ofnm;
  86. ftnlen ofnmlen;
  87. char *osta;
  88. char *oacc;
  89. char *ofm;
  90. ftnint orl;
  91. char *oblnk;
  92. } olist;
  93. /*close*/
  94. typedef struct
  95. { flag cerr;
  96. ftnint cunit;
  97. char *csta;
  98. } cllist;
  99. /*rewind, backspace, endfile*/
  100. typedef struct
  101. { flag aerr;
  102. ftnint aunit;
  103. } alist;
  104. /* inquire */
  105. typedef struct
  106. { flag inerr;
  107. ftnint inunit;
  108. char *infile;
  109. ftnlen infilen;
  110. ftnint *inex; /*parameters in standard's order*/
  111. ftnint *inopen;
  112. ftnint *innum;
  113. ftnint *innamed;
  114. char *inname;
  115. ftnlen innamlen;
  116. char *inacc;
  117. ftnlen inacclen;
  118. char *inseq;
  119. ftnlen inseqlen;
  120. char *indir;
  121. ftnlen indirlen;
  122. char *infmt;
  123. ftnlen infmtlen;
  124. char *inform;
  125. ftnint informlen;
  126. char *inunf;
  127. ftnlen inunflen;
  128. ftnint *inrecl;
  129. ftnint *innrec;
  130. char *inblank;
  131. ftnlen inblanklen;
  132. } inlist;
  133. #define VOID void
  134. union Multitype { /* for multiple entry points */
  135. integer1 g;
  136. shortint h;
  137. integer i;
  138. /* longint j; */
  139. real r;
  140. doublereal d;
  141. complex c;
  142. doublecomplex z;
  143. };
  144. typedef union Multitype Multitype;
  145. struct Vardesc { /* for Namelist */
  146. char *name;
  147. char *addr;
  148. ftnlen *dims;
  149. int type;
  150. };
  151. typedef struct Vardesc Vardesc;
  152. struct Namelist {
  153. char *name;
  154. Vardesc **vars;
  155. int nvars;
  156. };
  157. typedef struct Namelist Namelist;
  158. #define abs(x) ((x) >= 0 ? (x) : -(x))
  159. #define dabs(x) (fabs(x))
  160. #define f2cmin(a,b) ((a) <= (b) ? (a) : (b))
  161. #define f2cmax(a,b) ((a) >= (b) ? (a) : (b))
  162. #define dmin(a,b) (f2cmin(a,b))
  163. #define dmax(a,b) (f2cmax(a,b))
  164. #define bit_test(a,b) ((a) >> (b) & 1)
  165. #define bit_clear(a,b) ((a) & ~((uinteger)1 << (b)))
  166. #define bit_set(a,b) ((a) | ((uinteger)1 << (b)))
  167. #define abort_() { sig_die("Fortran abort routine called", 1); }
  168. #define c_abs(z) (cabsf(Cf(z)))
  169. #define c_cos(R,Z) { pCf(R)=ccos(Cf(Z)); }
  170. #ifdef _MSC_VER
  171. #define c_div(c, a, b) {Cf(c)._Val[0] = (Cf(a)._Val[0]/Cf(b)._Val[0]); Cf(c)._Val[1]=(Cf(a)._Val[1]/Cf(b)._Val[1]);}
  172. #define z_div(c, a, b) {Cd(c)._Val[0] = (Cd(a)._Val[0]/Cd(b)._Val[0]); Cd(c)._Val[1]=(Cd(a)._Val[1]/df(b)._Val[1]);}
  173. #else
  174. #define c_div(c, a, b) {pCf(c) = Cf(a)/Cf(b);}
  175. #define z_div(c, a, b) {pCd(c) = Cd(a)/Cd(b);}
  176. #endif
  177. #define c_exp(R, Z) {pCf(R) = cexpf(Cf(Z));}
  178. #define c_log(R, Z) {pCf(R) = clogf(Cf(Z));}
  179. #define c_sin(R, Z) {pCf(R) = csinf(Cf(Z));}
  180. //#define c_sqrt(R, Z) {*(R) = csqrtf(Cf(Z));}
  181. #define c_sqrt(R, Z) {pCf(R) = csqrtf(Cf(Z));}
  182. #define d_abs(x) (fabs(*(x)))
  183. #define d_acos(x) (acos(*(x)))
  184. #define d_asin(x) (asin(*(x)))
  185. #define d_atan(x) (atan(*(x)))
  186. #define d_atn2(x, y) (atan2(*(x),*(y)))
  187. #define d_cnjg(R, Z) { pCd(R) = conj(Cd(Z)); }
  188. #define r_cnjg(R, Z) { pCf(R) = conjf(Cf(Z)); }
  189. #define d_cos(x) (cos(*(x)))
  190. #define d_cosh(x) (cosh(*(x)))
  191. #define d_dim(__a, __b) ( *(__a) > *(__b) ? *(__a) - *(__b) : 0.0 )
  192. #define d_exp(x) (exp(*(x)))
  193. #define d_imag(z) (cimag(Cd(z)))
  194. #define r_imag(z) (cimagf(Cf(z)))
  195. #define d_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
  196. #define r_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
  197. #define d_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
  198. #define r_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
  199. #define d_log(x) (log(*(x)))
  200. #define d_mod(x, y) (fmod(*(x), *(y)))
  201. #define u_nint(__x) ((__x)>=0 ? floor((__x) + .5) : -floor(.5 - (__x)))
  202. #define d_nint(x) u_nint(*(x))
  203. #define u_sign(__a,__b) ((__b) >= 0 ? ((__a) >= 0 ? (__a) : -(__a)) : -((__a) >= 0 ? (__a) : -(__a)))
  204. #define d_sign(a,b) u_sign(*(a),*(b))
  205. #define r_sign(a,b) u_sign(*(a),*(b))
  206. #define d_sin(x) (sin(*(x)))
  207. #define d_sinh(x) (sinh(*(x)))
  208. #define d_sqrt(x) (sqrt(*(x)))
  209. #define d_tan(x) (tan(*(x)))
  210. #define d_tanh(x) (tanh(*(x)))
  211. #define i_abs(x) abs(*(x))
  212. #define i_dnnt(x) ((integer)u_nint(*(x)))
  213. #define i_len(s, n) (n)
  214. #define i_nint(x) ((integer)u_nint(*(x)))
  215. #define i_sign(a,b) ((integer)u_sign((integer)*(a),(integer)*(b)))
  216. #define s_cat(lpp, rpp, rnp, np, llp) { ftnlen i, nc, ll; char *f__rp, *lp; ll = (llp); lp = (lpp); for(i=0; i < (int)*(np); ++i) { nc = ll; if((rnp)[i] < nc) nc = (rnp)[i]; ll -= nc; f__rp = (rpp)[i]; while(--nc >= 0) *lp++ = *(f__rp)++; } while(--ll >= 0) *lp++ = ' '; }
  217. #define s_cmp(a,b,c,d) ((integer)strncmp((a),(b),f2cmin((c),(d))))
  218. #define s_copy(A,B,C,D) { int __i,__m; for (__i=0, __m=f2cmin((C),(D)); __i<__m && (B)[__i] != 0; ++__i) (A)[__i] = (B)[__i]; }
  219. #define sig_die(s, kill) { exit(1); }
  220. #define s_stop(s, n) {exit(0);}
  221. #define z_abs(z) (cabs(Cd(z)))
  222. #define z_exp(R, Z) {pCd(R) = cexp(Cd(Z));}
  223. #define z_sqrt(R, Z) {pCd(R) = csqrt(Cd(Z));}
  224. #define myexit_() break;
  225. #define mycycle() continue;
  226. #define myceiling(w) {ceil(w)}
  227. #define myhuge(w) {HUGE_VAL}
  228. //#define mymaxloc_(w,s,e,n) {if (sizeof(*(w)) == sizeof(double)) dmaxloc_((w),*(s),*(e),n); else dmaxloc_((w),*(s),*(e),n);}
  229. #define mymaxloc(w,s,e,n) {dmaxloc_(w,*(s),*(e),n)}
  230. /* -- translated by f2c (version 20000121).
  231. You must link the resulting object file with the libraries:
  232. -lf2c -lm (in that order)
  233. */
  234. /* Table of constant values */
  235. static integer c__1 = 1;
  236. static integer c_n1 = -1;
  237. static real c_b31 = -1.f;
  238. static real c_b33 = 1.f;
  239. /* > \brief <b> SGGLSE solves overdetermined or underdetermined systems for OTHER matrices</b> */
  240. /* =========== DOCUMENTATION =========== */
  241. /* Online html documentation available at */
  242. /* http://www.netlib.org/lapack/explore-html/ */
  243. /* > \htmlonly */
  244. /* > Download SGGLSE + dependencies */
  245. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/sgglse.
  246. f"> */
  247. /* > [TGZ]</a> */
  248. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/sgglse.
  249. f"> */
  250. /* > [ZIP]</a> */
  251. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/sgglse.
  252. f"> */
  253. /* > [TXT]</a> */
  254. /* > \endhtmlonly */
  255. /* Definition: */
  256. /* =========== */
  257. /* SUBROUTINE SGGLSE( M, N, P, A, LDA, B, LDB, C, D, X, WORK, LWORK, */
  258. /* INFO ) */
  259. /* INTEGER INFO, LDA, LDB, LWORK, M, N, P */
  260. /* REAL A( LDA, * ), B( LDB, * ), C( * ), D( * ), */
  261. /* $ WORK( * ), X( * ) */
  262. /* > \par Purpose: */
  263. /* ============= */
  264. /* > */
  265. /* > \verbatim */
  266. /* > */
  267. /* > SGGLSE solves the linear equality-constrained least squares (LSE) */
  268. /* > problem: */
  269. /* > */
  270. /* > minimize || c - A*x ||_2 subject to B*x = d */
  271. /* > */
  272. /* > where A is an M-by-N matrix, B is a P-by-N matrix, c is a given */
  273. /* > M-vector, and d is a given P-vector. It is assumed that */
  274. /* > P <= N <= M+P, and */
  275. /* > */
  276. /* > rank(B) = P and rank( (A) ) = N. */
  277. /* > ( (B) ) */
  278. /* > */
  279. /* > These conditions ensure that the LSE problem has a unique solution, */
  280. /* > which is obtained using a generalized RQ factorization of the */
  281. /* > matrices (B, A) given by */
  282. /* > */
  283. /* > B = (0 R)*Q, A = Z*T*Q. */
  284. /* > \endverbatim */
  285. /* Arguments: */
  286. /* ========== */
  287. /* > \param[in] M */
  288. /* > \verbatim */
  289. /* > M is INTEGER */
  290. /* > The number of rows of the matrix A. M >= 0. */
  291. /* > \endverbatim */
  292. /* > */
  293. /* > \param[in] N */
  294. /* > \verbatim */
  295. /* > N is INTEGER */
  296. /* > The number of columns of the matrices A and B. N >= 0. */
  297. /* > \endverbatim */
  298. /* > */
  299. /* > \param[in] P */
  300. /* > \verbatim */
  301. /* > P is INTEGER */
  302. /* > The number of rows of the matrix B. 0 <= P <= N <= M+P. */
  303. /* > \endverbatim */
  304. /* > */
  305. /* > \param[in,out] A */
  306. /* > \verbatim */
  307. /* > A is REAL array, dimension (LDA,N) */
  308. /* > On entry, the M-by-N matrix A. */
  309. /* > On exit, the elements on and above the diagonal of the array */
  310. /* > contain the f2cmin(M,N)-by-N upper trapezoidal matrix T. */
  311. /* > \endverbatim */
  312. /* > */
  313. /* > \param[in] LDA */
  314. /* > \verbatim */
  315. /* > LDA is INTEGER */
  316. /* > The leading dimension of the array A. LDA >= f2cmax(1,M). */
  317. /* > \endverbatim */
  318. /* > */
  319. /* > \param[in,out] B */
  320. /* > \verbatim */
  321. /* > B is REAL array, dimension (LDB,N) */
  322. /* > On entry, the P-by-N matrix B. */
  323. /* > On exit, the upper triangle of the subarray B(1:P,N-P+1:N) */
  324. /* > contains the P-by-P upper triangular matrix R. */
  325. /* > \endverbatim */
  326. /* > */
  327. /* > \param[in] LDB */
  328. /* > \verbatim */
  329. /* > LDB is INTEGER */
  330. /* > The leading dimension of the array B. LDB >= f2cmax(1,P). */
  331. /* > \endverbatim */
  332. /* > */
  333. /* > \param[in,out] C */
  334. /* > \verbatim */
  335. /* > C is REAL array, dimension (M) */
  336. /* > On entry, C contains the right hand side vector for the */
  337. /* > least squares part of the LSE problem. */
  338. /* > On exit, the residual sum of squares for the solution */
  339. /* > is given by the sum of squares of elements N-P+1 to M of */
  340. /* > vector C. */
  341. /* > \endverbatim */
  342. /* > */
  343. /* > \param[in,out] D */
  344. /* > \verbatim */
  345. /* > D is REAL array, dimension (P) */
  346. /* > On entry, D contains the right hand side vector for the */
  347. /* > constrained equation. */
  348. /* > On exit, D is destroyed. */
  349. /* > \endverbatim */
  350. /* > */
  351. /* > \param[out] X */
  352. /* > \verbatim */
  353. /* > X is REAL array, dimension (N) */
  354. /* > On exit, X is the solution of the LSE problem. */
  355. /* > \endverbatim */
  356. /* > */
  357. /* > \param[out] WORK */
  358. /* > \verbatim */
  359. /* > WORK is REAL array, dimension (MAX(1,LWORK)) */
  360. /* > On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
  361. /* > \endverbatim */
  362. /* > */
  363. /* > \param[in] LWORK */
  364. /* > \verbatim */
  365. /* > LWORK is INTEGER */
  366. /* > The dimension of the array WORK. LWORK >= f2cmax(1,M+N+P). */
  367. /* > For optimum performance LWORK >= P+f2cmin(M,N)+f2cmax(M,N)*NB, */
  368. /* > where NB is an upper bound for the optimal blocksizes for */
  369. /* > SGEQRF, SGERQF, SORMQR and SORMRQ. */
  370. /* > */
  371. /* > If LWORK = -1, then a workspace query is assumed; the routine */
  372. /* > only calculates the optimal size of the WORK array, returns */
  373. /* > this value as the first entry of the WORK array, and no error */
  374. /* > message related to LWORK is issued by XERBLA. */
  375. /* > \endverbatim */
  376. /* > */
  377. /* > \param[out] INFO */
  378. /* > \verbatim */
  379. /* > INFO is INTEGER */
  380. /* > = 0: successful exit. */
  381. /* > < 0: if INFO = -i, the i-th argument had an illegal value. */
  382. /* > = 1: the upper triangular factor R associated with B in the */
  383. /* > generalized RQ factorization of the pair (B, A) is */
  384. /* > singular, so that rank(B) < P; the least squares */
  385. /* > solution could not be computed. */
  386. /* > = 2: the (N-P) by (N-P) part of the upper trapezoidal factor */
  387. /* > T associated with A in the generalized RQ factorization */
  388. /* > of the pair (B, A) is singular, so that */
  389. /* > rank( (A) ) < N; the least squares solution could not */
  390. /* > ( (B) ) */
  391. /* > be computed. */
  392. /* > \endverbatim */
  393. /* Authors: */
  394. /* ======== */
  395. /* > \author Univ. of Tennessee */
  396. /* > \author Univ. of California Berkeley */
  397. /* > \author Univ. of Colorado Denver */
  398. /* > \author NAG Ltd. */
  399. /* > \date December 2016 */
  400. /* > \ingroup realOTHERsolve */
  401. /* ===================================================================== */
  402. /* Subroutine */ void sgglse_(integer *m, integer *n, integer *p, real *a,
  403. integer *lda, real *b, integer *ldb, real *c__, real *d__, real *x,
  404. real *work, integer *lwork, integer *info)
  405. {
  406. /* System generated locals */
  407. integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2;
  408. /* Local variables */
  409. integer lopt;
  410. extern /* Subroutine */ void sgemv_(char *, integer *, integer *, real *,
  411. real *, integer *, real *, integer *, real *, real *, integer *), scopy_(integer *, real *, integer *, real *, integer *),
  412. saxpy_(integer *, real *, real *, integer *, real *, integer *),
  413. strmv_(char *, char *, char *, integer *, real *, integer *, real
  414. *, integer *);
  415. integer nb, mn, nr;
  416. extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
  417. extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
  418. integer *, integer *, ftnlen, ftnlen);
  419. extern /* Subroutine */ void sggrqf_(integer *, integer *, integer *, real
  420. *, integer *, real *, real *, integer *, real *, real *, integer *
  421. , integer *);
  422. integer lwkmin, nb1, nb2, nb3, nb4, lwkopt;
  423. logical lquery;
  424. extern /* Subroutine */ void sormqr_(char *, char *, integer *, integer *,
  425. integer *, real *, integer *, real *, real *, integer *, real *,
  426. integer *, integer *), sormrq_(char *, char *,
  427. integer *, integer *, integer *, real *, integer *, real *, real *
  428. , integer *, real *, integer *, integer *);
  429. extern void strtrs_(char *, char *, char *, integer *, integer *, real *,
  430. integer *, real *, integer *, integer *);
  431. /* -- LAPACK driver routine (version 3.7.0) -- */
  432. /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
  433. /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
  434. /* December 2016 */
  435. /* ===================================================================== */
  436. /* Test the input parameters */
  437. /* Parameter adjustments */
  438. a_dim1 = *lda;
  439. a_offset = 1 + a_dim1 * 1;
  440. a -= a_offset;
  441. b_dim1 = *ldb;
  442. b_offset = 1 + b_dim1 * 1;
  443. b -= b_offset;
  444. --c__;
  445. --d__;
  446. --x;
  447. --work;
  448. /* Function Body */
  449. *info = 0;
  450. mn = f2cmin(*m,*n);
  451. lquery = *lwork == -1;
  452. if (*m < 0) {
  453. *info = -1;
  454. } else if (*n < 0) {
  455. *info = -2;
  456. } else if (*p < 0 || *p > *n || *p < *n - *m) {
  457. *info = -3;
  458. } else if (*lda < f2cmax(1,*m)) {
  459. *info = -5;
  460. } else if (*ldb < f2cmax(1,*p)) {
  461. *info = -7;
  462. }
  463. /* Calculate workspace */
  464. if (*info == 0) {
  465. if (*n == 0) {
  466. lwkmin = 1;
  467. lwkopt = 1;
  468. } else {
  469. nb1 = ilaenv_(&c__1, "SGEQRF", " ", m, n, &c_n1, &c_n1, (ftnlen)6,
  470. (ftnlen)1);
  471. nb2 = ilaenv_(&c__1, "SGERQF", " ", m, n, &c_n1, &c_n1, (ftnlen)6,
  472. (ftnlen)1);
  473. nb3 = ilaenv_(&c__1, "SORMQR", " ", m, n, p, &c_n1, (ftnlen)6, (
  474. ftnlen)1);
  475. nb4 = ilaenv_(&c__1, "SORMRQ", " ", m, n, p, &c_n1, (ftnlen)6, (
  476. ftnlen)1);
  477. /* Computing MAX */
  478. i__1 = f2cmax(nb1,nb2), i__1 = f2cmax(i__1,nb3);
  479. nb = f2cmax(i__1,nb4);
  480. lwkmin = *m + *n + *p;
  481. lwkopt = *p + mn + f2cmax(*m,*n) * nb;
  482. }
  483. work[1] = (real) lwkopt;
  484. if (*lwork < lwkmin && ! lquery) {
  485. *info = -12;
  486. }
  487. }
  488. if (*info != 0) {
  489. i__1 = -(*info);
  490. xerbla_("SGGLSE", &i__1, (ftnlen)6);
  491. return;
  492. } else if (lquery) {
  493. return;
  494. }
  495. /* Quick return if possible */
  496. if (*n == 0) {
  497. return;
  498. }
  499. /* Compute the GRQ factorization of matrices B and A: */
  500. /* B*Q**T = ( 0 T12 ) P Z**T*A*Q**T = ( R11 R12 ) N-P */
  501. /* N-P P ( 0 R22 ) M+P-N */
  502. /* N-P P */
  503. /* where T12 and R11 are upper triangular, and Q and Z are */
  504. /* orthogonal. */
  505. i__1 = *lwork - *p - mn;
  506. sggrqf_(p, m, n, &b[b_offset], ldb, &work[1], &a[a_offset], lda, &work[*p
  507. + 1], &work[*p + mn + 1], &i__1, info);
  508. lopt = work[*p + mn + 1];
  509. /* Update c = Z**T *c = ( c1 ) N-P */
  510. /* ( c2 ) M+P-N */
  511. i__1 = f2cmax(1,*m);
  512. i__2 = *lwork - *p - mn;
  513. sormqr_("Left", "Transpose", m, &c__1, &mn, &a[a_offset], lda, &work[*p +
  514. 1], &c__[1], &i__1, &work[*p + mn + 1], &i__2, info);
  515. /* Computing MAX */
  516. i__1 = lopt, i__2 = (integer) work[*p + mn + 1];
  517. lopt = f2cmax(i__1,i__2);
  518. /* Solve T12*x2 = d for x2 */
  519. if (*p > 0) {
  520. strtrs_("Upper", "No transpose", "Non-unit", p, &c__1, &b[(*n - *p +
  521. 1) * b_dim1 + 1], ldb, &d__[1], p, info);
  522. if (*info > 0) {
  523. *info = 1;
  524. return;
  525. }
  526. /* Put the solution in X */
  527. scopy_(p, &d__[1], &c__1, &x[*n - *p + 1], &c__1);
  528. /* Update c1 */
  529. i__1 = *n - *p;
  530. sgemv_("No transpose", &i__1, p, &c_b31, &a[(*n - *p + 1) * a_dim1 +
  531. 1], lda, &d__[1], &c__1, &c_b33, &c__[1], &c__1);
  532. }
  533. /* Solve R11*x1 = c1 for x1 */
  534. if (*n > *p) {
  535. i__1 = *n - *p;
  536. i__2 = *n - *p;
  537. strtrs_("Upper", "No transpose", "Non-unit", &i__1, &c__1, &a[
  538. a_offset], lda, &c__[1], &i__2, info);
  539. if (*info > 0) {
  540. *info = 2;
  541. return;
  542. }
  543. /* Put the solutions in X */
  544. i__1 = *n - *p;
  545. scopy_(&i__1, &c__[1], &c__1, &x[1], &c__1);
  546. }
  547. /* Compute the residual vector: */
  548. if (*m < *n) {
  549. nr = *m + *p - *n;
  550. if (nr > 0) {
  551. i__1 = *n - *m;
  552. sgemv_("No transpose", &nr, &i__1, &c_b31, &a[*n - *p + 1 + (*m +
  553. 1) * a_dim1], lda, &d__[nr + 1], &c__1, &c_b33, &c__[*n -
  554. *p + 1], &c__1);
  555. }
  556. } else {
  557. nr = *p;
  558. }
  559. if (nr > 0) {
  560. strmv_("Upper", "No transpose", "Non unit", &nr, &a[*n - *p + 1 + (*n
  561. - *p + 1) * a_dim1], lda, &d__[1], &c__1);
  562. saxpy_(&nr, &c_b31, &d__[1], &c__1, &c__[*n - *p + 1], &c__1);
  563. }
  564. /* Backward transformation x = Q**T*x */
  565. i__1 = *lwork - *p - mn;
  566. sormrq_("Left", "Transpose", n, &c__1, p, &b[b_offset], ldb, &work[1], &x[
  567. 1], n, &work[*p + mn + 1], &i__1, info);
  568. /* Computing MAX */
  569. i__1 = lopt, i__2 = (integer) work[*p + mn + 1];
  570. work[1] = (real) (*p + mn + f2cmax(i__1,i__2));
  571. return;
  572. /* End of SGGLSE */
  573. } /* sgglse_ */