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.

slatbs.c 38 kB

1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192939495969798991001011021031041051061071081091101111121131141151161171181191201211221231241251261271281291301311321331341351361371381391401411421431441451461471481491501511521531541551561571581591601611621631641651661671681691701711721731741751761771781791801811821831841851861871881891901911921931941951961971981992002012022032042052062072082092102112122132142152162172182192202212222232242252262272282292302312322332342352362372382392402412422432442452462472482492502512522532542552562572582592602612622632642652662672682692702712722732742752762772782792802812822832842852862872882892902912922932942952962972982993003013023033043053063073083093103113123133143153163173183193203213223233243253263273283293303313323333343353363373383393403413423433443453463473483493503513523533543553563573583593603613623633643653663673683693703713723733743753763773783793803813823833843853863873883893903913923933943953963973983994004014024034044054064074084094104114124134144154164174184194204214224234244254264274284294304314324334344354364374384394404414424434444454464474484494504514524534544554564574584594604614624634644654664674684694704714724734744754764774784794804814824834844854864874884894904914924934944954964974984995005015025035045055065075085095105115125135145155165175185195205215225235245255265275285295305315325335345355365375385395405415425435445455465475485495505515525535545555565575585595605615625635645655665675685695705715725735745755765775785795805815825835845855865875885895905915925935945955965975985996006016026036046056066076086096106116126136146156166176186196206216226236246256266276286296306316326336346356366376386396406416426436446456466476486496506516526536546556566576586596606616626636646656666676686696706716726736746756766776786796806816826836846856866876886896906916926936946956966976986997007017027037047057067077087097107117127137147157167177187197207217227237247257267277287297307317327337347357367377387397407417427437447457467477487497507517527537547557567577587597607617627637647657667677687697707717727737747757767777787797807817827837847857867877887897907917927937947957967977987998008018028038048058068078088098108118128138148158168178188198208218228238248258268278288298308318328338348358368378388398408418428438448458468478488498508518528538548558568578588598608618628638648658668678688698708718728738748758768778788798808818828838848858868878888898908918928938948958968978988999009019029039049059069079089099109119129139149159169179189199209219229239249259269279289299309319329339349359369379389399409419429439449459469479489499509519529539549559569579589599609619629639649659669679689699709719729739749759769779789799809819829839849859869879889899909919929939949959969979989991000100110021003100410051006100710081009101010111012101310141015101610171018101910201021102210231024102510261027102810291030103110321033103410351036103710381039104010411042104310441045104610471048104910501051105210531054105510561057105810591060106110621063106410651066106710681069107010711072107310741075107610771078107910801081108210831084108510861087108810891090109110921093109410951096109710981099110011011102110311041105110611071108110911101111111211131114111511161117111811191120112111221123112411251126112711281129113011311132113311341135113611371138113911401141114211431144114511461147114811491150115111521153115411551156115711581159116011611162116311641165116611671168116911701171117211731174117511761177117811791180118111821183118411851186118711881189119011911192119311941195119611971198119912001201120212031204120512061207120812091210121112121213121412151216121712181219122012211222122312241225122612271228122912301231123212331234123512361237123812391240124112421243124412451246124712481249125012511252125312541255125612571258125912601261126212631264126512661267126812691270127112721273127412751276127712781279128012811282128312841285128612871288128912901291129212931294129512961297129812991300130113021303130413051306130713081309131013111312131313141315131613171318131913201321132213231324132513261327132813291330133113321333133413351336133713381339134013411342134313441345134613471348134913501351135213531354135513561357135813591360136113621363136413651366136713681369137013711372137313741375137613771378137913801381138213831384138513861387138813891390139113921393139413951396139713981399140014011402140314041405140614071408140914101411141214131414141514161417141814191420
  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 int logical;
  52. typedef short int shortlogical;
  53. typedef char logical1;
  54. typedef char integer1;
  55. #define TRUE_ (1)
  56. #define FALSE_ (0)
  57. /* Extern is for use with -E */
  58. #ifndef Extern
  59. #define Extern extern
  60. #endif
  61. /* I/O stuff */
  62. typedef int flag;
  63. typedef int ftnlen;
  64. typedef int ftnint;
  65. /*external read, write*/
  66. typedef struct
  67. { flag cierr;
  68. ftnint ciunit;
  69. flag ciend;
  70. char *cifmt;
  71. ftnint cirec;
  72. } cilist;
  73. /*internal read, write*/
  74. typedef struct
  75. { flag icierr;
  76. char *iciunit;
  77. flag iciend;
  78. char *icifmt;
  79. ftnint icirlen;
  80. ftnint icirnum;
  81. } icilist;
  82. /*open*/
  83. typedef struct
  84. { flag oerr;
  85. ftnint ounit;
  86. char *ofnm;
  87. ftnlen ofnmlen;
  88. char *osta;
  89. char *oacc;
  90. char *ofm;
  91. ftnint orl;
  92. char *oblnk;
  93. } olist;
  94. /*close*/
  95. typedef struct
  96. { flag cerr;
  97. ftnint cunit;
  98. char *csta;
  99. } cllist;
  100. /*rewind, backspace, endfile*/
  101. typedef struct
  102. { flag aerr;
  103. ftnint aunit;
  104. } alist;
  105. /* inquire */
  106. typedef struct
  107. { flag inerr;
  108. ftnint inunit;
  109. char *infile;
  110. ftnlen infilen;
  111. ftnint *inex; /*parameters in standard's order*/
  112. ftnint *inopen;
  113. ftnint *innum;
  114. ftnint *innamed;
  115. char *inname;
  116. ftnlen innamlen;
  117. char *inacc;
  118. ftnlen inacclen;
  119. char *inseq;
  120. ftnlen inseqlen;
  121. char *indir;
  122. ftnlen indirlen;
  123. char *infmt;
  124. ftnlen infmtlen;
  125. char *inform;
  126. ftnint informlen;
  127. char *inunf;
  128. ftnlen inunflen;
  129. ftnint *inrecl;
  130. ftnint *innrec;
  131. char *inblank;
  132. ftnlen inblanklen;
  133. } inlist;
  134. #define VOID void
  135. union Multitype { /* for multiple entry points */
  136. integer1 g;
  137. shortint h;
  138. integer i;
  139. /* longint j; */
  140. real r;
  141. doublereal d;
  142. complex c;
  143. doublecomplex z;
  144. };
  145. typedef union Multitype Multitype;
  146. struct Vardesc { /* for Namelist */
  147. char *name;
  148. char *addr;
  149. ftnlen *dims;
  150. int type;
  151. };
  152. typedef struct Vardesc Vardesc;
  153. struct Namelist {
  154. char *name;
  155. Vardesc **vars;
  156. int nvars;
  157. };
  158. typedef struct Namelist Namelist;
  159. #define abs(x) ((x) >= 0 ? (x) : -(x))
  160. #define dabs(x) (fabs(x))
  161. #define f2cmin(a,b) ((a) <= (b) ? (a) : (b))
  162. #define f2cmax(a,b) ((a) >= (b) ? (a) : (b))
  163. #define dmin(a,b) (f2cmin(a,b))
  164. #define dmax(a,b) (f2cmax(a,b))
  165. #define bit_test(a,b) ((a) >> (b) & 1)
  166. #define bit_clear(a,b) ((a) & ~((uinteger)1 << (b)))
  167. #define bit_set(a,b) ((a) | ((uinteger)1 << (b)))
  168. #define abort_() { sig_die("Fortran abort routine called", 1); }
  169. #define c_abs(z) (cabsf(Cf(z)))
  170. #define c_cos(R,Z) { pCf(R)=ccos(Cf(Z)); }
  171. #ifdef _MSC_VER
  172. #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]);}
  173. #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]);}
  174. #else
  175. #define c_div(c, a, b) {pCf(c) = Cf(a)/Cf(b);}
  176. #define z_div(c, a, b) {pCd(c) = Cd(a)/Cd(b);}
  177. #endif
  178. #define c_exp(R, Z) {pCf(R) = cexpf(Cf(Z));}
  179. #define c_log(R, Z) {pCf(R) = clogf(Cf(Z));}
  180. #define c_sin(R, Z) {pCf(R) = csinf(Cf(Z));}
  181. //#define c_sqrt(R, Z) {*(R) = csqrtf(Cf(Z));}
  182. #define c_sqrt(R, Z) {pCf(R) = csqrtf(Cf(Z));}
  183. #define d_abs(x) (fabs(*(x)))
  184. #define d_acos(x) (acos(*(x)))
  185. #define d_asin(x) (asin(*(x)))
  186. #define d_atan(x) (atan(*(x)))
  187. #define d_atn2(x, y) (atan2(*(x),*(y)))
  188. #define d_cnjg(R, Z) { pCd(R) = conj(Cd(Z)); }
  189. #define r_cnjg(R, Z) { pCf(R) = conjf(Cf(Z)); }
  190. #define d_cos(x) (cos(*(x)))
  191. #define d_cosh(x) (cosh(*(x)))
  192. #define d_dim(__a, __b) ( *(__a) > *(__b) ? *(__a) - *(__b) : 0.0 )
  193. #define d_exp(x) (exp(*(x)))
  194. #define d_imag(z) (cimag(Cd(z)))
  195. #define r_imag(z) (cimagf(Cf(z)))
  196. #define d_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
  197. #define r_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
  198. #define d_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
  199. #define r_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
  200. #define d_log(x) (log(*(x)))
  201. #define d_mod(x, y) (fmod(*(x), *(y)))
  202. #define u_nint(__x) ((__x)>=0 ? floor((__x) + .5) : -floor(.5 - (__x)))
  203. #define d_nint(x) u_nint(*(x))
  204. #define u_sign(__a,__b) ((__b) >= 0 ? ((__a) >= 0 ? (__a) : -(__a)) : -((__a) >= 0 ? (__a) : -(__a)))
  205. #define d_sign(a,b) u_sign(*(a),*(b))
  206. #define r_sign(a,b) u_sign(*(a),*(b))
  207. #define d_sin(x) (sin(*(x)))
  208. #define d_sinh(x) (sinh(*(x)))
  209. #define d_sqrt(x) (sqrt(*(x)))
  210. #define d_tan(x) (tan(*(x)))
  211. #define d_tanh(x) (tanh(*(x)))
  212. #define i_abs(x) abs(*(x))
  213. #define i_dnnt(x) ((integer)u_nint(*(x)))
  214. #define i_len(s, n) (n)
  215. #define i_nint(x) ((integer)u_nint(*(x)))
  216. #define i_sign(a,b) ((integer)u_sign((integer)*(a),(integer)*(b)))
  217. #define pow_dd(ap, bp) ( pow(*(ap), *(bp)))
  218. #define pow_si(B,E) spow_ui(*(B),*(E))
  219. #define pow_ri(B,E) spow_ui(*(B),*(E))
  220. #define pow_di(B,E) dpow_ui(*(B),*(E))
  221. #define pow_zi(p, a, b) {pCd(p) = zpow_ui(Cd(a), *(b));}
  222. #define pow_ci(p, a, b) {pCf(p) = cpow_ui(Cf(a), *(b));}
  223. #define pow_zz(R,A,B) {pCd(R) = cpow(Cd(A),*(B));}
  224. #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++ = ' '; }
  225. #define s_cmp(a,b,c,d) ((integer)strncmp((a),(b),f2cmin((c),(d))))
  226. #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]; }
  227. #define sig_die(s, kill) { exit(1); }
  228. #define s_stop(s, n) {exit(0);}
  229. static char junk[] = "\n@(#)LIBF77 VERSION 19990503\n";
  230. #define z_abs(z) (cabs(Cd(z)))
  231. #define z_exp(R, Z) {pCd(R) = cexp(Cd(Z));}
  232. #define z_sqrt(R, Z) {pCd(R) = csqrt(Cd(Z));}
  233. #define myexit_() break;
  234. #define mycycle() continue;
  235. #define myceiling(w) {ceil(w)}
  236. #define myhuge(w) {HUGE_VAL}
  237. //#define mymaxloc_(w,s,e,n) {if (sizeof(*(w)) == sizeof(double)) dmaxloc_((w),*(s),*(e),n); else dmaxloc_((w),*(s),*(e),n);}
  238. #define mymaxloc(w,s,e,n) {dmaxloc_(w,*(s),*(e),n)}
  239. /* procedure parameter types for -A and -C++ */
  240. #define F2C_proc_par_types 1
  241. #ifdef __cplusplus
  242. typedef logical (*L_fp)(...);
  243. #else
  244. typedef logical (*L_fp)();
  245. #endif
  246. static float spow_ui(float x, integer n) {
  247. float pow=1.0; unsigned long int u;
  248. if(n != 0) {
  249. if(n < 0) n = -n, x = 1/x;
  250. for(u = n; ; ) {
  251. if(u & 01) pow *= x;
  252. if(u >>= 1) x *= x;
  253. else break;
  254. }
  255. }
  256. return pow;
  257. }
  258. static double dpow_ui(double x, integer n) {
  259. double pow=1.0; unsigned long int u;
  260. if(n != 0) {
  261. if(n < 0) n = -n, x = 1/x;
  262. for(u = n; ; ) {
  263. if(u & 01) pow *= x;
  264. if(u >>= 1) x *= x;
  265. else break;
  266. }
  267. }
  268. return pow;
  269. }
  270. #ifdef _MSC_VER
  271. static _Fcomplex cpow_ui(complex x, integer n) {
  272. complex pow={1.0,0.0}; unsigned long int u;
  273. if(n != 0) {
  274. if(n < 0) n = -n, x.r = 1/x.r, x.i=1/x.i;
  275. for(u = n; ; ) {
  276. if(u & 01) pow.r *= x.r, pow.i *= x.i;
  277. if(u >>= 1) x.r *= x.r, x.i *= x.i;
  278. else break;
  279. }
  280. }
  281. _Fcomplex p={pow.r, pow.i};
  282. return p;
  283. }
  284. #else
  285. static _Complex float cpow_ui(_Complex float x, integer n) {
  286. _Complex float pow=1.0; unsigned long int u;
  287. if(n != 0) {
  288. if(n < 0) n = -n, x = 1/x;
  289. for(u = n; ; ) {
  290. if(u & 01) pow *= x;
  291. if(u >>= 1) x *= x;
  292. else break;
  293. }
  294. }
  295. return pow;
  296. }
  297. #endif
  298. #ifdef _MSC_VER
  299. static _Dcomplex zpow_ui(_Dcomplex x, integer n) {
  300. _Dcomplex pow={1.0,0.0}; unsigned long int u;
  301. if(n != 0) {
  302. if(n < 0) n = -n, x._Val[0] = 1/x._Val[0], x._Val[1] =1/x._Val[1];
  303. for(u = n; ; ) {
  304. if(u & 01) pow._Val[0] *= x._Val[0], pow._Val[1] *= x._Val[1];
  305. if(u >>= 1) x._Val[0] *= x._Val[0], x._Val[1] *= x._Val[1];
  306. else break;
  307. }
  308. }
  309. _Dcomplex p = {pow._Val[0], pow._Val[1]};
  310. return p;
  311. }
  312. #else
  313. static _Complex double zpow_ui(_Complex double x, integer n) {
  314. _Complex double pow=1.0; unsigned long int u;
  315. if(n != 0) {
  316. if(n < 0) n = -n, x = 1/x;
  317. for(u = n; ; ) {
  318. if(u & 01) pow *= x;
  319. if(u >>= 1) x *= x;
  320. else break;
  321. }
  322. }
  323. return pow;
  324. }
  325. #endif
  326. static integer pow_ii(integer x, integer n) {
  327. integer pow; unsigned long int u;
  328. if (n <= 0) {
  329. if (n == 0 || x == 1) pow = 1;
  330. else if (x != -1) pow = x == 0 ? 1/x : 0;
  331. else n = -n;
  332. }
  333. if ((n > 0) || !(n == 0 || x == 1 || x != -1)) {
  334. u = n;
  335. for(pow = 1; ; ) {
  336. if(u & 01) pow *= x;
  337. if(u >>= 1) x *= x;
  338. else break;
  339. }
  340. }
  341. return pow;
  342. }
  343. static integer dmaxloc_(double *w, integer s, integer e, integer *n)
  344. {
  345. double m; integer i, mi;
  346. for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
  347. if (w[i-1]>m) mi=i ,m=w[i-1];
  348. return mi-s+1;
  349. }
  350. static integer smaxloc_(float *w, integer s, integer e, integer *n)
  351. {
  352. float m; integer i, mi;
  353. for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
  354. if (w[i-1]>m) mi=i ,m=w[i-1];
  355. return mi-s+1;
  356. }
  357. static inline void cdotc_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
  358. integer n = *n_, incx = *incx_, incy = *incy_, i;
  359. #ifdef _MSC_VER
  360. _Fcomplex zdotc = {0.0, 0.0};
  361. if (incx == 1 && incy == 1) {
  362. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  363. zdotc._Val[0] += conjf(Cf(&x[i]))._Val[0] * Cf(&y[i])._Val[0];
  364. zdotc._Val[1] += conjf(Cf(&x[i]))._Val[1] * Cf(&y[i])._Val[1];
  365. }
  366. } else {
  367. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  368. zdotc._Val[0] += conjf(Cf(&x[i*incx]))._Val[0] * Cf(&y[i*incy])._Val[0];
  369. zdotc._Val[1] += conjf(Cf(&x[i*incx]))._Val[1] * Cf(&y[i*incy])._Val[1];
  370. }
  371. }
  372. pCf(z) = zdotc;
  373. }
  374. #else
  375. _Complex float zdotc = 0.0;
  376. if (incx == 1 && incy == 1) {
  377. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  378. zdotc += conjf(Cf(&x[i])) * Cf(&y[i]);
  379. }
  380. } else {
  381. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  382. zdotc += conjf(Cf(&x[i*incx])) * Cf(&y[i*incy]);
  383. }
  384. }
  385. pCf(z) = zdotc;
  386. }
  387. #endif
  388. static inline void zdotc_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
  389. integer n = *n_, incx = *incx_, incy = *incy_, i;
  390. #ifdef _MSC_VER
  391. _Dcomplex zdotc = {0.0, 0.0};
  392. if (incx == 1 && incy == 1) {
  393. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  394. zdotc._Val[0] += conj(Cd(&x[i]))._Val[0] * Cd(&y[i])._Val[0];
  395. zdotc._Val[1] += conj(Cd(&x[i]))._Val[1] * Cd(&y[i])._Val[1];
  396. }
  397. } else {
  398. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  399. zdotc._Val[0] += conj(Cd(&x[i*incx]))._Val[0] * Cd(&y[i*incy])._Val[0];
  400. zdotc._Val[1] += conj(Cd(&x[i*incx]))._Val[1] * Cd(&y[i*incy])._Val[1];
  401. }
  402. }
  403. pCd(z) = zdotc;
  404. }
  405. #else
  406. _Complex double zdotc = 0.0;
  407. if (incx == 1 && incy == 1) {
  408. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  409. zdotc += conj(Cd(&x[i])) * Cd(&y[i]);
  410. }
  411. } else {
  412. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  413. zdotc += conj(Cd(&x[i*incx])) * Cd(&y[i*incy]);
  414. }
  415. }
  416. pCd(z) = zdotc;
  417. }
  418. #endif
  419. static inline void cdotu_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
  420. integer n = *n_, incx = *incx_, incy = *incy_, i;
  421. #ifdef _MSC_VER
  422. _Fcomplex zdotc = {0.0, 0.0};
  423. if (incx == 1 && incy == 1) {
  424. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  425. zdotc._Val[0] += Cf(&x[i])._Val[0] * Cf(&y[i])._Val[0];
  426. zdotc._Val[1] += Cf(&x[i])._Val[1] * Cf(&y[i])._Val[1];
  427. }
  428. } else {
  429. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  430. zdotc._Val[0] += Cf(&x[i*incx])._Val[0] * Cf(&y[i*incy])._Val[0];
  431. zdotc._Val[1] += Cf(&x[i*incx])._Val[1] * Cf(&y[i*incy])._Val[1];
  432. }
  433. }
  434. pCf(z) = zdotc;
  435. }
  436. #else
  437. _Complex float zdotc = 0.0;
  438. if (incx == 1 && incy == 1) {
  439. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  440. zdotc += Cf(&x[i]) * Cf(&y[i]);
  441. }
  442. } else {
  443. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  444. zdotc += Cf(&x[i*incx]) * Cf(&y[i*incy]);
  445. }
  446. }
  447. pCf(z) = zdotc;
  448. }
  449. #endif
  450. static inline void zdotu_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
  451. integer n = *n_, incx = *incx_, incy = *incy_, i;
  452. #ifdef _MSC_VER
  453. _Dcomplex zdotc = {0.0, 0.0};
  454. if (incx == 1 && incy == 1) {
  455. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  456. zdotc._Val[0] += Cd(&x[i])._Val[0] * Cd(&y[i])._Val[0];
  457. zdotc._Val[1] += Cd(&x[i])._Val[1] * Cd(&y[i])._Val[1];
  458. }
  459. } else {
  460. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  461. zdotc._Val[0] += Cd(&x[i*incx])._Val[0] * Cd(&y[i*incy])._Val[0];
  462. zdotc._Val[1] += Cd(&x[i*incx])._Val[1] * Cd(&y[i*incy])._Val[1];
  463. }
  464. }
  465. pCd(z) = zdotc;
  466. }
  467. #else
  468. _Complex double zdotc = 0.0;
  469. if (incx == 1 && incy == 1) {
  470. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  471. zdotc += Cd(&x[i]) * Cd(&y[i]);
  472. }
  473. } else {
  474. for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
  475. zdotc += Cd(&x[i*incx]) * Cd(&y[i*incy]);
  476. }
  477. }
  478. pCd(z) = zdotc;
  479. }
  480. #endif
  481. /* -- translated by f2c (version 20000121).
  482. You must link the resulting object file with the libraries:
  483. -lf2c -lm (in that order)
  484. */
  485. /* Table of constant values */
  486. static integer c__1 = 1;
  487. static real c_b36 = .5f;
  488. /* > \brief \b SLATBS solves a triangular banded system of equations. */
  489. /* =========== DOCUMENTATION =========== */
  490. /* Online html documentation available at */
  491. /* http://www.netlib.org/lapack/explore-html/ */
  492. /* > \htmlonly */
  493. /* > Download SLATBS + dependencies */
  494. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/slatbs.
  495. f"> */
  496. /* > [TGZ]</a> */
  497. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/slatbs.
  498. f"> */
  499. /* > [ZIP]</a> */
  500. /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/slatbs.
  501. f"> */
  502. /* > [TXT]</a> */
  503. /* > \endhtmlonly */
  504. /* Definition: */
  505. /* =========== */
  506. /* SUBROUTINE SLATBS( UPLO, TRANS, DIAG, NORMIN, N, KD, AB, LDAB, X, */
  507. /* SCALE, CNORM, INFO ) */
  508. /* CHARACTER DIAG, NORMIN, TRANS, UPLO */
  509. /* INTEGER INFO, KD, LDAB, N */
  510. /* REAL SCALE */
  511. /* REAL AB( LDAB, * ), CNORM( * ), X( * ) */
  512. /* > \par Purpose: */
  513. /* ============= */
  514. /* > */
  515. /* > \verbatim */
  516. /* > */
  517. /* > SLATBS solves one of the triangular systems */
  518. /* > */
  519. /* > A *x = s*b or A**T*x = s*b */
  520. /* > */
  521. /* > with scaling to prevent overflow, where A is an upper or lower */
  522. /* > triangular band matrix. Here A**T denotes the transpose of A, x and b */
  523. /* > are n-element vectors, and s is a scaling factor, usually less than */
  524. /* > or equal to 1, chosen so that the components of x will be less than */
  525. /* > the overflow threshold. If the unscaled problem will not cause */
  526. /* > overflow, the Level 2 BLAS routine STBSV is called. If the matrix A */
  527. /* > is singular (A(j,j) = 0 for some j), then s is set to 0 and a */
  528. /* > non-trivial solution to A*x = 0 is returned. */
  529. /* > \endverbatim */
  530. /* Arguments: */
  531. /* ========== */
  532. /* > \param[in] UPLO */
  533. /* > \verbatim */
  534. /* > UPLO is CHARACTER*1 */
  535. /* > Specifies whether the matrix A is upper or lower triangular. */
  536. /* > = 'U': Upper triangular */
  537. /* > = 'L': Lower triangular */
  538. /* > \endverbatim */
  539. /* > */
  540. /* > \param[in] TRANS */
  541. /* > \verbatim */
  542. /* > TRANS is CHARACTER*1 */
  543. /* > Specifies the operation applied to A. */
  544. /* > = 'N': Solve A * x = s*b (No transpose) */
  545. /* > = 'T': Solve A**T* x = s*b (Transpose) */
  546. /* > = 'C': Solve A**T* x = s*b (Conjugate transpose = Transpose) */
  547. /* > \endverbatim */
  548. /* > */
  549. /* > \param[in] DIAG */
  550. /* > \verbatim */
  551. /* > DIAG is CHARACTER*1 */
  552. /* > Specifies whether or not the matrix A is unit triangular. */
  553. /* > = 'N': Non-unit triangular */
  554. /* > = 'U': Unit triangular */
  555. /* > \endverbatim */
  556. /* > */
  557. /* > \param[in] NORMIN */
  558. /* > \verbatim */
  559. /* > NORMIN is CHARACTER*1 */
  560. /* > Specifies whether CNORM has been set or not. */
  561. /* > = 'Y': CNORM contains the column norms on entry */
  562. /* > = 'N': CNORM is not set on entry. On exit, the norms will */
  563. /* > be computed and stored in CNORM. */
  564. /* > \endverbatim */
  565. /* > */
  566. /* > \param[in] N */
  567. /* > \verbatim */
  568. /* > N is INTEGER */
  569. /* > The order of the matrix A. N >= 0. */
  570. /* > \endverbatim */
  571. /* > */
  572. /* > \param[in] KD */
  573. /* > \verbatim */
  574. /* > KD is INTEGER */
  575. /* > The number of subdiagonals or superdiagonals in the */
  576. /* > triangular matrix A. KD >= 0. */
  577. /* > \endverbatim */
  578. /* > */
  579. /* > \param[in] AB */
  580. /* > \verbatim */
  581. /* > AB is REAL array, dimension (LDAB,N) */
  582. /* > The upper or lower triangular band matrix A, stored in the */
  583. /* > first KD+1 rows of the array. The j-th column of A is stored */
  584. /* > in the j-th column of the array AB as follows: */
  585. /* > if UPLO = 'U', AB(kd+1+i-j,j) = A(i,j) for f2cmax(1,j-kd)<=i<=j; */
  586. /* > if UPLO = 'L', AB(1+i-j,j) = A(i,j) for j<=i<=f2cmin(n,j+kd). */
  587. /* > \endverbatim */
  588. /* > */
  589. /* > \param[in] LDAB */
  590. /* > \verbatim */
  591. /* > LDAB is INTEGER */
  592. /* > The leading dimension of the array AB. LDAB >= KD+1. */
  593. /* > \endverbatim */
  594. /* > */
  595. /* > \param[in,out] X */
  596. /* > \verbatim */
  597. /* > X is REAL array, dimension (N) */
  598. /* > On entry, the right hand side b of the triangular system. */
  599. /* > On exit, X is overwritten by the solution vector x. */
  600. /* > \endverbatim */
  601. /* > */
  602. /* > \param[out] SCALE */
  603. /* > \verbatim */
  604. /* > SCALE is REAL */
  605. /* > The scaling factor s for the triangular system */
  606. /* > A * x = s*b or A**T* x = s*b. */
  607. /* > If SCALE = 0, the matrix A is singular or badly scaled, and */
  608. /* > the vector x is an exact or approximate solution to A*x = 0. */
  609. /* > \endverbatim */
  610. /* > */
  611. /* > \param[in,out] CNORM */
  612. /* > \verbatim */
  613. /* > CNORM is REAL array, dimension (N) */
  614. /* > */
  615. /* > If NORMIN = 'Y', CNORM is an input argument and CNORM(j) */
  616. /* > contains the norm of the off-diagonal part of the j-th column */
  617. /* > of A. If TRANS = 'N', CNORM(j) must be greater than or equal */
  618. /* > to the infinity-norm, and if TRANS = 'T' or 'C', CNORM(j) */
  619. /* > must be greater than or equal to the 1-norm. */
  620. /* > */
  621. /* > If NORMIN = 'N', CNORM is an output argument and CNORM(j) */
  622. /* > returns the 1-norm of the offdiagonal part of the j-th column */
  623. /* > of A. */
  624. /* > \endverbatim */
  625. /* > */
  626. /* > \param[out] INFO */
  627. /* > \verbatim */
  628. /* > INFO is INTEGER */
  629. /* > = 0: successful exit */
  630. /* > < 0: if INFO = -k, the k-th argument had an illegal value */
  631. /* > \endverbatim */
  632. /* Authors: */
  633. /* ======== */
  634. /* > \author Univ. of Tennessee */
  635. /* > \author Univ. of California Berkeley */
  636. /* > \author Univ. of Colorado Denver */
  637. /* > \author NAG Ltd. */
  638. /* > \date December 2016 */
  639. /* > \ingroup realOTHERauxiliary */
  640. /* > \par Further Details: */
  641. /* ===================== */
  642. /* > */
  643. /* > \verbatim */
  644. /* > */
  645. /* > A rough bound on x is computed; if that is less than overflow, STBSV */
  646. /* > is called, otherwise, specific code is used which checks for possible */
  647. /* > overflow or divide-by-zero at every operation. */
  648. /* > */
  649. /* > A columnwise scheme is used for solving A*x = b. The basic algorithm */
  650. /* > if A is lower triangular is */
  651. /* > */
  652. /* > x[1:n] := b[1:n] */
  653. /* > for j = 1, ..., n */
  654. /* > x(j) := x(j) / A(j,j) */
  655. /* > x[j+1:n] := x[j+1:n] - x(j) * A[j+1:n,j] */
  656. /* > end */
  657. /* > */
  658. /* > Define bounds on the components of x after j iterations of the loop: */
  659. /* > M(j) = bound on x[1:j] */
  660. /* > G(j) = bound on x[j+1:n] */
  661. /* > Initially, let M(0) = 0 and G(0) = f2cmax{x(i), i=1,...,n}. */
  662. /* > */
  663. /* > Then for iteration j+1 we have */
  664. /* > M(j+1) <= G(j) / | A(j+1,j+1) | */
  665. /* > G(j+1) <= G(j) + M(j+1) * | A[j+2:n,j+1] | */
  666. /* > <= G(j) ( 1 + CNORM(j+1) / | A(j+1,j+1) | ) */
  667. /* > */
  668. /* > where CNORM(j+1) is greater than or equal to the infinity-norm of */
  669. /* > column j+1 of A, not counting the diagonal. Hence */
  670. /* > */
  671. /* > G(j) <= G(0) product ( 1 + CNORM(i) / | A(i,i) | ) */
  672. /* > 1<=i<=j */
  673. /* > and */
  674. /* > */
  675. /* > |x(j)| <= ( G(0) / |A(j,j)| ) product ( 1 + CNORM(i) / |A(i,i)| ) */
  676. /* > 1<=i< j */
  677. /* > */
  678. /* > Since |x(j)| <= M(j), we use the Level 2 BLAS routine STBSV if the */
  679. /* > reciprocal of the largest M(j), j=1,..,n, is larger than */
  680. /* > f2cmax(underflow, 1/overflow). */
  681. /* > */
  682. /* > The bound on x(j) is also used to determine when a step in the */
  683. /* > columnwise method can be performed without fear of overflow. If */
  684. /* > the computed bound is greater than a large constant, x is scaled to */
  685. /* > prevent overflow, but if the bound overflows, x is set to 0, x(j) to */
  686. /* > 1, and scale to 0, and a non-trivial solution to A*x = 0 is found. */
  687. /* > */
  688. /* > Similarly, a row-wise scheme is used to solve A**T*x = b. The basic */
  689. /* > algorithm for A upper triangular is */
  690. /* > */
  691. /* > for j = 1, ..., n */
  692. /* > x(j) := ( b(j) - A[1:j-1,j]**T * x[1:j-1] ) / A(j,j) */
  693. /* > end */
  694. /* > */
  695. /* > We simultaneously compute two bounds */
  696. /* > G(j) = bound on ( b(i) - A[1:i-1,i]**T * x[1:i-1] ), 1<=i<=j */
  697. /* > M(j) = bound on x(i), 1<=i<=j */
  698. /* > */
  699. /* > The initial values are G(0) = 0, M(0) = f2cmax{b(i), i=1,..,n}, and we */
  700. /* > add the constraint G(j) >= G(j-1) and M(j) >= M(j-1) for j >= 1. */
  701. /* > Then the bound on x(j) is */
  702. /* > */
  703. /* > M(j) <= M(j-1) * ( 1 + CNORM(j) ) / | A(j,j) | */
  704. /* > */
  705. /* > <= M(0) * product ( ( 1 + CNORM(i) ) / |A(i,i)| ) */
  706. /* > 1<=i<=j */
  707. /* > */
  708. /* > and we can safely call STBSV if 1/M(n) and 1/G(n) are both greater */
  709. /* > than f2cmax(underflow, 1/overflow). */
  710. /* > \endverbatim */
  711. /* > */
  712. /* ===================================================================== */
  713. /* Subroutine */ int slatbs_(char *uplo, char *trans, char *diag, char *
  714. normin, integer *n, integer *kd, real *ab, integer *ldab, real *x,
  715. real *scale, real *cnorm, integer *info)
  716. {
  717. /* System generated locals */
  718. integer ab_dim1, ab_offset, i__1, i__2, i__3, i__4;
  719. real r__1, r__2, r__3;
  720. /* Local variables */
  721. integer jinc, jlen;
  722. real xbnd;
  723. integer imax;
  724. real tmax, tjjs;
  725. extern real sdot_(integer *, real *, integer *, real *, integer *);
  726. real xmax, grow, sumj;
  727. integer i__, j, maind;
  728. extern logical lsame_(char *, char *);
  729. extern /* Subroutine */ int sscal_(integer *, real *, real *, integer *);
  730. real tscal, uscal;
  731. integer jlast;
  732. extern real sasum_(integer *, real *, integer *);
  733. logical upper;
  734. extern /* Subroutine */ int stbsv_(char *, char *, char *, integer *,
  735. integer *, real *, integer *, real *, integer *), saxpy_(integer *, real *, real *, integer *, real *,
  736. integer *);
  737. real xj;
  738. extern real slamch_(char *);
  739. extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
  740. real bignum;
  741. extern integer isamax_(integer *, real *, integer *);
  742. logical notran;
  743. integer jfirst;
  744. real smlnum;
  745. logical nounit;
  746. real rec, tjj;
  747. /* -- LAPACK auxiliary routine (version 3.7.0) -- */
  748. /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
  749. /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
  750. /* December 2016 */
  751. /* ===================================================================== */
  752. /* Parameter adjustments */
  753. ab_dim1 = *ldab;
  754. ab_offset = 1 + ab_dim1 * 1;
  755. ab -= ab_offset;
  756. --x;
  757. --cnorm;
  758. /* Function Body */
  759. *info = 0;
  760. upper = lsame_(uplo, "U");
  761. notran = lsame_(trans, "N");
  762. nounit = lsame_(diag, "N");
  763. /* Test the input parameters. */
  764. if (! upper && ! lsame_(uplo, "L")) {
  765. *info = -1;
  766. } else if (! notran && ! lsame_(trans, "T") && !
  767. lsame_(trans, "C")) {
  768. *info = -2;
  769. } else if (! nounit && ! lsame_(diag, "U")) {
  770. *info = -3;
  771. } else if (! lsame_(normin, "Y") && ! lsame_(normin,
  772. "N")) {
  773. *info = -4;
  774. } else if (*n < 0) {
  775. *info = -5;
  776. } else if (*kd < 0) {
  777. *info = -6;
  778. } else if (*ldab < *kd + 1) {
  779. *info = -8;
  780. }
  781. if (*info != 0) {
  782. i__1 = -(*info);
  783. xerbla_("SLATBS", &i__1, (ftnlen)6);
  784. return 0;
  785. }
  786. /* Quick return if possible */
  787. if (*n == 0) {
  788. return 0;
  789. }
  790. /* Determine machine dependent parameters to control overflow. */
  791. smlnum = slamch_("Safe minimum") / slamch_("Precision");
  792. bignum = 1.f / smlnum;
  793. *scale = 1.f;
  794. if (lsame_(normin, "N")) {
  795. /* Compute the 1-norm of each column, not including the diagonal. */
  796. if (upper) {
  797. /* A is upper triangular. */
  798. i__1 = *n;
  799. for (j = 1; j <= i__1; ++j) {
  800. /* Computing MIN */
  801. i__2 = *kd, i__3 = j - 1;
  802. jlen = f2cmin(i__2,i__3);
  803. cnorm[j] = sasum_(&jlen, &ab[*kd + 1 - jlen + j * ab_dim1], &
  804. c__1);
  805. /* L10: */
  806. }
  807. } else {
  808. /* A is lower triangular. */
  809. i__1 = *n;
  810. for (j = 1; j <= i__1; ++j) {
  811. /* Computing MIN */
  812. i__2 = *kd, i__3 = *n - j;
  813. jlen = f2cmin(i__2,i__3);
  814. if (jlen > 0) {
  815. cnorm[j] = sasum_(&jlen, &ab[j * ab_dim1 + 2], &c__1);
  816. } else {
  817. cnorm[j] = 0.f;
  818. }
  819. /* L20: */
  820. }
  821. }
  822. }
  823. /* Scale the column norms by TSCAL if the maximum element in CNORM is */
  824. /* greater than BIGNUM. */
  825. imax = isamax_(n, &cnorm[1], &c__1);
  826. tmax = cnorm[imax];
  827. if (tmax <= bignum) {
  828. tscal = 1.f;
  829. } else {
  830. tscal = 1.f / (smlnum * tmax);
  831. sscal_(n, &tscal, &cnorm[1], &c__1);
  832. }
  833. /* Compute a bound on the computed solution vector to see if the */
  834. /* Level 2 BLAS routine STBSV can be used. */
  835. j = isamax_(n, &x[1], &c__1);
  836. xmax = (r__1 = x[j], abs(r__1));
  837. xbnd = xmax;
  838. if (notran) {
  839. /* Compute the growth in A * x = b. */
  840. if (upper) {
  841. jfirst = *n;
  842. jlast = 1;
  843. jinc = -1;
  844. maind = *kd + 1;
  845. } else {
  846. jfirst = 1;
  847. jlast = *n;
  848. jinc = 1;
  849. maind = 1;
  850. }
  851. if (tscal != 1.f) {
  852. grow = 0.f;
  853. goto L50;
  854. }
  855. if (nounit) {
  856. /* A is non-unit triangular. */
  857. /* Compute GROW = 1/G(j) and XBND = 1/M(j). */
  858. /* Initially, G(0) = f2cmax{x(i), i=1,...,n}. */
  859. grow = 1.f / f2cmax(xbnd,smlnum);
  860. xbnd = grow;
  861. i__1 = jlast;
  862. i__2 = jinc;
  863. for (j = jfirst; i__2 < 0 ? j >= i__1 : j <= i__1; j += i__2) {
  864. /* Exit the loop if the growth factor is too small. */
  865. if (grow <= smlnum) {
  866. goto L50;
  867. }
  868. /* M(j) = G(j-1) / abs(A(j,j)) */
  869. tjj = (r__1 = ab[maind + j * ab_dim1], abs(r__1));
  870. /* Computing MIN */
  871. r__1 = xbnd, r__2 = f2cmin(1.f,tjj) * grow;
  872. xbnd = f2cmin(r__1,r__2);
  873. if (tjj + cnorm[j] >= smlnum) {
  874. /* G(j) = G(j-1)*( 1 + CNORM(j) / abs(A(j,j)) ) */
  875. grow *= tjj / (tjj + cnorm[j]);
  876. } else {
  877. /* G(j) could overflow, set GROW to 0. */
  878. grow = 0.f;
  879. }
  880. /* L30: */
  881. }
  882. grow = xbnd;
  883. } else {
  884. /* A is unit triangular. */
  885. /* Compute GROW = 1/G(j), where G(0) = f2cmax{x(i), i=1,...,n}. */
  886. /* Computing MIN */
  887. r__1 = 1.f, r__2 = 1.f / f2cmax(xbnd,smlnum);
  888. grow = f2cmin(r__1,r__2);
  889. i__2 = jlast;
  890. i__1 = jinc;
  891. for (j = jfirst; i__1 < 0 ? j >= i__2 : j <= i__2; j += i__1) {
  892. /* Exit the loop if the growth factor is too small. */
  893. if (grow <= smlnum) {
  894. goto L50;
  895. }
  896. /* G(j) = G(j-1)*( 1 + CNORM(j) ) */
  897. grow *= 1.f / (cnorm[j] + 1.f);
  898. /* L40: */
  899. }
  900. }
  901. L50:
  902. ;
  903. } else {
  904. /* Compute the growth in A**T * x = b. */
  905. if (upper) {
  906. jfirst = 1;
  907. jlast = *n;
  908. jinc = 1;
  909. maind = *kd + 1;
  910. } else {
  911. jfirst = *n;
  912. jlast = 1;
  913. jinc = -1;
  914. maind = 1;
  915. }
  916. if (tscal != 1.f) {
  917. grow = 0.f;
  918. goto L80;
  919. }
  920. if (nounit) {
  921. /* A is non-unit triangular. */
  922. /* Compute GROW = 1/G(j) and XBND = 1/M(j). */
  923. /* Initially, M(0) = f2cmax{x(i), i=1,...,n}. */
  924. grow = 1.f / f2cmax(xbnd,smlnum);
  925. xbnd = grow;
  926. i__1 = jlast;
  927. i__2 = jinc;
  928. for (j = jfirst; i__2 < 0 ? j >= i__1 : j <= i__1; j += i__2) {
  929. /* Exit the loop if the growth factor is too small. */
  930. if (grow <= smlnum) {
  931. goto L80;
  932. }
  933. /* G(j) = f2cmax( G(j-1), M(j-1)*( 1 + CNORM(j) ) ) */
  934. xj = cnorm[j] + 1.f;
  935. /* Computing MIN */
  936. r__1 = grow, r__2 = xbnd / xj;
  937. grow = f2cmin(r__1,r__2);
  938. /* M(j) = M(j-1)*( 1 + CNORM(j) ) / abs(A(j,j)) */
  939. tjj = (r__1 = ab[maind + j * ab_dim1], abs(r__1));
  940. if (xj > tjj) {
  941. xbnd *= tjj / xj;
  942. }
  943. /* L60: */
  944. }
  945. grow = f2cmin(grow,xbnd);
  946. } else {
  947. /* A is unit triangular. */
  948. /* Compute GROW = 1/G(j), where G(0) = f2cmax{x(i), i=1,...,n}. */
  949. /* Computing MIN */
  950. r__1 = 1.f, r__2 = 1.f / f2cmax(xbnd,smlnum);
  951. grow = f2cmin(r__1,r__2);
  952. i__2 = jlast;
  953. i__1 = jinc;
  954. for (j = jfirst; i__1 < 0 ? j >= i__2 : j <= i__2; j += i__1) {
  955. /* Exit the loop if the growth factor is too small. */
  956. if (grow <= smlnum) {
  957. goto L80;
  958. }
  959. /* G(j) = ( 1 + CNORM(j) )*G(j-1) */
  960. xj = cnorm[j] + 1.f;
  961. grow /= xj;
  962. /* L70: */
  963. }
  964. }
  965. L80:
  966. ;
  967. }
  968. if (grow * tscal > smlnum) {
  969. /* Use the Level 2 BLAS solve if the reciprocal of the bound on */
  970. /* elements of X is not too small. */
  971. stbsv_(uplo, trans, diag, n, kd, &ab[ab_offset], ldab, &x[1], &c__1);
  972. } else {
  973. /* Use a Level 1 BLAS solve, scaling intermediate results. */
  974. if (xmax > bignum) {
  975. /* Scale X so that its components are less than or equal to */
  976. /* BIGNUM in absolute value. */
  977. *scale = bignum / xmax;
  978. sscal_(n, scale, &x[1], &c__1);
  979. xmax = bignum;
  980. }
  981. if (notran) {
  982. /* Solve A * x = b */
  983. i__1 = jlast;
  984. i__2 = jinc;
  985. for (j = jfirst; i__2 < 0 ? j >= i__1 : j <= i__1; j += i__2) {
  986. /* Compute x(j) = b(j) / A(j,j), scaling x if necessary. */
  987. xj = (r__1 = x[j], abs(r__1));
  988. if (nounit) {
  989. tjjs = ab[maind + j * ab_dim1] * tscal;
  990. } else {
  991. tjjs = tscal;
  992. if (tscal == 1.f) {
  993. goto L95;
  994. }
  995. }
  996. tjj = abs(tjjs);
  997. if (tjj > smlnum) {
  998. /* abs(A(j,j)) > SMLNUM: */
  999. if (tjj < 1.f) {
  1000. if (xj > tjj * bignum) {
  1001. /* Scale x by 1/b(j). */
  1002. rec = 1.f / xj;
  1003. sscal_(n, &rec, &x[1], &c__1);
  1004. *scale *= rec;
  1005. xmax *= rec;
  1006. }
  1007. }
  1008. x[j] /= tjjs;
  1009. xj = (r__1 = x[j], abs(r__1));
  1010. } else if (tjj > 0.f) {
  1011. /* 0 < abs(A(j,j)) <= SMLNUM: */
  1012. if (xj > tjj * bignum) {
  1013. /* Scale x by (1/abs(x(j)))*abs(A(j,j))*BIGNUM */
  1014. /* to avoid overflow when dividing by A(j,j). */
  1015. rec = tjj * bignum / xj;
  1016. if (cnorm[j] > 1.f) {
  1017. /* Scale by 1/CNORM(j) to avoid overflow when */
  1018. /* multiplying x(j) times column j. */
  1019. rec /= cnorm[j];
  1020. }
  1021. sscal_(n, &rec, &x[1], &c__1);
  1022. *scale *= rec;
  1023. xmax *= rec;
  1024. }
  1025. x[j] /= tjjs;
  1026. xj = (r__1 = x[j], abs(r__1));
  1027. } else {
  1028. /* A(j,j) = 0: Set x(1:n) = 0, x(j) = 1, and */
  1029. /* scale = 0, and compute a solution to A*x = 0. */
  1030. i__3 = *n;
  1031. for (i__ = 1; i__ <= i__3; ++i__) {
  1032. x[i__] = 0.f;
  1033. /* L90: */
  1034. }
  1035. x[j] = 1.f;
  1036. xj = 1.f;
  1037. *scale = 0.f;
  1038. xmax = 0.f;
  1039. }
  1040. L95:
  1041. /* Scale x if necessary to avoid overflow when adding a */
  1042. /* multiple of column j of A. */
  1043. if (xj > 1.f) {
  1044. rec = 1.f / xj;
  1045. if (cnorm[j] > (bignum - xmax) * rec) {
  1046. /* Scale x by 1/(2*abs(x(j))). */
  1047. rec *= .5f;
  1048. sscal_(n, &rec, &x[1], &c__1);
  1049. *scale *= rec;
  1050. }
  1051. } else if (xj * cnorm[j] > bignum - xmax) {
  1052. /* Scale x by 1/2. */
  1053. sscal_(n, &c_b36, &x[1], &c__1);
  1054. *scale *= .5f;
  1055. }
  1056. if (upper) {
  1057. if (j > 1) {
  1058. /* Compute the update */
  1059. /* x(f2cmax(1,j-kd):j-1) := x(f2cmax(1,j-kd):j-1) - */
  1060. /* x(j)* A(f2cmax(1,j-kd):j-1,j) */
  1061. /* Computing MIN */
  1062. i__3 = *kd, i__4 = j - 1;
  1063. jlen = f2cmin(i__3,i__4);
  1064. r__1 = -x[j] * tscal;
  1065. saxpy_(&jlen, &r__1, &ab[*kd + 1 - jlen + j * ab_dim1]
  1066. , &c__1, &x[j - jlen], &c__1);
  1067. i__3 = j - 1;
  1068. i__ = isamax_(&i__3, &x[1], &c__1);
  1069. xmax = (r__1 = x[i__], abs(r__1));
  1070. }
  1071. } else if (j < *n) {
  1072. /* Compute the update */
  1073. /* x(j+1:f2cmin(j+kd,n)) := x(j+1:f2cmin(j+kd,n)) - */
  1074. /* x(j) * A(j+1:f2cmin(j+kd,n),j) */
  1075. /* Computing MIN */
  1076. i__3 = *kd, i__4 = *n - j;
  1077. jlen = f2cmin(i__3,i__4);
  1078. if (jlen > 0) {
  1079. r__1 = -x[j] * tscal;
  1080. saxpy_(&jlen, &r__1, &ab[j * ab_dim1 + 2], &c__1, &x[
  1081. j + 1], &c__1);
  1082. }
  1083. i__3 = *n - j;
  1084. i__ = j + isamax_(&i__3, &x[j + 1], &c__1);
  1085. xmax = (r__1 = x[i__], abs(r__1));
  1086. }
  1087. /* L100: */
  1088. }
  1089. } else {
  1090. /* Solve A**T * x = b */
  1091. i__2 = jlast;
  1092. i__1 = jinc;
  1093. for (j = jfirst; i__1 < 0 ? j >= i__2 : j <= i__2; j += i__1) {
  1094. /* Compute x(j) = b(j) - sum A(k,j)*x(k). */
  1095. /* k<>j */
  1096. xj = (r__1 = x[j], abs(r__1));
  1097. uscal = tscal;
  1098. rec = 1.f / f2cmax(xmax,1.f);
  1099. if (cnorm[j] > (bignum - xj) * rec) {
  1100. /* If x(j) could overflow, scale x by 1/(2*XMAX). */
  1101. rec *= .5f;
  1102. if (nounit) {
  1103. tjjs = ab[maind + j * ab_dim1] * tscal;
  1104. } else {
  1105. tjjs = tscal;
  1106. }
  1107. tjj = abs(tjjs);
  1108. if (tjj > 1.f) {
  1109. /* Divide by A(j,j) when scaling x if A(j,j) > 1. */
  1110. /* Computing MIN */
  1111. r__1 = 1.f, r__2 = rec * tjj;
  1112. rec = f2cmin(r__1,r__2);
  1113. uscal /= tjjs;
  1114. }
  1115. if (rec < 1.f) {
  1116. sscal_(n, &rec, &x[1], &c__1);
  1117. *scale *= rec;
  1118. xmax *= rec;
  1119. }
  1120. }
  1121. sumj = 0.f;
  1122. if (uscal == 1.f) {
  1123. /* If the scaling needed for A in the dot product is 1, */
  1124. /* call SDOT to perform the dot product. */
  1125. if (upper) {
  1126. /* Computing MIN */
  1127. i__3 = *kd, i__4 = j - 1;
  1128. jlen = f2cmin(i__3,i__4);
  1129. sumj = sdot_(&jlen, &ab[*kd + 1 - jlen + j * ab_dim1],
  1130. &c__1, &x[j - jlen], &c__1);
  1131. } else {
  1132. /* Computing MIN */
  1133. i__3 = *kd, i__4 = *n - j;
  1134. jlen = f2cmin(i__3,i__4);
  1135. if (jlen > 0) {
  1136. sumj = sdot_(&jlen, &ab[j * ab_dim1 + 2], &c__1, &
  1137. x[j + 1], &c__1);
  1138. }
  1139. }
  1140. } else {
  1141. /* Otherwise, use in-line code for the dot product. */
  1142. if (upper) {
  1143. /* Computing MIN */
  1144. i__3 = *kd, i__4 = j - 1;
  1145. jlen = f2cmin(i__3,i__4);
  1146. i__3 = jlen;
  1147. for (i__ = 1; i__ <= i__3; ++i__) {
  1148. sumj += ab[*kd + i__ - jlen + j * ab_dim1] *
  1149. uscal * x[j - jlen - 1 + i__];
  1150. /* L110: */
  1151. }
  1152. } else {
  1153. /* Computing MIN */
  1154. i__3 = *kd, i__4 = *n - j;
  1155. jlen = f2cmin(i__3,i__4);
  1156. i__3 = jlen;
  1157. for (i__ = 1; i__ <= i__3; ++i__) {
  1158. sumj += ab[i__ + 1 + j * ab_dim1] * uscal * x[j +
  1159. i__];
  1160. /* L120: */
  1161. }
  1162. }
  1163. }
  1164. if (uscal == tscal) {
  1165. /* Compute x(j) := ( x(j) - sumj ) / A(j,j) if 1/A(j,j) */
  1166. /* was not used to scale the dotproduct. */
  1167. x[j] -= sumj;
  1168. xj = (r__1 = x[j], abs(r__1));
  1169. if (nounit) {
  1170. /* Compute x(j) = x(j) / A(j,j), scaling if necessary. */
  1171. tjjs = ab[maind + j * ab_dim1] * tscal;
  1172. } else {
  1173. tjjs = tscal;
  1174. if (tscal == 1.f) {
  1175. goto L135;
  1176. }
  1177. }
  1178. tjj = abs(tjjs);
  1179. if (tjj > smlnum) {
  1180. /* abs(A(j,j)) > SMLNUM: */
  1181. if (tjj < 1.f) {
  1182. if (xj > tjj * bignum) {
  1183. /* Scale X by 1/abs(x(j)). */
  1184. rec = 1.f / xj;
  1185. sscal_(n, &rec, &x[1], &c__1);
  1186. *scale *= rec;
  1187. xmax *= rec;
  1188. }
  1189. }
  1190. x[j] /= tjjs;
  1191. } else if (tjj > 0.f) {
  1192. /* 0 < abs(A(j,j)) <= SMLNUM: */
  1193. if (xj > tjj * bignum) {
  1194. /* Scale x by (1/abs(x(j)))*abs(A(j,j))*BIGNUM. */
  1195. rec = tjj * bignum / xj;
  1196. sscal_(n, &rec, &x[1], &c__1);
  1197. *scale *= rec;
  1198. xmax *= rec;
  1199. }
  1200. x[j] /= tjjs;
  1201. } else {
  1202. /* A(j,j) = 0: Set x(1:n) = 0, x(j) = 1, and */
  1203. /* scale = 0, and compute a solution to A**T*x = 0. */
  1204. i__3 = *n;
  1205. for (i__ = 1; i__ <= i__3; ++i__) {
  1206. x[i__] = 0.f;
  1207. /* L130: */
  1208. }
  1209. x[j] = 1.f;
  1210. *scale = 0.f;
  1211. xmax = 0.f;
  1212. }
  1213. L135:
  1214. ;
  1215. } else {
  1216. /* Compute x(j) := x(j) / A(j,j) - sumj if the dot */
  1217. /* product has already been divided by 1/A(j,j). */
  1218. x[j] = x[j] / tjjs - sumj;
  1219. }
  1220. /* Computing MAX */
  1221. r__2 = xmax, r__3 = (r__1 = x[j], abs(r__1));
  1222. xmax = f2cmax(r__2,r__3);
  1223. /* L140: */
  1224. }
  1225. }
  1226. *scale /= tscal;
  1227. }
  1228. /* Scale the column norms by 1/TSCAL for return. */
  1229. if (tscal != 1.f) {
  1230. r__1 = 1.f / tscal;
  1231. sscal_(n, &r__1, &cnorm[1], &c__1);
  1232. }
  1233. return 0;
  1234. /* End of SLATBS */
  1235. } /* slatbs_ */