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.

sggglm.c 19 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659
  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_b32 = -1.f;
  238. static real c_b34 = 1.f;
  239. /* > \brief \b SGGGLM */
  240. /* =========== DOCUMENTATION =========== */
  241. /* Online html documentation available at */
  242. /* http://www.netlib.org/lapack/explore-html/ */
  243. /* > \htmlonly */
  244. /* > Download SGGGLM + dependencies */
  245. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/sggglm.
  246. f"> */
  247. /* > [TGZ]</a> */
  248. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/sggglm.
  249. f"> */
  250. /* > [ZIP]</a> */
  251. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/sggglm.
  252. f"> */
  253. /* > [TXT]</a> */
  254. /* > \endhtmlonly */
  255. /* Definition: */
  256. /* =========== */
  257. /* SUBROUTINE SGGGLM( N, M, P, A, LDA, B, LDB, D, X, Y, WORK, LWORK, */
  258. /* INFO ) */
  259. /* INTEGER INFO, LDA, LDB, LWORK, M, N, P */
  260. /* REAL A( LDA, * ), B( LDB, * ), D( * ), WORK( * ), */
  261. /* $ X( * ), Y( * ) */
  262. /* > \par Purpose: */
  263. /* ============= */
  264. /* > */
  265. /* > \verbatim */
  266. /* > */
  267. /* > SGGGLM solves a general Gauss-Markov linear model (GLM) problem: */
  268. /* > */
  269. /* > minimize || y ||_2 subject to d = A*x + B*y */
  270. /* > x */
  271. /* > */
  272. /* > where A is an N-by-M matrix, B is an N-by-P matrix, and d is a */
  273. /* > given N-vector. It is assumed that M <= N <= M+P, and */
  274. /* > */
  275. /* > rank(A) = M and rank( A B ) = N. */
  276. /* > */
  277. /* > Under these assumptions, the constrained equation is always */
  278. /* > consistent, and there is a unique solution x and a minimal 2-norm */
  279. /* > solution y, which is obtained using a generalized QR factorization */
  280. /* > of the matrices (A, B) given by */
  281. /* > */
  282. /* > A = Q*(R), B = Q*T*Z. */
  283. /* > (0) */
  284. /* > */
  285. /* > In particular, if matrix B is square nonsingular, then the problem */
  286. /* > GLM is equivalent to the following weighted linear least squares */
  287. /* > problem */
  288. /* > */
  289. /* > minimize || inv(B)*(d-A*x) ||_2 */
  290. /* > x */
  291. /* > */
  292. /* > where inv(B) denotes the inverse of B. */
  293. /* > \endverbatim */
  294. /* Arguments: */
  295. /* ========== */
  296. /* > \param[in] N */
  297. /* > \verbatim */
  298. /* > N is INTEGER */
  299. /* > The number of rows of the matrices A and B. N >= 0. */
  300. /* > \endverbatim */
  301. /* > */
  302. /* > \param[in] M */
  303. /* > \verbatim */
  304. /* > M is INTEGER */
  305. /* > The number of columns of the matrix A. 0 <= M <= N. */
  306. /* > \endverbatim */
  307. /* > */
  308. /* > \param[in] P */
  309. /* > \verbatim */
  310. /* > P is INTEGER */
  311. /* > The number of columns of the matrix B. P >= N-M. */
  312. /* > \endverbatim */
  313. /* > */
  314. /* > \param[in,out] A */
  315. /* > \verbatim */
  316. /* > A is REAL array, dimension (LDA,M) */
  317. /* > On entry, the N-by-M matrix A. */
  318. /* > On exit, the upper triangular part of the array A contains */
  319. /* > the M-by-M upper triangular matrix R. */
  320. /* > \endverbatim */
  321. /* > */
  322. /* > \param[in] LDA */
  323. /* > \verbatim */
  324. /* > LDA is INTEGER */
  325. /* > The leading dimension of the array A. LDA >= f2cmax(1,N). */
  326. /* > \endverbatim */
  327. /* > */
  328. /* > \param[in,out] B */
  329. /* > \verbatim */
  330. /* > B is REAL array, dimension (LDB,P) */
  331. /* > On entry, the N-by-P matrix B. */
  332. /* > On exit, if N <= P, the upper triangle of the subarray */
  333. /* > B(1:N,P-N+1:P) contains the N-by-N upper triangular matrix T; */
  334. /* > if N > P, the elements on and above the (N-P)th subdiagonal */
  335. /* > contain the N-by-P upper trapezoidal matrix T. */
  336. /* > \endverbatim */
  337. /* > */
  338. /* > \param[in] LDB */
  339. /* > \verbatim */
  340. /* > LDB is INTEGER */
  341. /* > The leading dimension of the array B. LDB >= f2cmax(1,N). */
  342. /* > \endverbatim */
  343. /* > */
  344. /* > \param[in,out] D */
  345. /* > \verbatim */
  346. /* > D is REAL array, dimension (N) */
  347. /* > On entry, D is the left hand side of the GLM equation. */
  348. /* > On exit, D is destroyed. */
  349. /* > \endverbatim */
  350. /* > */
  351. /* > \param[out] X */
  352. /* > \verbatim */
  353. /* > X is REAL array, dimension (M) */
  354. /* > \endverbatim */
  355. /* > */
  356. /* > \param[out] Y */
  357. /* > \verbatim */
  358. /* > Y is REAL array, dimension (P) */
  359. /* > */
  360. /* > On exit, X and Y are the solutions of the GLM problem. */
  361. /* > \endverbatim */
  362. /* > */
  363. /* > \param[out] WORK */
  364. /* > \verbatim */
  365. /* > WORK is REAL array, dimension (MAX(1,LWORK)) */
  366. /* > On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
  367. /* > \endverbatim */
  368. /* > */
  369. /* > \param[in] LWORK */
  370. /* > \verbatim */
  371. /* > LWORK is INTEGER */
  372. /* > The dimension of the array WORK. LWORK >= f2cmax(1,N+M+P). */
  373. /* > For optimum performance, LWORK >= M+f2cmin(N,P)+f2cmax(N,P)*NB, */
  374. /* > where NB is an upper bound for the optimal blocksizes for */
  375. /* > SGEQRF, SGERQF, SORMQR and SORMRQ. */
  376. /* > */
  377. /* > If LWORK = -1, then a workspace query is assumed; the routine */
  378. /* > only calculates the optimal size of the WORK array, returns */
  379. /* > this value as the first entry of the WORK array, and no error */
  380. /* > message related to LWORK is issued by XERBLA. */
  381. /* > \endverbatim */
  382. /* > */
  383. /* > \param[out] INFO */
  384. /* > \verbatim */
  385. /* > INFO is INTEGER */
  386. /* > = 0: successful exit. */
  387. /* > < 0: if INFO = -i, the i-th argument had an illegal value. */
  388. /* > = 1: the upper triangular factor R associated with A in the */
  389. /* > generalized QR factorization of the pair (A, B) is */
  390. /* > singular, so that rank(A) < M; the least squares */
  391. /* > solution could not be computed. */
  392. /* > = 2: the bottom (N-M) by (N-M) part of the upper trapezoidal */
  393. /* > factor T associated with B in the generalized QR */
  394. /* > factorization of the pair (A, B) is singular, so that */
  395. /* > rank( A B ) < N; the least squares solution could not */
  396. /* > be computed. */
  397. /* > \endverbatim */
  398. /* Authors: */
  399. /* ======== */
  400. /* > \author Univ. of Tennessee */
  401. /* > \author Univ. of California Berkeley */
  402. /* > \author Univ. of Colorado Denver */
  403. /* > \author NAG Ltd. */
  404. /* > \date December 2016 */
  405. /* > \ingroup realOTHEReigen */
  406. /* ===================================================================== */
  407. /* Subroutine */ void sggglm_(integer *n, integer *m, integer *p, real *a,
  408. integer *lda, real *b, integer *ldb, real *d__, real *x, real *y,
  409. real *work, integer *lwork, integer *info)
  410. {
  411. /* System generated locals */
  412. integer a_dim1, a_offset, b_dim1, b_offset, i__1, i__2, i__3, i__4;
  413. /* Local variables */
  414. integer lopt, i__;
  415. extern /* Subroutine */ void sgemv_(char *, integer *, integer *, real *,
  416. real *, integer *, real *, integer *, real *, real *, integer *), scopy_(integer *, real *, integer *, real *, integer *);
  417. integer nb, np;
  418. extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
  419. extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
  420. integer *, integer *, ftnlen, ftnlen);
  421. extern /* Subroutine */ void sggqrf_(integer *, integer *, integer *, real
  422. *, integer *, real *, real *, integer *, real *, real *, integer *
  423. , integer *);
  424. integer lwkmin, nb1, nb2, nb3, nb4, lwkopt;
  425. logical lquery;
  426. extern /* Subroutine */ void sormqr_(char *, char *, integer *, integer *,
  427. integer *, real *, integer *, real *, real *, integer *, real *,
  428. integer *, integer *), sormrq_(char *, char *,
  429. integer *, integer *, integer *, real *, integer *, real *, real *
  430. , integer *, real *, integer *, integer *);
  431. extern void strtrs_(char *, char *, char *, integer *, integer *, real *,
  432. integer *, real *, integer *, integer *);
  433. /* -- LAPACK driver routine (version 3.7.0) -- */
  434. /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
  435. /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
  436. /* December 2016 */
  437. /* =================================================================== */
  438. /* Test the input parameters */
  439. /* Parameter adjustments */
  440. a_dim1 = *lda;
  441. a_offset = 1 + a_dim1 * 1;
  442. a -= a_offset;
  443. b_dim1 = *ldb;
  444. b_offset = 1 + b_dim1 * 1;
  445. b -= b_offset;
  446. --d__;
  447. --x;
  448. --y;
  449. --work;
  450. /* Function Body */
  451. *info = 0;
  452. np = f2cmin(*n,*p);
  453. lquery = *lwork == -1;
  454. if (*n < 0) {
  455. *info = -1;
  456. } else if (*m < 0 || *m > *n) {
  457. *info = -2;
  458. } else if (*p < 0 || *p < *n - *m) {
  459. *info = -3;
  460. } else if (*lda < f2cmax(1,*n)) {
  461. *info = -5;
  462. } else if (*ldb < f2cmax(1,*n)) {
  463. *info = -7;
  464. }
  465. /* Calculate workspace */
  466. if (*info == 0) {
  467. if (*n == 0) {
  468. lwkmin = 1;
  469. lwkopt = 1;
  470. } else {
  471. nb1 = ilaenv_(&c__1, "SGEQRF", " ", n, m, &c_n1, &c_n1, (ftnlen)6,
  472. (ftnlen)1);
  473. nb2 = ilaenv_(&c__1, "SGERQF", " ", n, m, &c_n1, &c_n1, (ftnlen)6,
  474. (ftnlen)1);
  475. nb3 = ilaenv_(&c__1, "SORMQR", " ", n, m, p, &c_n1, (ftnlen)6, (
  476. ftnlen)1);
  477. nb4 = ilaenv_(&c__1, "SORMRQ", " ", n, m, p, &c_n1, (ftnlen)6, (
  478. ftnlen)1);
  479. /* Computing MAX */
  480. i__1 = f2cmax(nb1,nb2), i__1 = f2cmax(i__1,nb3);
  481. nb = f2cmax(i__1,nb4);
  482. lwkmin = *m + *n + *p;
  483. lwkopt = *m + np + f2cmax(*n,*p) * nb;
  484. }
  485. work[1] = (real) lwkopt;
  486. if (*lwork < lwkmin && ! lquery) {
  487. *info = -12;
  488. }
  489. }
  490. if (*info != 0) {
  491. i__1 = -(*info);
  492. xerbla_("SGGGLM", &i__1, (ftnlen)6);
  493. return;
  494. } else if (lquery) {
  495. return;
  496. }
  497. /* Quick return if possible */
  498. if (*n == 0) {
  499. i__1 = *m;
  500. for (i__ = 1; i__ <= i__1; ++i__) {
  501. x[i__] = 0.f;
  502. }
  503. i__1 = *p;
  504. for (i__ = 1; i__ <= i__1; ++i__) {
  505. y[i__] = 0.f;
  506. }
  507. return;
  508. }
  509. /* Compute the GQR factorization of matrices A and B: */
  510. /* Q**T*A = ( R11 ) M, Q**T*B*Z**T = ( T11 T12 ) M */
  511. /* ( 0 ) N-M ( 0 T22 ) N-M */
  512. /* M M+P-N N-M */
  513. /* where R11 and T22 are upper triangular, and Q and Z are */
  514. /* orthogonal. */
  515. i__1 = *lwork - *m - np;
  516. sggqrf_(n, m, p, &a[a_offset], lda, &work[1], &b[b_offset], ldb, &work[*m
  517. + 1], &work[*m + np + 1], &i__1, info);
  518. lopt = work[*m + np + 1];
  519. /* Update left-hand-side vector d = Q**T*d = ( d1 ) M */
  520. /* ( d2 ) N-M */
  521. i__1 = f2cmax(1,*n);
  522. i__2 = *lwork - *m - np;
  523. sormqr_("Left", "Transpose", n, &c__1, m, &a[a_offset], lda, &work[1], &
  524. d__[1], &i__1, &work[*m + np + 1], &i__2, info);
  525. /* Computing MAX */
  526. i__1 = lopt, i__2 = (integer) work[*m + np + 1];
  527. lopt = f2cmax(i__1,i__2);
  528. /* Solve T22*y2 = d2 for y2 */
  529. if (*n > *m) {
  530. i__1 = *n - *m;
  531. i__2 = *n - *m;
  532. strtrs_("Upper", "No transpose", "Non unit", &i__1, &c__1, &b[*m + 1
  533. + (*m + *p - *n + 1) * b_dim1], ldb, &d__[*m + 1], &i__2,
  534. info);
  535. if (*info > 0) {
  536. *info = 1;
  537. return;
  538. }
  539. i__1 = *n - *m;
  540. scopy_(&i__1, &d__[*m + 1], &c__1, &y[*m + *p - *n + 1], &c__1);
  541. }
  542. /* Set y1 = 0 */
  543. i__1 = *m + *p - *n;
  544. for (i__ = 1; i__ <= i__1; ++i__) {
  545. y[i__] = 0.f;
  546. /* L10: */
  547. }
  548. /* Update d1 = d1 - T12*y2 */
  549. i__1 = *n - *m;
  550. sgemv_("No transpose", m, &i__1, &c_b32, &b[(*m + *p - *n + 1) * b_dim1 +
  551. 1], ldb, &y[*m + *p - *n + 1], &c__1, &c_b34, &d__[1], &c__1);
  552. /* Solve triangular system: R11*x = d1 */
  553. if (*m > 0) {
  554. strtrs_("Upper", "No Transpose", "Non unit", m, &c__1, &a[a_offset],
  555. lda, &d__[1], m, info);
  556. if (*info > 0) {
  557. *info = 2;
  558. return;
  559. }
  560. /* Copy D to X */
  561. scopy_(m, &d__[1], &c__1, &x[1], &c__1);
  562. }
  563. /* Backward transformation y = Z**T *y */
  564. /* Computing MAX */
  565. i__1 = 1, i__2 = *n - *p + 1;
  566. i__3 = f2cmax(1,*p);
  567. i__4 = *lwork - *m - np;
  568. sormrq_("Left", "Transpose", p, &c__1, &np, &b[f2cmax(i__1,i__2) + b_dim1],
  569. ldb, &work[*m + 1], &y[1], &i__3, &work[*m + np + 1], &i__4, info);
  570. /* Computing MAX */
  571. i__1 = lopt, i__2 = (integer) work[*m + np + 1];
  572. work[1] = (real) (*m + np + f2cmax(i__1,i__2));
  573. return;
  574. /* End of SGGGLM */
  575. } /* sggglm_ */