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.

zlaqps.c 24 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844
  1. /* f2c.h -- Standard Fortran to C header file */
  2. /** barf [ba:rf] 2. "He suggested using FORTRAN, and everybody barfed."
  3. - From The Shogakukan DICTIONARY OF NEW ENGLISH (Second edition) */
  4. #ifndef F2C_INCLUDE
  5. #define F2C_INCLUDE
  6. #include <math.h>
  7. #include <stdlib.h>
  8. #include <string.h>
  9. #include <stdio.h>
  10. #include <complex.h>
  11. #ifdef complex
  12. #undef complex
  13. #endif
  14. #ifdef I
  15. #undef I
  16. #endif
  17. #if defined(_WIN64)
  18. typedef long long BLASLONG;
  19. typedef unsigned long long BLASULONG;
  20. #else
  21. typedef long BLASLONG;
  22. typedef unsigned long BLASULONG;
  23. #endif
  24. #ifdef LAPACK_ILP64
  25. typedef BLASLONG blasint;
  26. #if defined(_WIN64)
  27. #define blasabs(x) llabs(x)
  28. #else
  29. #define blasabs(x) labs(x)
  30. #endif
  31. #else
  32. typedef int blasint;
  33. #define blasabs(x) abs(x)
  34. #endif
  35. typedef blasint integer;
  36. typedef unsigned int uinteger;
  37. typedef char *address;
  38. typedef short int shortint;
  39. typedef float real;
  40. typedef double doublereal;
  41. typedef struct { real r, i; } complex;
  42. typedef struct { doublereal r, i; } doublecomplex;
  43. static inline _Complex float Cf(complex *z) {return z->r + z->i*_Complex_I;}
  44. static inline _Complex double Cd(doublecomplex *z) {return z->r + z->i*_Complex_I;}
  45. static inline _Complex float * _pCf(complex *z) {return (_Complex float*)z;}
  46. static inline _Complex double * _pCd(doublecomplex *z) {return (_Complex double*)z;}
  47. #define pCf(z) (*_pCf(z))
  48. #define pCd(z) (*_pCd(z))
  49. typedef int logical;
  50. typedef short int shortlogical;
  51. typedef char logical1;
  52. typedef char integer1;
  53. #define TRUE_ (1)
  54. #define FALSE_ (0)
  55. /* Extern is for use with -E */
  56. #ifndef Extern
  57. #define Extern extern
  58. #endif
  59. /* I/O stuff */
  60. typedef int flag;
  61. typedef int ftnlen;
  62. typedef int ftnint;
  63. /*external read, write*/
  64. typedef struct
  65. { flag cierr;
  66. ftnint ciunit;
  67. flag ciend;
  68. char *cifmt;
  69. ftnint cirec;
  70. } cilist;
  71. /*internal read, write*/
  72. typedef struct
  73. { flag icierr;
  74. char *iciunit;
  75. flag iciend;
  76. char *icifmt;
  77. ftnint icirlen;
  78. ftnint icirnum;
  79. } icilist;
  80. /*open*/
  81. typedef struct
  82. { flag oerr;
  83. ftnint ounit;
  84. char *ofnm;
  85. ftnlen ofnmlen;
  86. char *osta;
  87. char *oacc;
  88. char *ofm;
  89. ftnint orl;
  90. char *oblnk;
  91. } olist;
  92. /*close*/
  93. typedef struct
  94. { flag cerr;
  95. ftnint cunit;
  96. char *csta;
  97. } cllist;
  98. /*rewind, backspace, endfile*/
  99. typedef struct
  100. { flag aerr;
  101. ftnint aunit;
  102. } alist;
  103. /* inquire */
  104. typedef struct
  105. { flag inerr;
  106. ftnint inunit;
  107. char *infile;
  108. ftnlen infilen;
  109. ftnint *inex; /*parameters in standard's order*/
  110. ftnint *inopen;
  111. ftnint *innum;
  112. ftnint *innamed;
  113. char *inname;
  114. ftnlen innamlen;
  115. char *inacc;
  116. ftnlen inacclen;
  117. char *inseq;
  118. ftnlen inseqlen;
  119. char *indir;
  120. ftnlen indirlen;
  121. char *infmt;
  122. ftnlen infmtlen;
  123. char *inform;
  124. ftnint informlen;
  125. char *inunf;
  126. ftnlen inunflen;
  127. ftnint *inrecl;
  128. ftnint *innrec;
  129. char *inblank;
  130. ftnlen inblanklen;
  131. } inlist;
  132. #define VOID void
  133. union Multitype { /* for multiple entry points */
  134. integer1 g;
  135. shortint h;
  136. integer i;
  137. /* longint j; */
  138. real r;
  139. doublereal d;
  140. complex c;
  141. doublecomplex z;
  142. };
  143. typedef union Multitype Multitype;
  144. struct Vardesc { /* for Namelist */
  145. char *name;
  146. char *addr;
  147. ftnlen *dims;
  148. int type;
  149. };
  150. typedef struct Vardesc Vardesc;
  151. struct Namelist {
  152. char *name;
  153. Vardesc **vars;
  154. int nvars;
  155. };
  156. typedef struct Namelist Namelist;
  157. #define abs(x) ((x) >= 0 ? (x) : -(x))
  158. #define dabs(x) (fabs(x))
  159. #define f2cmin(a,b) ((a) <= (b) ? (a) : (b))
  160. #define f2cmax(a,b) ((a) >= (b) ? (a) : (b))
  161. #define dmin(a,b) (f2cmin(a,b))
  162. #define dmax(a,b) (f2cmax(a,b))
  163. #define bit_test(a,b) ((a) >> (b) & 1)
  164. #define bit_clear(a,b) ((a) & ~((uinteger)1 << (b)))
  165. #define bit_set(a,b) ((a) | ((uinteger)1 << (b)))
  166. #define abort_() { sig_die("Fortran abort routine called", 1); }
  167. #define c_abs(z) (cabsf(Cf(z)))
  168. #define c_cos(R,Z) { pCf(R)=ccos(Cf(Z)); }
  169. #define c_div(c, a, b) {pCf(c) = Cf(a)/Cf(b);}
  170. #define z_div(c, a, b) {pCd(c) = Cd(a)/Cd(b);}
  171. #define c_exp(R, Z) {pCf(R) = cexpf(Cf(Z));}
  172. #define c_log(R, Z) {pCf(R) = clogf(Cf(Z));}
  173. #define c_sin(R, Z) {pCf(R) = csinf(Cf(Z));}
  174. //#define c_sqrt(R, Z) {*(R) = csqrtf(Cf(Z));}
  175. #define c_sqrt(R, Z) {pCf(R) = csqrtf(Cf(Z));}
  176. #define d_abs(x) (fabs(*(x)))
  177. #define d_acos(x) (acos(*(x)))
  178. #define d_asin(x) (asin(*(x)))
  179. #define d_atan(x) (atan(*(x)))
  180. #define d_atn2(x, y) (atan2(*(x),*(y)))
  181. #define d_cnjg(R, Z) { pCd(R) = conj(Cd(Z)); }
  182. #define r_cnjg(R, Z) { pCf(R) = conj(Cf(Z)); }
  183. #define d_cos(x) (cos(*(x)))
  184. #define d_cosh(x) (cosh(*(x)))
  185. #define d_dim(__a, __b) ( *(__a) > *(__b) ? *(__a) - *(__b) : 0.0 )
  186. #define d_exp(x) (exp(*(x)))
  187. #define d_imag(z) (cimag(Cd(z)))
  188. #define r_imag(z) (cimag(Cf(z)))
  189. #define d_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
  190. #define r_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
  191. #define d_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
  192. #define r_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
  193. #define d_log(x) (log(*(x)))
  194. #define d_mod(x, y) (fmod(*(x), *(y)))
  195. #define u_nint(__x) ((__x)>=0 ? floor((__x) + .5) : -floor(.5 - (__x)))
  196. #define d_nint(x) u_nint(*(x))
  197. #define u_sign(__a,__b) ((__b) >= 0 ? ((__a) >= 0 ? (__a) : -(__a)) : -((__a) >= 0 ? (__a) : -(__a)))
  198. #define d_sign(a,b) u_sign(*(a),*(b))
  199. #define r_sign(a,b) u_sign(*(a),*(b))
  200. #define d_sin(x) (sin(*(x)))
  201. #define d_sinh(x) (sinh(*(x)))
  202. #define d_sqrt(x) (sqrt(*(x)))
  203. #define d_tan(x) (tan(*(x)))
  204. #define d_tanh(x) (tanh(*(x)))
  205. #define i_abs(x) abs(*(x))
  206. #define i_dnnt(x) ((integer)u_nint(*(x)))
  207. #define i_len(s, n) (n)
  208. #define i_nint(x) ((integer)u_nint(*(x)))
  209. #define i_sign(a,b) ((integer)u_sign((integer)*(a),(integer)*(b)))
  210. #define pow_dd(ap, bp) ( pow(*(ap), *(bp)))
  211. #define pow_si(B,E) spow_ui(*(B),*(E))
  212. #define pow_ri(B,E) spow_ui(*(B),*(E))
  213. #define pow_di(B,E) dpow_ui(*(B),*(E))
  214. #define pow_zi(p, a, b) {pCd(p) = zpow_ui(Cd(a), *(b));}
  215. #define pow_ci(p, a, b) {pCf(p) = cpow_ui(Cf(a), *(b));}
  216. #define pow_zz(R,A,B) {pCd(R) = cpow(Cd(A),*(B));}
  217. #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++ = ' '; }
  218. #define s_cmp(a,b,c,d) ((integer)strncmp((a),(b),f2cmin((c),(d))))
  219. #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]; }
  220. #define sig_die(s, kill) { exit(1); }
  221. #define s_stop(s, n) {exit(0);}
  222. static char junk[] = "\n@(#)LIBF77 VERSION 19990503\n";
  223. #define z_abs(z) (cabs(Cd(z)))
  224. #define z_exp(R, Z) {pCd(R) = cexp(Cd(Z));}
  225. #define z_sqrt(R, Z) {pCd(R) = csqrt(Cd(Z));}
  226. #define myexit_() break;
  227. #define mycycle() continue;
  228. #define myceiling(w) {ceil(w)}
  229. #define myhuge(w) {HUGE_VAL}
  230. //#define mymaxloc_(w,s,e,n) {if (sizeof(*(w)) == sizeof(double)) dmaxloc_((w),*(s),*(e),n); else dmaxloc_((w),*(s),*(e),n);}
  231. #define mymaxloc(w,s,e,n) {dmaxloc_(w,*(s),*(e),n)}
  232. /* procedure parameter types for -A and -C++ */
  233. #define F2C_proc_par_types 1
  234. #ifdef __cplusplus
  235. typedef logical (*L_fp)(...);
  236. #else
  237. typedef logical (*L_fp)();
  238. #endif
  239. static float spow_ui(float x, integer n) {
  240. float pow=1.0; unsigned long int u;
  241. if(n != 0) {
  242. if(n < 0) n = -n, x = 1/x;
  243. for(u = n; ; ) {
  244. if(u & 01) pow *= x;
  245. if(u >>= 1) x *= x;
  246. else break;
  247. }
  248. }
  249. return pow;
  250. }
  251. static double dpow_ui(double x, integer n) {
  252. double pow=1.0; unsigned long int u;
  253. if(n != 0) {
  254. if(n < 0) n = -n, x = 1/x;
  255. for(u = n; ; ) {
  256. if(u & 01) pow *= x;
  257. if(u >>= 1) x *= x;
  258. else break;
  259. }
  260. }
  261. return pow;
  262. }
  263. static _Complex float cpow_ui(_Complex float x, integer n) {
  264. _Complex float pow=1.0; unsigned long int u;
  265. if(n != 0) {
  266. if(n < 0) n = -n, x = 1/x;
  267. for(u = n; ; ) {
  268. if(u & 01) pow *= x;
  269. if(u >>= 1) x *= x;
  270. else break;
  271. }
  272. }
  273. return pow;
  274. }
  275. static _Complex double zpow_ui(_Complex double x, integer n) {
  276. _Complex double pow=1.0; unsigned long int u;
  277. if(n != 0) {
  278. if(n < 0) n = -n, x = 1/x;
  279. for(u = n; ; ) {
  280. if(u & 01) pow *= x;
  281. if(u >>= 1) x *= x;
  282. else break;
  283. }
  284. }
  285. return pow;
  286. }
  287. static integer pow_ii(integer x, integer n) {
  288. integer pow; unsigned long int u;
  289. if (n <= 0) {
  290. if (n == 0 || x == 1) pow = 1;
  291. else if (x != -1) pow = x == 0 ? 1/x : 0;
  292. else n = -n;
  293. }
  294. if ((n > 0) || !(n == 0 || x == 1 || x != -1)) {
  295. u = n;
  296. for(pow = 1; ; ) {
  297. if(u & 01) pow *= x;
  298. if(u >>= 1) x *= x;
  299. else break;
  300. }
  301. }
  302. return pow;
  303. }
  304. static integer dmaxloc_(double *w, integer s, integer e, integer *n)
  305. {
  306. double m; integer i, mi;
  307. for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
  308. if (w[i-1]>m) mi=i ,m=w[i-1];
  309. return mi-s+1;
  310. }
  311. static integer smaxloc_(float *w, integer s, integer e, integer *n)
  312. {
  313. float m; integer i, mi;
  314. for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
  315. if (w[i-1]>m) mi=i ,m=w[i-1];
  316. return mi-s+1;
  317. }
  318. static inline void cdotc_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
  319. integer n = *n_, incx = *incx_, incy = *incy_, i;
  320. _Complex float zdotc = 0.0;
  321. if (incx == 1 && incy == 1) {
  322. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  323. zdotc += conjf(Cf(&x[i])) * Cf(&y[i]);
  324. }
  325. } else {
  326. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  327. zdotc += conjf(Cf(&x[i*incx])) * Cf(&y[i*incy]);
  328. }
  329. }
  330. pCf(z) = zdotc;
  331. }
  332. static inline void zdotc_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
  333. integer n = *n_, incx = *incx_, incy = *incy_, i;
  334. _Complex double zdotc = 0.0;
  335. if (incx == 1 && incy == 1) {
  336. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  337. zdotc += conj(Cd(&x[i])) * Cd(&y[i]);
  338. }
  339. } else {
  340. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  341. zdotc += conj(Cd(&x[i*incx])) * Cd(&y[i*incy]);
  342. }
  343. }
  344. pCd(z) = zdotc;
  345. }
  346. static inline void cdotu_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
  347. integer n = *n_, incx = *incx_, incy = *incy_, i;
  348. _Complex float zdotc = 0.0;
  349. if (incx == 1 && incy == 1) {
  350. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  351. zdotc += Cf(&x[i]) * Cf(&y[i]);
  352. }
  353. } else {
  354. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  355. zdotc += Cf(&x[i*incx]) * Cf(&y[i*incy]);
  356. }
  357. }
  358. pCf(z) = zdotc;
  359. }
  360. static inline void zdotu_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
  361. integer n = *n_, incx = *incx_, incy = *incy_, i;
  362. _Complex double zdotc = 0.0;
  363. if (incx == 1 && incy == 1) {
  364. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  365. zdotc += Cd(&x[i]) * Cd(&y[i]);
  366. }
  367. } else {
  368. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  369. zdotc += Cd(&x[i*incx]) * Cd(&y[i*incy]);
  370. }
  371. }
  372. pCd(z) = zdotc;
  373. }
  374. #endif
  375. /* -- translated by f2c (version 20000121).
  376. You must link the resulting object file with the libraries:
  377. -lf2c -lm (in that order)
  378. */
  379. /* Table of constant values */
  380. static doublecomplex c_b1 = {0.,0.};
  381. static doublecomplex c_b2 = {1.,0.};
  382. static integer c__1 = 1;
  383. /* > \brief \b ZLAQPS computes a step of QR factorization with column pivoting of a real m-by-n matrix A by us
  384. ing BLAS level 3. */
  385. /* =========== DOCUMENTATION =========== */
  386. /* Online html documentation available at */
  387. /* http://www.netlib.org/lapack/explore-html/ */
  388. /* > \htmlonly */
  389. /* > Download ZLAQPS + dependencies */
  390. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/zlaqps.
  391. f"> */
  392. /* > [TGZ]</a> */
  393. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/zlaqps.
  394. f"> */
  395. /* > [ZIP]</a> */
  396. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/zlaqps.
  397. f"> */
  398. /* > [TXT]</a> */
  399. /* > \endhtmlonly */
  400. /* Definition: */
  401. /* =========== */
  402. /* SUBROUTINE ZLAQPS( M, N, OFFSET, NB, KB, A, LDA, JPVT, TAU, VN1, */
  403. /* VN2, AUXV, F, LDF ) */
  404. /* INTEGER KB, LDA, LDF, M, N, NB, OFFSET */
  405. /* INTEGER JPVT( * ) */
  406. /* DOUBLE PRECISION VN1( * ), VN2( * ) */
  407. /* COMPLEX*16 A( LDA, * ), AUXV( * ), F( LDF, * ), TAU( * ) */
  408. /* > \par Purpose: */
  409. /* ============= */
  410. /* > */
  411. /* > \verbatim */
  412. /* > */
  413. /* > ZLAQPS computes a step of QR factorization with column pivoting */
  414. /* > of a complex M-by-N matrix A by using Blas-3. It tries to factorize */
  415. /* > NB columns from A starting from the row OFFSET+1, and updates all */
  416. /* > of the matrix with Blas-3 xGEMM. */
  417. /* > */
  418. /* > In some cases, due to catastrophic cancellations, it cannot */
  419. /* > factorize NB columns. Hence, the actual number of factorized */
  420. /* > columns is returned in KB. */
  421. /* > */
  422. /* > Block A(1:OFFSET,1:N) is accordingly pivoted, but not factorized. */
  423. /* > \endverbatim */
  424. /* Arguments: */
  425. /* ========== */
  426. /* > \param[in] M */
  427. /* > \verbatim */
  428. /* > M is INTEGER */
  429. /* > The number of rows of the matrix A. M >= 0. */
  430. /* > \endverbatim */
  431. /* > */
  432. /* > \param[in] N */
  433. /* > \verbatim */
  434. /* > N is INTEGER */
  435. /* > The number of columns of the matrix A. N >= 0 */
  436. /* > \endverbatim */
  437. /* > */
  438. /* > \param[in] OFFSET */
  439. /* > \verbatim */
  440. /* > OFFSET is INTEGER */
  441. /* > The number of rows of A that have been factorized in */
  442. /* > previous steps. */
  443. /* > \endverbatim */
  444. /* > */
  445. /* > \param[in] NB */
  446. /* > \verbatim */
  447. /* > NB is INTEGER */
  448. /* > The number of columns to factorize. */
  449. /* > \endverbatim */
  450. /* > */
  451. /* > \param[out] KB */
  452. /* > \verbatim */
  453. /* > KB is INTEGER */
  454. /* > The number of columns actually factorized. */
  455. /* > \endverbatim */
  456. /* > */
  457. /* > \param[in,out] A */
  458. /* > \verbatim */
  459. /* > A is COMPLEX*16 array, dimension (LDA,N) */
  460. /* > On entry, the M-by-N matrix A. */
  461. /* > On exit, block A(OFFSET+1:M,1:KB) is the triangular */
  462. /* > factor obtained and block A(1:OFFSET,1:N) has been */
  463. /* > accordingly pivoted, but no factorized. */
  464. /* > The rest of the matrix, block A(OFFSET+1:M,KB+1:N) has */
  465. /* > been updated. */
  466. /* > \endverbatim */
  467. /* > */
  468. /* > \param[in] LDA */
  469. /* > \verbatim */
  470. /* > LDA is INTEGER */
  471. /* > The leading dimension of the array A. LDA >= f2cmax(1,M). */
  472. /* > \endverbatim */
  473. /* > */
  474. /* > \param[in,out] JPVT */
  475. /* > \verbatim */
  476. /* > JPVT is INTEGER array, dimension (N) */
  477. /* > JPVT(I) = K <==> Column K of the full matrix A has been */
  478. /* > permuted into position I in AP. */
  479. /* > \endverbatim */
  480. /* > */
  481. /* > \param[out] TAU */
  482. /* > \verbatim */
  483. /* > TAU is COMPLEX*16 array, dimension (KB) */
  484. /* > The scalar factors of the elementary reflectors. */
  485. /* > \endverbatim */
  486. /* > */
  487. /* > \param[in,out] VN1 */
  488. /* > \verbatim */
  489. /* > VN1 is DOUBLE PRECISION array, dimension (N) */
  490. /* > The vector with the partial column norms. */
  491. /* > \endverbatim */
  492. /* > */
  493. /* > \param[in,out] VN2 */
  494. /* > \verbatim */
  495. /* > VN2 is DOUBLE PRECISION array, dimension (N) */
  496. /* > The vector with the exact column norms. */
  497. /* > \endverbatim */
  498. /* > */
  499. /* > \param[in,out] AUXV */
  500. /* > \verbatim */
  501. /* > AUXV is COMPLEX*16 array, dimension (NB) */
  502. /* > Auxiliary vector. */
  503. /* > \endverbatim */
  504. /* > */
  505. /* > \param[in,out] F */
  506. /* > \verbatim */
  507. /* > F is COMPLEX*16 array, dimension (LDF,NB) */
  508. /* > Matrix F**H = L * Y**H * A. */
  509. /* > \endverbatim */
  510. /* > */
  511. /* > \param[in] LDF */
  512. /* > \verbatim */
  513. /* > LDF is INTEGER */
  514. /* > The leading dimension of the array F. LDF >= f2cmax(1,N). */
  515. /* > \endverbatim */
  516. /* Authors: */
  517. /* ======== */
  518. /* > \author Univ. of Tennessee */
  519. /* > \author Univ. of California Berkeley */
  520. /* > \author Univ. of Colorado Denver */
  521. /* > \author NAG Ltd. */
  522. /* > \date December 2016 */
  523. /* > \ingroup complex16OTHERauxiliary */
  524. /* > \par Contributors: */
  525. /* ================== */
  526. /* > */
  527. /* > G. Quintana-Orti, Depto. de Informatica, Universidad Jaime I, Spain */
  528. /* > X. Sun, Computer Science Dept., Duke University, USA */
  529. /* > \n */
  530. /* > Partial column norm updating strategy modified on April 2011 */
  531. /* > Z. Drmac and Z. Bujanovic, Dept. of Mathematics, */
  532. /* > University of Zagreb, Croatia. */
  533. /* > \par References: */
  534. /* ================ */
  535. /* > */
  536. /* > LAPACK Working Note 176 */
  537. /* > \htmlonly */
  538. /* > <a href="http://www.netlib.org/lapack/lawnspdf/lawn176.pdf">[PDF]</a> */
  539. /* > \endhtmlonly */
  540. /* ===================================================================== */
  541. /* Subroutine */ int zlaqps_(integer *m, integer *n, integer *offset, integer
  542. *nb, integer *kb, doublecomplex *a, integer *lda, integer *jpvt,
  543. doublecomplex *tau, doublereal *vn1, doublereal *vn2, doublecomplex *
  544. auxv, doublecomplex *f, integer *ldf)
  545. {
  546. /* System generated locals */
  547. integer a_dim1, a_offset, f_dim1, f_offset, i__1, i__2, i__3;
  548. doublereal d__1, d__2;
  549. doublecomplex z__1;
  550. /* Local variables */
  551. doublereal temp, temp2;
  552. integer j, k;
  553. doublereal tol3z;
  554. integer itemp;
  555. extern /* Subroutine */ int zgemm_(char *, char *, integer *, integer *,
  556. integer *, doublecomplex *, doublecomplex *, integer *,
  557. doublecomplex *, integer *, doublecomplex *, doublecomplex *,
  558. integer *), zgemv_(char *, integer *, integer *,
  559. doublecomplex *, doublecomplex *, integer *, doublecomplex *,
  560. integer *, doublecomplex *, doublecomplex *, integer *),
  561. zswap_(integer *, doublecomplex *, integer *, doublecomplex *,
  562. integer *);
  563. extern doublereal dznrm2_(integer *, doublecomplex *, integer *), dlamch_(
  564. char *);
  565. integer rk;
  566. extern integer idamax_(integer *, doublereal *, integer *);
  567. integer lsticc;
  568. extern /* Subroutine */ int zlarfg_(integer *, doublecomplex *,
  569. doublecomplex *, integer *, doublecomplex *);
  570. integer lastrk;
  571. doublecomplex akk;
  572. integer pvt;
  573. /* -- LAPACK auxiliary routine (version 3.7.0) -- */
  574. /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
  575. /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
  576. /* December 2016 */
  577. /* ===================================================================== */
  578. /* Parameter adjustments */
  579. a_dim1 = *lda;
  580. a_offset = 1 + a_dim1 * 1;
  581. a -= a_offset;
  582. --jpvt;
  583. --tau;
  584. --vn1;
  585. --vn2;
  586. --auxv;
  587. f_dim1 = *ldf;
  588. f_offset = 1 + f_dim1 * 1;
  589. f -= f_offset;
  590. /* Function Body */
  591. /* Computing MIN */
  592. i__1 = *m, i__2 = *n + *offset;
  593. lastrk = f2cmin(i__1,i__2);
  594. lsticc = 0;
  595. k = 0;
  596. tol3z = sqrt(dlamch_("Epsilon"));
  597. /* Beginning of while loop. */
  598. L10:
  599. if (k < *nb && lsticc == 0) {
  600. ++k;
  601. rk = *offset + k;
  602. /* Determine ith pivot column and swap if necessary */
  603. i__1 = *n - k + 1;
  604. pvt = k - 1 + idamax_(&i__1, &vn1[k], &c__1);
  605. if (pvt != k) {
  606. zswap_(m, &a[pvt * a_dim1 + 1], &c__1, &a[k * a_dim1 + 1], &c__1);
  607. i__1 = k - 1;
  608. zswap_(&i__1, &f[pvt + f_dim1], ldf, &f[k + f_dim1], ldf);
  609. itemp = jpvt[pvt];
  610. jpvt[pvt] = jpvt[k];
  611. jpvt[k] = itemp;
  612. vn1[pvt] = vn1[k];
  613. vn2[pvt] = vn2[k];
  614. }
  615. /* Apply previous Householder reflectors to column K: */
  616. /* A(RK:M,K) := A(RK:M,K) - A(RK:M,1:K-1)*F(K,1:K-1)**H. */
  617. if (k > 1) {
  618. i__1 = k - 1;
  619. for (j = 1; j <= i__1; ++j) {
  620. i__2 = k + j * f_dim1;
  621. d_cnjg(&z__1, &f[k + j * f_dim1]);
  622. f[i__2].r = z__1.r, f[i__2].i = z__1.i;
  623. /* L20: */
  624. }
  625. i__1 = *m - rk + 1;
  626. i__2 = k - 1;
  627. z__1.r = -1., z__1.i = 0.;
  628. zgemv_("No transpose", &i__1, &i__2, &z__1, &a[rk + a_dim1], lda,
  629. &f[k + f_dim1], ldf, &c_b2, &a[rk + k * a_dim1], &c__1);
  630. i__1 = k - 1;
  631. for (j = 1; j <= i__1; ++j) {
  632. i__2 = k + j * f_dim1;
  633. d_cnjg(&z__1, &f[k + j * f_dim1]);
  634. f[i__2].r = z__1.r, f[i__2].i = z__1.i;
  635. /* L30: */
  636. }
  637. }
  638. /* Generate elementary reflector H(k). */
  639. if (rk < *m) {
  640. i__1 = *m - rk + 1;
  641. zlarfg_(&i__1, &a[rk + k * a_dim1], &a[rk + 1 + k * a_dim1], &
  642. c__1, &tau[k]);
  643. } else {
  644. zlarfg_(&c__1, &a[rk + k * a_dim1], &a[rk + k * a_dim1], &c__1, &
  645. tau[k]);
  646. }
  647. i__1 = rk + k * a_dim1;
  648. akk.r = a[i__1].r, akk.i = a[i__1].i;
  649. i__1 = rk + k * a_dim1;
  650. a[i__1].r = 1., a[i__1].i = 0.;
  651. /* Compute Kth column of F: */
  652. /* Compute F(K+1:N,K) := tau(K)*A(RK:M,K+1:N)**H*A(RK:M,K). */
  653. if (k < *n) {
  654. i__1 = *m - rk + 1;
  655. i__2 = *n - k;
  656. zgemv_("Conjugate transpose", &i__1, &i__2, &tau[k], &a[rk + (k +
  657. 1) * a_dim1], lda, &a[rk + k * a_dim1], &c__1, &c_b1, &f[
  658. k + 1 + k * f_dim1], &c__1);
  659. }
  660. /* Padding F(1:K,K) with zeros. */
  661. i__1 = k;
  662. for (j = 1; j <= i__1; ++j) {
  663. i__2 = j + k * f_dim1;
  664. f[i__2].r = 0., f[i__2].i = 0.;
  665. /* L40: */
  666. }
  667. /* Incremental updating of F: */
  668. /* F(1:N,K) := F(1:N,K) - tau(K)*F(1:N,1:K-1)*A(RK:M,1:K-1)**H */
  669. /* *A(RK:M,K). */
  670. if (k > 1) {
  671. i__1 = *m - rk + 1;
  672. i__2 = k - 1;
  673. i__3 = k;
  674. z__1.r = -tau[i__3].r, z__1.i = -tau[i__3].i;
  675. zgemv_("Conjugate transpose", &i__1, &i__2, &z__1, &a[rk + a_dim1]
  676. , lda, &a[rk + k * a_dim1], &c__1, &c_b1, &auxv[1], &c__1);
  677. i__1 = k - 1;
  678. zgemv_("No transpose", n, &i__1, &c_b2, &f[f_dim1 + 1], ldf, &
  679. auxv[1], &c__1, &c_b2, &f[k * f_dim1 + 1], &c__1);
  680. }
  681. /* Update the current row of A: */
  682. /* A(RK,K+1:N) := A(RK,K+1:N) - A(RK,1:K)*F(K+1:N,1:K)**H. */
  683. if (k < *n) {
  684. i__1 = *n - k;
  685. z__1.r = -1., z__1.i = 0.;
  686. zgemm_("No transpose", "Conjugate transpose", &c__1, &i__1, &k, &
  687. z__1, &a[rk + a_dim1], lda, &f[k + 1 + f_dim1], ldf, &
  688. c_b2, &a[rk + (k + 1) * a_dim1], lda);
  689. }
  690. /* Update partial column norms. */
  691. if (rk < lastrk) {
  692. i__1 = *n;
  693. for (j = k + 1; j <= i__1; ++j) {
  694. if (vn1[j] != 0.) {
  695. /* NOTE: The following 4 lines follow from the analysis in */
  696. /* Lapack Working Note 176. */
  697. temp = z_abs(&a[rk + j * a_dim1]) / vn1[j];
  698. /* Computing MAX */
  699. d__1 = 0., d__2 = (temp + 1.) * (1. - temp);
  700. temp = f2cmax(d__1,d__2);
  701. /* Computing 2nd power */
  702. d__1 = vn1[j] / vn2[j];
  703. temp2 = temp * (d__1 * d__1);
  704. if (temp2 <= tol3z) {
  705. vn2[j] = (doublereal) lsticc;
  706. lsticc = j;
  707. } else {
  708. vn1[j] *= sqrt(temp);
  709. }
  710. }
  711. /* L50: */
  712. }
  713. }
  714. i__1 = rk + k * a_dim1;
  715. a[i__1].r = akk.r, a[i__1].i = akk.i;
  716. /* End of while loop. */
  717. goto L10;
  718. }
  719. *kb = k;
  720. rk = *offset + *kb;
  721. /* Apply the block reflector to the rest of the matrix: */
  722. /* A(OFFSET+KB+1:M,KB+1:N) := A(OFFSET+KB+1:M,KB+1:N) - */
  723. /* A(OFFSET+KB+1:M,1:KB)*F(KB+1:N,1:KB)**H. */
  724. /* Computing MIN */
  725. i__1 = *n, i__2 = *m - *offset;
  726. if (*kb < f2cmin(i__1,i__2)) {
  727. i__1 = *m - rk;
  728. i__2 = *n - *kb;
  729. z__1.r = -1., z__1.i = 0.;
  730. zgemm_("No transpose", "Conjugate transpose", &i__1, &i__2, kb, &z__1,
  731. &a[rk + 1 + a_dim1], lda, &f[*kb + 1 + f_dim1], ldf, &c_b2, &
  732. a[rk + 1 + (*kb + 1) * a_dim1], lda);
  733. }
  734. /* Recomputation of difficult columns. */
  735. L60:
  736. if (lsticc > 0) {
  737. itemp = i_dnnt(&vn2[lsticc]);
  738. i__1 = *m - rk;
  739. vn1[lsticc] = dznrm2_(&i__1, &a[rk + 1 + lsticc * a_dim1], &c__1);
  740. /* NOTE: The computation of VN1( LSTICC ) relies on the fact that */
  741. /* SNRM2 does not fail on vectors with norm below the value of */
  742. /* SQRT(DLAMCH('S')) */
  743. vn2[lsticc] = vn1[lsticc];
  744. lsticc = itemp;
  745. goto L60;
  746. }
  747. return 0;
  748. /* End of ZLAQPS */
  749. } /* zlaqps_ */