14 typedef long long BLASLONG;
15 typedef unsigned long long BLASULONG;
17 typedef long BLASLONG;
18 typedef unsigned long BLASULONG;
22 typedef BLASLONG blasint;
24 #define blasabs(x) llabs(x)
26 #define blasabs(x) labs(x)
30 #define blasabs(x) abs(x)
33 typedef blasint integer;
35 typedef unsigned int uinteger;
36 typedef char *address;
37 typedef short int shortint;
39 typedef double doublereal;
40 typedef struct { real r, i; } complex;
41 typedef struct { doublereal r, i; } doublecomplex;
43 static inline _Fcomplex Cf(complex *z) {_Fcomplex zz={z->r , z->i}; return zz;}
44 static inline _Dcomplex Cd(doublecomplex *z) {_Dcomplex zz={z->r , z->i};return zz;}
45 static inline _Fcomplex * _pCf(complex *z) {return (_Fcomplex*)z;}
46 static inline _Dcomplex * _pCd(doublecomplex *z) {return (_Dcomplex*)z;}
48 static inline _Complex float Cf(complex *z) {return z->r + z->i*_Complex_I;}
49 static inline _Complex double Cd(doublecomplex *z) {return z->r + z->i*_Complex_I;}
50 static inline _Complex float * _pCf(complex *z) {return (_Complex float*)z;}
51 static inline _Complex double * _pCd(doublecomplex *z) {return (_Complex double*)z;}
53 #define pCf(z) (*_pCf(z))
54 #define pCd(z) (*_pCd(z))
56 typedef short int shortlogical;
57 typedef char logical1;
58 typedef char integer1;
63 /* Extern is for use with -E */
74 /*external read, write*/
83 /*internal read, write*/
113 /*rewind, backspace, endfile*/
125 ftnint *inex; /*parameters in standard's order*/
151 union Multitype { /* for multiple entry points */
162 typedef union Multitype Multitype;
164 struct Vardesc { /* for Namelist */
170 typedef struct Vardesc Vardesc;
177 typedef struct Namelist Namelist;
179 #define abs(x) ((x) >= 0 ? (x) : -(x))
180 #define dabs(x) (fabs(x))
181 #define f2cmin(a,b) ((a) <= (b) ? (a) : (b))
182 #define f2cmax(a,b) ((a) >= (b) ? (a) : (b))
183 #define dmin(a,b) (f2cmin(a,b))
184 #define dmax(a,b) (f2cmax(a,b))
185 #define bit_test(a,b) ((a) >> (b) & 1)
186 #define bit_clear(a,b) ((a) & ~((uinteger)1 << (b)))
187 #define bit_set(a,b) ((a) | ((uinteger)1 << (b)))
189 #define abort_() { sig_die("Fortran abort routine called", 1); }
190 #define c_abs(z) (cabsf(Cf(z)))
191 #define c_cos(R,Z) { pCf(R)=ccos(Cf(Z)); }
193 #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]);}
194 #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]);}
196 #define c_div(c, a, b) {pCf(c) = Cf(a)/Cf(b);}
197 #define z_div(c, a, b) {pCd(c) = Cd(a)/Cd(b);}
199 #define c_exp(R, Z) {pCf(R) = cexpf(Cf(Z));}
200 #define c_log(R, Z) {pCf(R) = clogf(Cf(Z));}
201 #define c_sin(R, Z) {pCf(R) = csinf(Cf(Z));}
202 //#define c_sqrt(R, Z) {*(R) = csqrtf(Cf(Z));}
203 #define c_sqrt(R, Z) {pCf(R) = csqrtf(Cf(Z));}
204 #define d_abs(x) (fabs(*(x)))
205 #define d_acos(x) (acos(*(x)))
206 #define d_asin(x) (asin(*(x)))
207 #define d_atan(x) (atan(*(x)))
208 #define d_atn2(x, y) (atan2(*(x),*(y)))
209 #define d_cnjg(R, Z) { pCd(R) = conj(Cd(Z)); }
210 #define r_cnjg(R, Z) { pCf(R) = conjf(Cf(Z)); }
211 #define d_cos(x) (cos(*(x)))
212 #define d_cosh(x) (cosh(*(x)))
213 #define d_dim(__a, __b) ( *(__a) > *(__b) ? *(__a) - *(__b) : 0.0 )
214 #define d_exp(x) (exp(*(x)))
215 #define d_imag(z) (cimag(Cd(z)))
216 #define r_imag(z) (cimagf(Cf(z)))
217 #define d_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
218 #define r_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
219 #define d_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
220 #define r_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
221 #define d_log(x) (log(*(x)))
222 #define d_mod(x, y) (fmod(*(x), *(y)))
223 #define u_nint(__x) ((__x)>=0 ? floor((__x) + .5) : -floor(.5 - (__x)))
224 #define d_nint(x) u_nint(*(x))
225 #define u_sign(__a,__b) ((__b) >= 0 ? ((__a) >= 0 ? (__a) : -(__a)) : -((__a) >= 0 ? (__a) : -(__a)))
226 #define d_sign(a,b) u_sign(*(a),*(b))
227 #define r_sign(a,b) u_sign(*(a),*(b))
228 #define d_sin(x) (sin(*(x)))
229 #define d_sinh(x) (sinh(*(x)))
230 #define d_sqrt(x) (sqrt(*(x)))
231 #define d_tan(x) (tan(*(x)))
232 #define d_tanh(x) (tanh(*(x)))
233 #define i_abs(x) abs(*(x))
234 #define i_dnnt(x) ((integer)u_nint(*(x)))
235 #define i_len(s, n) (n)
236 #define i_nint(x) ((integer)u_nint(*(x)))
237 #define i_sign(a,b) ((integer)u_sign((integer)*(a),(integer)*(b)))
238 #define pow_dd(ap, bp) ( pow(*(ap), *(bp)))
239 #define pow_si(B,E) spow_ui(*(B),*(E))
240 #define pow_ri(B,E) spow_ui(*(B),*(E))
241 #define pow_di(B,E) dpow_ui(*(B),*(E))
242 #define pow_zi(p, a, b) {pCd(p) = zpow_ui(Cd(a), *(b));}
243 #define pow_ci(p, a, b) {pCf(p) = cpow_ui(Cf(a), *(b));}
244 #define pow_zz(R,A,B) {pCd(R) = cpow(Cd(A),*(B));}
245 #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++ = ' '; }
246 #define s_cmp(a,b,c,d) ((integer)strncmp((a),(b),f2cmin((c),(d))))
247 #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]; }
248 #define sig_die(s, kill) { exit(1); }
249 #define s_stop(s, n) {exit(0);}
250 static char junk[] = "\n@(#)LIBF77 VERSION 19990503\n";
251 #define z_abs(z) (cabs(Cd(z)))
252 #define z_exp(R, Z) {pCd(R) = cexp(Cd(Z));}
253 #define z_sqrt(R, Z) {pCd(R) = csqrt(Cd(Z));}
254 #define myexit_() break;
255 #define mycycle() continue;
256 #define myceiling(w) {ceil(w)}
257 #define myhuge(w) {HUGE_VAL}
258 //#define mymaxloc_(w,s,e,n) {if (sizeof(*(w)) == sizeof(double)) dmaxloc_((w),*(s),*(e),n); else dmaxloc_((w),*(s),*(e),n);}
259 #define mymaxloc(w,s,e,n) {dmaxloc_(w,*(s),*(e),n)}
261 /* procedure parameter types for -A and -C++ */
263 #define F2C_proc_par_types 1
265 typedef logical (*L_fp)(...);
267 typedef logical (*L_fp)();
270 static float spow_ui(float x, integer n) {
271 float pow=1.0; unsigned long int u;
273 if(n < 0) n = -n, x = 1/x;
282 static double dpow_ui(double x, integer n) {
283 double pow=1.0; unsigned long int u;
285 if(n < 0) n = -n, x = 1/x;
295 static _Fcomplex cpow_ui(complex x, integer n) {
296 complex pow={1.0,0.0}; unsigned long int u;
298 if(n < 0) n = -n, x.r = 1/x.r, x.i=1/x.i;
300 if(u & 01) pow.r *= x.r, pow.i *= x.i;
301 if(u >>= 1) x.r *= x.r, x.i *= x.i;
305 _Fcomplex p={pow.r, pow.i};
309 static _Complex float cpow_ui(_Complex float x, integer n) {
310 _Complex float pow=1.0; unsigned long int u;
312 if(n < 0) n = -n, x = 1/x;
323 static _Dcomplex zpow_ui(_Dcomplex x, integer n) {
324 _Dcomplex pow={1.0,0.0}; unsigned long int u;
326 if(n < 0) n = -n, x._Val[0] = 1/x._Val[0], x._Val[1] =1/x._Val[1];
328 if(u & 01) pow._Val[0] *= x._Val[0], pow._Val[1] *= x._Val[1];
329 if(u >>= 1) x._Val[0] *= x._Val[0], x._Val[1] *= x._Val[1];
333 _Dcomplex p = {pow._Val[0], pow._Val[1]};
337 static _Complex double zpow_ui(_Complex double x, integer n) {
338 _Complex double pow=1.0; unsigned long int u;
340 if(n < 0) n = -n, x = 1/x;
350 static integer pow_ii(integer x, integer n) {
351 integer pow; unsigned long int u;
353 if (n == 0 || x == 1) pow = 1;
354 else if (x != -1) pow = x == 0 ? 1/x : 0;
357 if ((n > 0) || !(n == 0 || x == 1 || x != -1)) {
367 static integer dmaxloc_(double *w, integer s, integer e, integer *n)
369 double m; integer i, mi;
370 for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
371 if (w[i-1]>m) mi=i ,m=w[i-1];
374 static integer smaxloc_(float *w, integer s, integer e, integer *n)
376 float m; integer i, mi;
377 for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
378 if (w[i-1]>m) mi=i ,m=w[i-1];
381 static inline void cdotc_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
382 integer n = *n_, incx = *incx_, incy = *incy_, i;
384 _Fcomplex zdotc = {0.0, 0.0};
385 if (incx == 1 && incy == 1) {
386 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
387 zdotc._Val[0] += conjf(Cf(&x[i]))._Val[0] * Cf(&y[i])._Val[0];
388 zdotc._Val[1] += conjf(Cf(&x[i]))._Val[1] * Cf(&y[i])._Val[1];
391 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
392 zdotc._Val[0] += conjf(Cf(&x[i*incx]))._Val[0] * Cf(&y[i*incy])._Val[0];
393 zdotc._Val[1] += conjf(Cf(&x[i*incx]))._Val[1] * Cf(&y[i*incy])._Val[1];
399 _Complex float zdotc = 0.0;
400 if (incx == 1 && incy == 1) {
401 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
402 zdotc += conjf(Cf(&x[i])) * Cf(&y[i]);
405 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
406 zdotc += conjf(Cf(&x[i*incx])) * Cf(&y[i*incy]);
412 static inline void zdotc_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
413 integer n = *n_, incx = *incx_, incy = *incy_, i;
415 _Dcomplex zdotc = {0.0, 0.0};
416 if (incx == 1 && incy == 1) {
417 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
418 zdotc._Val[0] += conj(Cd(&x[i]))._Val[0] * Cd(&y[i])._Val[0];
419 zdotc._Val[1] += conj(Cd(&x[i]))._Val[1] * Cd(&y[i])._Val[1];
422 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
423 zdotc._Val[0] += conj(Cd(&x[i*incx]))._Val[0] * Cd(&y[i*incy])._Val[0];
424 zdotc._Val[1] += conj(Cd(&x[i*incx]))._Val[1] * Cd(&y[i*incy])._Val[1];
430 _Complex double zdotc = 0.0;
431 if (incx == 1 && incy == 1) {
432 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
433 zdotc += conj(Cd(&x[i])) * Cd(&y[i]);
436 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
437 zdotc += conj(Cd(&x[i*incx])) * Cd(&y[i*incy]);
443 static inline void cdotu_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
444 integer n = *n_, incx = *incx_, incy = *incy_, i;
446 _Fcomplex zdotc = {0.0, 0.0};
447 if (incx == 1 && incy == 1) {
448 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
449 zdotc._Val[0] += Cf(&x[i])._Val[0] * Cf(&y[i])._Val[0];
450 zdotc._Val[1] += Cf(&x[i])._Val[1] * Cf(&y[i])._Val[1];
453 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
454 zdotc._Val[0] += Cf(&x[i*incx])._Val[0] * Cf(&y[i*incy])._Val[0];
455 zdotc._Val[1] += Cf(&x[i*incx])._Val[1] * Cf(&y[i*incy])._Val[1];
461 _Complex float zdotc = 0.0;
462 if (incx == 1 && incy == 1) {
463 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
464 zdotc += Cf(&x[i]) * Cf(&y[i]);
467 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
468 zdotc += Cf(&x[i*incx]) * Cf(&y[i*incy]);
474 static inline void zdotu_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
475 integer n = *n_, incx = *incx_, incy = *incy_, i;
477 _Dcomplex zdotc = {0.0, 0.0};
478 if (incx == 1 && incy == 1) {
479 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
480 zdotc._Val[0] += Cd(&x[i])._Val[0] * Cd(&y[i])._Val[0];
481 zdotc._Val[1] += Cd(&x[i])._Val[1] * Cd(&y[i])._Val[1];
484 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
485 zdotc._Val[0] += Cd(&x[i*incx])._Val[0] * Cd(&y[i*incy])._Val[0];
486 zdotc._Val[1] += Cd(&x[i*incx])._Val[1] * Cd(&y[i*incy])._Val[1];
492 _Complex double zdotc = 0.0;
493 if (incx == 1 && incy == 1) {
494 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
495 zdotc += Cd(&x[i]) * Cd(&y[i]);
498 for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
499 zdotc += Cd(&x[i*incx]) * Cd(&y[i*incy]);
505 /* -- translated by f2c (version 20000121).
506 You must link the resulting object file with the libraries:
507 -lf2c -lm (in that order)
513 /* Table of constant values */
515 static integer c__1 = 1;
516 static integer c_n1 = -1;
517 static integer c__3 = 3;
518 static integer c__2 = 2;
519 static real c_b21 = -1.f;
520 static real c_b22 = 1.f;
522 /* > \brief \b SGEBRD */
524 /* =========== DOCUMENTATION =========== */
526 /* Online html documentation available at */
527 /* http://www.netlib.org/lapack/explore-html/ */
530 /* > Download SGEBRD + dependencies */
531 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/sgebrd.
534 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/sgebrd.
537 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/sgebrd.
545 /* SUBROUTINE SGEBRD( M, N, A, LDA, D, E, TAUQ, TAUP, WORK, LWORK, */
548 /* INTEGER INFO, LDA, LWORK, M, N */
549 /* REAL A( LDA, * ), D( * ), E( * ), TAUP( * ), */
550 /* $ TAUQ( * ), WORK( * ) */
553 /* > \par Purpose: */
558 /* > SGEBRD reduces a general real M-by-N matrix A to upper or lower */
559 /* > bidiagonal form B by an orthogonal transformation: Q**T * A * P = B. */
561 /* > If m >= n, B is upper bidiagonal; if m < n, B is lower bidiagonal. */
570 /* > The number of rows in the matrix A. M >= 0. */
576 /* > The number of columns in the matrix A. N >= 0. */
579 /* > \param[in,out] A */
581 /* > A is REAL array, dimension (LDA,N) */
582 /* > On entry, the M-by-N general matrix to be reduced. */
584 /* > if m >= n, the diagonal and the first superdiagonal are */
585 /* > overwritten with the upper bidiagonal matrix B; the */
586 /* > elements below the diagonal, with the array TAUQ, represent */
587 /* > the orthogonal matrix Q as a product of elementary */
588 /* > reflectors, and the elements above the first superdiagonal, */
589 /* > with the array TAUP, represent the orthogonal matrix P as */
590 /* > a product of elementary reflectors; */
591 /* > if m < n, the diagonal and the first subdiagonal are */
592 /* > overwritten with the lower bidiagonal matrix B; the */
593 /* > elements below the first subdiagonal, with the array TAUQ, */
594 /* > represent the orthogonal matrix Q as a product of */
595 /* > elementary reflectors, and the elements above the diagonal, */
596 /* > with the array TAUP, represent the orthogonal matrix P as */
597 /* > a product of elementary reflectors. */
598 /* > See Further Details. */
601 /* > \param[in] LDA */
603 /* > LDA is INTEGER */
604 /* > The leading dimension of the array A. LDA >= f2cmax(1,M). */
607 /* > \param[out] D */
609 /* > D is REAL array, dimension (f2cmin(M,N)) */
610 /* > The diagonal elements of the bidiagonal matrix B: */
611 /* > D(i) = A(i,i). */
614 /* > \param[out] E */
616 /* > E is REAL array, dimension (f2cmin(M,N)-1) */
617 /* > The off-diagonal elements of the bidiagonal matrix B: */
618 /* > if m >= n, E(i) = A(i,i+1) for i = 1,2,...,n-1; */
619 /* > if m < n, E(i) = A(i+1,i) for i = 1,2,...,m-1. */
622 /* > \param[out] TAUQ */
624 /* > TAUQ is REAL array, dimension (f2cmin(M,N)) */
625 /* > The scalar factors of the elementary reflectors which */
626 /* > represent the orthogonal matrix Q. See Further Details. */
629 /* > \param[out] TAUP */
631 /* > TAUP is REAL array, dimension (f2cmin(M,N)) */
632 /* > The scalar factors of the elementary reflectors which */
633 /* > represent the orthogonal matrix P. See Further Details. */
636 /* > \param[out] WORK */
638 /* > WORK is REAL array, dimension (MAX(1,LWORK)) */
639 /* > On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
642 /* > \param[in] LWORK */
644 /* > LWORK is INTEGER */
645 /* > The length of the array WORK. LWORK >= f2cmax(1,M,N). */
646 /* > For optimum performance LWORK >= (M+N)*NB, where NB */
647 /* > is the optimal blocksize. */
649 /* > If LWORK = -1, then a workspace query is assumed; the routine */
650 /* > only calculates the optimal size of the WORK array, returns */
651 /* > this value as the first entry of the WORK array, and no error */
652 /* > message related to LWORK is issued by XERBLA. */
655 /* > \param[out] INFO */
657 /* > INFO is INTEGER */
658 /* > = 0: successful exit */
659 /* > < 0: if INFO = -i, the i-th argument had an illegal value. */
665 /* > \author Univ. of Tennessee */
666 /* > \author Univ. of California Berkeley */
667 /* > \author Univ. of Colorado Denver */
668 /* > \author NAG Ltd. */
670 /* > \date November 2017 */
672 /* > \ingroup realGEcomputational */
674 /* > \par Further Details: */
675 /* ===================== */
679 /* > The matrices Q and P are represented as products of elementary */
684 /* > Q = H(1) H(2) . . . H(n) and P = G(1) G(2) . . . G(n-1) */
686 /* > Each H(i) and G(i) has the form: */
688 /* > H(i) = I - tauq * v * v**T and G(i) = I - taup * u * u**T */
690 /* > where tauq and taup are real scalars, and v and u are real vectors; */
691 /* > v(1:i-1) = 0, v(i) = 1, and v(i+1:m) is stored on exit in A(i+1:m,i); */
692 /* > u(1:i) = 0, u(i+1) = 1, and u(i+2:n) is stored on exit in A(i,i+2:n); */
693 /* > tauq is stored in TAUQ(i) and taup in TAUP(i). */
697 /* > Q = H(1) H(2) . . . H(m-1) and P = G(1) G(2) . . . G(m) */
699 /* > Each H(i) and G(i) has the form: */
701 /* > H(i) = I - tauq * v * v**T and G(i) = I - taup * u * u**T */
703 /* > where tauq and taup are real scalars, and v and u are real vectors; */
704 /* > v(1:i) = 0, v(i+1) = 1, and v(i+2:m) is stored on exit in A(i+2:m,i); */
705 /* > u(1:i-1) = 0, u(i) = 1, and u(i+1:n) is stored on exit in A(i,i+1:n); */
706 /* > tauq is stored in TAUQ(i) and taup in TAUP(i). */
708 /* > The contents of A on exit are illustrated by the following examples: */
710 /* > m = 6 and n = 5 (m > n): m = 5 and n = 6 (m < n): */
712 /* > ( d e u1 u1 u1 ) ( d u1 u1 u1 u1 u1 ) */
713 /* > ( v1 d e u2 u2 ) ( e d u2 u2 u2 u2 ) */
714 /* > ( v1 v2 d e u3 ) ( v1 e d u3 u3 u3 ) */
715 /* > ( v1 v2 v3 d e ) ( v1 v2 e d u4 u4 ) */
716 /* > ( v1 v2 v3 v4 d ) ( v1 v2 v3 e d u5 ) */
717 /* > ( v1 v2 v3 v4 v5 ) */
719 /* > where d and e denote diagonal and off-diagonal elements of B, vi */
720 /* > denotes an element of the vector defining H(i), and ui an element of */
721 /* > the vector defining G(i). */
724 /* ===================================================================== */
725 /* Subroutine */ int sgebrd_(integer *m, integer *n, real *a, integer *lda,
726 real *d__, real *e, real *tauq, real *taup, real *work, integer *
727 lwork, integer *info)
729 /* System generated locals */
730 integer a_dim1, a_offset, i__1, i__2, i__3, i__4;
732 /* Local variables */
733 integer i__, j, nbmin, iinfo;
734 extern /* Subroutine */ int sgemm_(char *, char *, integer *, integer *,
735 integer *, real *, real *, integer *, real *, integer *, real *,
738 extern /* Subroutine */ int sgebd2_(integer *, integer *, real *, integer
739 *, real *, real *, real *, real *, real *, integer *);
741 extern /* Subroutine */ int slabrd_(integer *, integer *, integer *, real
742 *, integer *, real *, real *, real *, real *, real *, integer *,
745 extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
746 extern integer ilaenv_(integer *, char *, char *, integer *, integer *,
747 integer *, integer *, ftnlen, ftnlen);
748 integer ldwrkx, ldwrky, lwkopt;
752 /* -- LAPACK computational routine (version 3.8.0) -- */
753 /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
754 /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
758 /* ===================================================================== */
761 /* Test the input parameters */
763 /* Parameter adjustments */
765 a_offset = 1 + a_dim1 * 1;
776 i__1 = 1, i__2 = ilaenv_(&c__1, "SGEBRD", " ", m, n, &c_n1, &c_n1, (
777 ftnlen)6, (ftnlen)1);
778 nb = f2cmax(i__1,i__2);
779 lwkopt = (*m + *n) * nb;
780 work[1] = (real) lwkopt;
781 lquery = *lwork == -1;
786 } else if (*lda < f2cmax(1,*m)) {
788 } else /* if(complicated condition) */ {
791 if (*lwork < f2cmax(i__1,*n) && ! lquery) {
797 xerbla_("SGEBRD", &i__1, (ftnlen)6);
803 /* Quick return if possible */
805 minmn = f2cmin(*m,*n);
815 if (nb > 1 && nb < minmn) {
817 /* Set the crossover point NX. */
820 i__1 = nb, i__2 = ilaenv_(&c__3, "SGEBRD", " ", m, n, &c_n1, &c_n1, (
821 ftnlen)6, (ftnlen)1);
822 nx = f2cmax(i__1,i__2);
824 /* Determine when to switch from blocked to unblocked code. */
830 /* Not enough work space for the optimal NB, consider using */
831 /* a smaller block size. */
833 nbmin = ilaenv_(&c__2, "SGEBRD", " ", m, n, &c_n1, &c_n1, (
834 ftnlen)6, (ftnlen)1);
835 if (*lwork >= (*m + *n) * nbmin) {
836 nb = *lwork / (*m + *n);
849 for (i__ = 1; i__2 < 0 ? i__ >= i__1 : i__ <= i__1; i__ += i__2) {
851 /* Reduce rows and columns i:i+nb-1 to bidiagonal form and return */
852 /* the matrices X and Y which are needed to update the unreduced */
853 /* part of the matrix */
857 slabrd_(&i__3, &i__4, &nb, &a[i__ + i__ * a_dim1], lda, &d__[i__], &e[
858 i__], &tauq[i__], &taup[i__], &work[1], &ldwrkx, &work[ldwrkx
861 /* Update the trailing submatrix A(i+nb:m,i+nb:n), using an update */
862 /* of the form A := A - V*Y**T - X*U**T */
864 i__3 = *m - i__ - nb + 1;
865 i__4 = *n - i__ - nb + 1;
866 sgemm_("No transpose", "Transpose", &i__3, &i__4, &nb, &c_b21, &a[i__
867 + nb + i__ * a_dim1], lda, &work[ldwrkx * nb + nb + 1], &
868 ldwrky, &c_b22, &a[i__ + nb + (i__ + nb) * a_dim1], lda);
869 i__3 = *m - i__ - nb + 1;
870 i__4 = *n - i__ - nb + 1;
871 sgemm_("No transpose", "No transpose", &i__3, &i__4, &nb, &c_b21, &
872 work[nb + 1], &ldwrkx, &a[i__ + (i__ + nb) * a_dim1], lda, &
873 c_b22, &a[i__ + nb + (i__ + nb) * a_dim1], lda);
875 /* Copy diagonal and off-diagonal elements of B back into A */
879 for (j = i__; j <= i__3; ++j) {
880 a[j + j * a_dim1] = d__[j];
881 a[j + (j + 1) * a_dim1] = e[j];
886 for (j = i__; j <= i__3; ++j) {
887 a[j + j * a_dim1] = d__[j];
888 a[j + 1 + j * a_dim1] = e[j];
895 /* Use unblocked code to reduce the remainder of the matrix */
899 sgebd2_(&i__2, &i__1, &a[i__ + i__ * a_dim1], lda, &d__[i__], &e[i__], &
900 tauq[i__], &taup[i__], &work[1], &iinfo);