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.

spbtrf.c 26 kB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924
  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 integer c__1 = 1;
  381. static integer c_n1 = -1;
  382. static real c_b18 = 1.f;
  383. static real c_b21 = -1.f;
  384. static integer c__33 = 33;
  385. /* > \brief \b SPBTRF */
  386. /* =========== DOCUMENTATION =========== */
  387. /* Online html documentation available at */
  388. /* http://www.netlib.org/lapack/explore-html/ */
  389. /* > \htmlonly */
  390. /* > Download SPBTRF + dependencies */
  391. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/spbtrf.
  392. f"> */
  393. /* > [TGZ]</a> */
  394. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/spbtrf.
  395. f"> */
  396. /* > [ZIP]</a> */
  397. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/spbtrf.
  398. f"> */
  399. /* > [TXT]</a> */
  400. /* > \endhtmlonly */
  401. /* Definition: */
  402. /* =========== */
  403. /* SUBROUTINE SPBTRF( UPLO, N, KD, AB, LDAB, INFO ) */
  404. /* CHARACTER UPLO */
  405. /* INTEGER INFO, KD, LDAB, N */
  406. /* REAL AB( LDAB, * ) */
  407. /* > \par Purpose: */
  408. /* ============= */
  409. /* > */
  410. /* > \verbatim */
  411. /* > */
  412. /* > SPBTRF computes the Cholesky factorization of a real symmetric */
  413. /* > positive definite band matrix A. */
  414. /* > */
  415. /* > The factorization has the form */
  416. /* > A = U**T * U, if UPLO = 'U', or */
  417. /* > A = L * L**T, if UPLO = 'L', */
  418. /* > where U is an upper triangular matrix and L is lower triangular. */
  419. /* > \endverbatim */
  420. /* Arguments: */
  421. /* ========== */
  422. /* > \param[in] UPLO */
  423. /* > \verbatim */
  424. /* > UPLO is CHARACTER*1 */
  425. /* > = 'U': Upper triangle of A is stored; */
  426. /* > = 'L': Lower triangle of A is stored. */
  427. /* > \endverbatim */
  428. /* > */
  429. /* > \param[in] N */
  430. /* > \verbatim */
  431. /* > N is INTEGER */
  432. /* > The order of the matrix A. N >= 0. */
  433. /* > \endverbatim */
  434. /* > */
  435. /* > \param[in] KD */
  436. /* > \verbatim */
  437. /* > KD is INTEGER */
  438. /* > The number of superdiagonals of the matrix A if UPLO = 'U', */
  439. /* > or the number of subdiagonals if UPLO = 'L'. KD >= 0. */
  440. /* > \endverbatim */
  441. /* > */
  442. /* > \param[in,out] AB */
  443. /* > \verbatim */
  444. /* > AB is REAL array, dimension (LDAB,N) */
  445. /* > On entry, the upper or lower triangle of the symmetric band */
  446. /* > matrix A, stored in the first KD+1 rows of the array. The */
  447. /* > j-th column of A is stored in the j-th column of the array AB */
  448. /* > as follows: */
  449. /* > if UPLO = 'U', AB(kd+1+i-j,j) = A(i,j) for f2cmax(1,j-kd)<=i<=j; */
  450. /* > if UPLO = 'L', AB(1+i-j,j) = A(i,j) for j<=i<=f2cmin(n,j+kd). */
  451. /* > */
  452. /* > On exit, if INFO = 0, the triangular factor U or L from the */
  453. /* > Cholesky factorization A = U**T*U or A = L*L**T of the band */
  454. /* > matrix A, in the same storage format as A. */
  455. /* > \endverbatim */
  456. /* > */
  457. /* > \param[in] LDAB */
  458. /* > \verbatim */
  459. /* > LDAB is INTEGER */
  460. /* > The leading dimension of the array AB. LDAB >= KD+1. */
  461. /* > \endverbatim */
  462. /* > */
  463. /* > \param[out] INFO */
  464. /* > \verbatim */
  465. /* > INFO is INTEGER */
  466. /* > = 0: successful exit */
  467. /* > < 0: if INFO = -i, the i-th argument had an illegal value */
  468. /* > > 0: if INFO = i, the leading minor of order i is not */
  469. /* > positive definite, and the factorization could not be */
  470. /* > completed. */
  471. /* > \endverbatim */
  472. /* Authors: */
  473. /* ======== */
  474. /* > \author Univ. of Tennessee */
  475. /* > \author Univ. of California Berkeley */
  476. /* > \author Univ. of Colorado Denver */
  477. /* > \author NAG Ltd. */
  478. /* > \date December 2016 */
  479. /* > \ingroup realOTHERcomputational */
  480. /* > \par Further Details: */
  481. /* ===================== */
  482. /* > */
  483. /* > \verbatim */
  484. /* > */
  485. /* > The band storage scheme is illustrated by the following example, when */
  486. /* > N = 6, KD = 2, and UPLO = 'U': */
  487. /* > */
  488. /* > On entry: On exit: */
  489. /* > */
  490. /* > * * a13 a24 a35 a46 * * u13 u24 u35 u46 */
  491. /* > * a12 a23 a34 a45 a56 * u12 u23 u34 u45 u56 */
  492. /* > a11 a22 a33 a44 a55 a66 u11 u22 u33 u44 u55 u66 */
  493. /* > */
  494. /* > Similarly, if UPLO = 'L' the format of A is as follows: */
  495. /* > */
  496. /* > On entry: On exit: */
  497. /* > */
  498. /* > a11 a22 a33 a44 a55 a66 l11 l22 l33 l44 l55 l66 */
  499. /* > a21 a32 a43 a54 a65 * l21 l32 l43 l54 l65 * */
  500. /* > a31 a42 a53 a64 * * l31 l42 l53 l64 * * */
  501. /* > */
  502. /* > Array elements marked * are not used by the routine. */
  503. /* > \endverbatim */
  504. /* > \par Contributors: */
  505. /* ================== */
  506. /* > */
  507. /* > Peter Mayes and Giuseppe Radicati, IBM ECSEC, Rome, March 23, 1989 */
  508. /* ===================================================================== */
  509. /* Subroutine */ int spbtrf_(char *uplo, integer *n, integer *kd, real *ab,
  510. integer *ldab, integer *info)
  511. {
  512. /* System generated locals */
  513. integer ab_dim1, ab_offset, i__1, i__2, i__3, i__4;
  514. /* Local variables */
  515. real work[1056] /* was [33][32] */;
  516. integer i__, j;
  517. extern logical lsame_(char *, char *);
  518. extern /* Subroutine */ int sgemm_(char *, char *, integer *, integer *,
  519. integer *, real *, real *, integer *, real *, integer *, real *,
  520. real *, integer *);
  521. integer i2, i3;
  522. extern /* Subroutine */ int strsm_(char *, char *, char *, char *,
  523. integer *, integer *, real *, real *, integer *, real *, integer *
  524. ), ssyrk_(char *, char *, integer
  525. *, integer *, real *, real *, integer *, real *, real *, integer *
  526. ), spbtf2_(char *, integer *, integer *, real *,
  527. integer *, integer *);
  528. integer ib;
  529. extern /* Subroutine */ int spotf2_(char *, integer *, real *, integer *,
  530. integer *);
  531. integer nb, ii, jj;
  532. extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
  533. extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
  534. integer *, integer *, ftnlen, ftnlen);
  535. /* -- LAPACK computational routine (version 3.7.0) -- */
  536. /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
  537. /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
  538. /* December 2016 */
  539. /* ===================================================================== */
  540. /* Test the input parameters. */
  541. /* Parameter adjustments */
  542. ab_dim1 = *ldab;
  543. ab_offset = 1 + ab_dim1 * 1;
  544. ab -= ab_offset;
  545. /* Function Body */
  546. *info = 0;
  547. if (! lsame_(uplo, "U") && ! lsame_(uplo, "L")) {
  548. *info = -1;
  549. } else if (*n < 0) {
  550. *info = -2;
  551. } else if (*kd < 0) {
  552. *info = -3;
  553. } else if (*ldab < *kd + 1) {
  554. *info = -5;
  555. }
  556. if (*info != 0) {
  557. i__1 = -(*info);
  558. xerbla_("SPBTRF", &i__1, (ftnlen)6);
  559. return 0;
  560. }
  561. /* Quick return if possible */
  562. if (*n == 0) {
  563. return 0;
  564. }
  565. /* Determine the block size for this environment */
  566. nb = ilaenv_(&c__1, "SPBTRF", uplo, n, kd, &c_n1, &c_n1, (ftnlen)6, (
  567. ftnlen)1);
  568. /* The block size must not exceed the semi-bandwidth KD, and must not */
  569. /* exceed the limit set by the size of the local array WORK. */
  570. nb = f2cmin(nb,32);
  571. if (nb <= 1 || nb > *kd) {
  572. /* Use unblocked code */
  573. spbtf2_(uplo, n, kd, &ab[ab_offset], ldab, info);
  574. } else {
  575. /* Use blocked code */
  576. if (lsame_(uplo, "U")) {
  577. /* Compute the Cholesky factorization of a symmetric band */
  578. /* matrix, given the upper triangle of the matrix in band */
  579. /* storage. */
  580. /* Zero the upper triangle of the work array. */
  581. i__1 = nb;
  582. for (j = 1; j <= i__1; ++j) {
  583. i__2 = j - 1;
  584. for (i__ = 1; i__ <= i__2; ++i__) {
  585. work[i__ + j * 33 - 34] = 0.f;
  586. /* L10: */
  587. }
  588. /* L20: */
  589. }
  590. /* Process the band matrix one diagonal block at a time. */
  591. i__1 = *n;
  592. i__2 = nb;
  593. for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) {
  594. /* Computing MIN */
  595. i__3 = nb, i__4 = *n - i__ + 1;
  596. ib = f2cmin(i__3,i__4);
  597. /* Factorize the diagonal block */
  598. i__3 = *ldab - 1;
  599. spotf2_(uplo, &ib, &ab[*kd + 1 + i__ * ab_dim1], &i__3, &ii);
  600. if (ii != 0) {
  601. *info = i__ + ii - 1;
  602. goto L150;
  603. }
  604. if (i__ + ib <= *n) {
  605. /* Update the relevant part of the trailing submatrix. */
  606. /* If A11 denotes the diagonal block which has just been */
  607. /* factorized, then we need to update the remaining */
  608. /* blocks in the diagram: */
  609. /* A11 A12 A13 */
  610. /* A22 A23 */
  611. /* A33 */
  612. /* The numbers of rows and columns in the partitioning */
  613. /* are IB, I2, I3 respectively. The blocks A12, A22 and */
  614. /* A23 are empty if IB = KD. The upper triangle of A13 */
  615. /* lies outside the band. */
  616. /* Computing MIN */
  617. i__3 = *kd - ib, i__4 = *n - i__ - ib + 1;
  618. i2 = f2cmin(i__3,i__4);
  619. /* Computing MIN */
  620. i__3 = ib, i__4 = *n - i__ - *kd + 1;
  621. i3 = f2cmin(i__3,i__4);
  622. if (i2 > 0) {
  623. /* Update A12 */
  624. i__3 = *ldab - 1;
  625. i__4 = *ldab - 1;
  626. strsm_("Left", "Upper", "Transpose", "Non-unit", &ib,
  627. &i2, &c_b18, &ab[*kd + 1 + i__ * ab_dim1], &
  628. i__3, &ab[*kd + 1 - ib + (i__ + ib) * ab_dim1]
  629. , &i__4);
  630. /* Update A22 */
  631. i__3 = *ldab - 1;
  632. i__4 = *ldab - 1;
  633. ssyrk_("Upper", "Transpose", &i2, &ib, &c_b21, &ab[*
  634. kd + 1 - ib + (i__ + ib) * ab_dim1], &i__3, &
  635. c_b18, &ab[*kd + 1 + (i__ + ib) * ab_dim1], &
  636. i__4);
  637. }
  638. if (i3 > 0) {
  639. /* Copy the lower triangle of A13 into the work array. */
  640. i__3 = i3;
  641. for (jj = 1; jj <= i__3; ++jj) {
  642. i__4 = ib;
  643. for (ii = jj; ii <= i__4; ++ii) {
  644. work[ii + jj * 33 - 34] = ab[ii - jj + 1 + (
  645. jj + i__ + *kd - 1) * ab_dim1];
  646. /* L30: */
  647. }
  648. /* L40: */
  649. }
  650. /* Update A13 (in the work array). */
  651. i__3 = *ldab - 1;
  652. strsm_("Left", "Upper", "Transpose", "Non-unit", &ib,
  653. &i3, &c_b18, &ab[*kd + 1 + i__ * ab_dim1], &
  654. i__3, work, &c__33);
  655. /* Update A23 */
  656. if (i2 > 0) {
  657. i__3 = *ldab - 1;
  658. i__4 = *ldab - 1;
  659. sgemm_("Transpose", "No Transpose", &i2, &i3, &ib,
  660. &c_b21, &ab[*kd + 1 - ib + (i__ + ib) *
  661. ab_dim1], &i__3, work, &c__33, &c_b18, &
  662. ab[ib + 1 + (i__ + *kd) * ab_dim1], &i__4);
  663. }
  664. /* Update A33 */
  665. i__3 = *ldab - 1;
  666. ssyrk_("Upper", "Transpose", &i3, &ib, &c_b21, work, &
  667. c__33, &c_b18, &ab[*kd + 1 + (i__ + *kd) *
  668. ab_dim1], &i__3);
  669. /* Copy the lower triangle of A13 back into place. */
  670. i__3 = i3;
  671. for (jj = 1; jj <= i__3; ++jj) {
  672. i__4 = ib;
  673. for (ii = jj; ii <= i__4; ++ii) {
  674. ab[ii - jj + 1 + (jj + i__ + *kd - 1) *
  675. ab_dim1] = work[ii + jj * 33 - 34];
  676. /* L50: */
  677. }
  678. /* L60: */
  679. }
  680. }
  681. }
  682. /* L70: */
  683. }
  684. } else {
  685. /* Compute the Cholesky factorization of a symmetric band */
  686. /* matrix, given the lower triangle of the matrix in band */
  687. /* storage. */
  688. /* Zero the lower triangle of the work array. */
  689. i__2 = nb;
  690. for (j = 1; j <= i__2; ++j) {
  691. i__1 = nb;
  692. for (i__ = j + 1; i__ <= i__1; ++i__) {
  693. work[i__ + j * 33 - 34] = 0.f;
  694. /* L80: */
  695. }
  696. /* L90: */
  697. }
  698. /* Process the band matrix one diagonal block at a time. */
  699. i__2 = *n;
  700. i__1 = nb;
  701. for (i__ = 1; i__1 < 0 ? i__ >= i__2 : i__ <= i__2; i__ += i__1) {
  702. /* Computing MIN */
  703. i__3 = nb, i__4 = *n - i__ + 1;
  704. ib = f2cmin(i__3,i__4);
  705. /* Factorize the diagonal block */
  706. i__3 = *ldab - 1;
  707. spotf2_(uplo, &ib, &ab[i__ * ab_dim1 + 1], &i__3, &ii);
  708. if (ii != 0) {
  709. *info = i__ + ii - 1;
  710. goto L150;
  711. }
  712. if (i__ + ib <= *n) {
  713. /* Update the relevant part of the trailing submatrix. */
  714. /* If A11 denotes the diagonal block which has just been */
  715. /* factorized, then we need to update the remaining */
  716. /* blocks in the diagram: */
  717. /* A11 */
  718. /* A21 A22 */
  719. /* A31 A32 A33 */
  720. /* The numbers of rows and columns in the partitioning */
  721. /* are IB, I2, I3 respectively. The blocks A21, A22 and */
  722. /* A32 are empty if IB = KD. The lower triangle of A31 */
  723. /* lies outside the band. */
  724. /* Computing MIN */
  725. i__3 = *kd - ib, i__4 = *n - i__ - ib + 1;
  726. i2 = f2cmin(i__3,i__4);
  727. /* Computing MIN */
  728. i__3 = ib, i__4 = *n - i__ - *kd + 1;
  729. i3 = f2cmin(i__3,i__4);
  730. if (i2 > 0) {
  731. /* Update A21 */
  732. i__3 = *ldab - 1;
  733. i__4 = *ldab - 1;
  734. strsm_("Right", "Lower", "Transpose", "Non-unit", &i2,
  735. &ib, &c_b18, &ab[i__ * ab_dim1 + 1], &i__3, &
  736. ab[ib + 1 + i__ * ab_dim1], &i__4);
  737. /* Update A22 */
  738. i__3 = *ldab - 1;
  739. i__4 = *ldab - 1;
  740. ssyrk_("Lower", "No Transpose", &i2, &ib, &c_b21, &ab[
  741. ib + 1 + i__ * ab_dim1], &i__3, &c_b18, &ab[(
  742. i__ + ib) * ab_dim1 + 1], &i__4);
  743. }
  744. if (i3 > 0) {
  745. /* Copy the upper triangle of A31 into the work array. */
  746. i__3 = ib;
  747. for (jj = 1; jj <= i__3; ++jj) {
  748. i__4 = f2cmin(jj,i3);
  749. for (ii = 1; ii <= i__4; ++ii) {
  750. work[ii + jj * 33 - 34] = ab[*kd + 1 - jj +
  751. ii + (jj + i__ - 1) * ab_dim1];
  752. /* L100: */
  753. }
  754. /* L110: */
  755. }
  756. /* Update A31 (in the work array). */
  757. i__3 = *ldab - 1;
  758. strsm_("Right", "Lower", "Transpose", "Non-unit", &i3,
  759. &ib, &c_b18, &ab[i__ * ab_dim1 + 1], &i__3,
  760. work, &c__33);
  761. /* Update A32 */
  762. if (i2 > 0) {
  763. i__3 = *ldab - 1;
  764. i__4 = *ldab - 1;
  765. sgemm_("No transpose", "Transpose", &i3, &i2, &ib,
  766. &c_b21, work, &c__33, &ab[ib + 1 + i__ *
  767. ab_dim1], &i__3, &c_b18, &ab[*kd + 1 - ib
  768. + (i__ + ib) * ab_dim1], &i__4);
  769. }
  770. /* Update A33 */
  771. i__3 = *ldab - 1;
  772. ssyrk_("Lower", "No Transpose", &i3, &ib, &c_b21,
  773. work, &c__33, &c_b18, &ab[(i__ + *kd) *
  774. ab_dim1 + 1], &i__3);
  775. /* Copy the upper triangle of A31 back into place. */
  776. i__3 = ib;
  777. for (jj = 1; jj <= i__3; ++jj) {
  778. i__4 = f2cmin(jj,i3);
  779. for (ii = 1; ii <= i__4; ++ii) {
  780. ab[*kd + 1 - jj + ii + (jj + i__ - 1) *
  781. ab_dim1] = work[ii + jj * 33 - 34];
  782. /* L120: */
  783. }
  784. /* L130: */
  785. }
  786. }
  787. }
  788. /* L140: */
  789. }
  790. }
  791. }
  792. return 0;
  793. L150:
  794. return 0;
  795. /* End of SPBTRF */
  796. } /* spbtrf_ */