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]/Cd(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 doublecomplex c_b1 = {1.,0.};
516 static doublecomplex c_b2 = {0.,0.};
517 static integer c__0 = 0;
518 static integer c__1 = 1;
520 /* > \brief \b ZUNGTSQR_ROW */
522 /* =========== DOCUMENTATION =========== */
524 /* Online html documentation available at */
525 /* http://www.netlib.org/lapack/explore-html/ */
528 /* > Download ZUNGTSQR_ROW + dependencies */
529 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/zunrgts
532 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/zunrgts
535 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/zunrgts
543 /* SUBROUTINE ZUNGTSQR_ROW( M, N, MB, NB, A, LDA, T, LDT, WORK, */
544 /* $ LWORK, INFO ) */
547 /* INTEGER INFO, LDA, LDT, LWORK, M, N, MB, NB */
548 /* COMPLEX*16 A( LDA, * ), T( LDT, * ), WORK( * ) */
550 /* > \par Purpose: */
555 /* > ZUNGTSQR_ROW generates an M-by-N complex matrix Q_out with */
556 /* > orthonormal columns from the output of ZLATSQR. These N orthonormal */
557 /* > columns are the first N columns of a product of complex unitary */
558 /* > matrices Q(k)_in of order M, which are returned by ZLATSQR in */
559 /* > a special format. */
561 /* > Q_out = first_N_columns_of( Q(1)_in * Q(2)_in * ... * Q(k)_in ). */
563 /* > The input matrices Q(k)_in are stored in row and column blocks in A. */
564 /* > See the documentation of ZLATSQR for more details on the format of */
565 /* > Q(k)_in, where each Q(k)_in is represented by block Householder */
566 /* > transformations. This routine calls an auxiliary routine ZLARFB_GETT, */
567 /* > where the computation is performed on each individual block. The */
568 /* > algorithm first sweeps NB-sized column blocks from the right to left */
569 /* > starting in the bottom row block and continues to the top row block */
570 /* > (hence _ROW in the routine name). This sweep is in reverse order of */
571 /* > the order in which ZLATSQR generates the output blocks. */
580 /* > The number of rows of the matrix A. M >= 0. */
586 /* > The number of columns of the matrix A. M >= N >= 0. */
589 /* > \param[in] MB */
591 /* > MB is INTEGER */
592 /* > The row block size used by ZLATSQR to return */
593 /* > arrays A and T. MB > N. */
594 /* > (Note that if MB > M, then M is used instead of MB */
595 /* > as the row block size). */
598 /* > \param[in] NB */
600 /* > NB is INTEGER */
601 /* > The column block size used by ZLATSQR to return */
602 /* > arrays A and T. NB >= 1. */
603 /* > (Note that if NB > N, then N is used instead of NB */
604 /* > as the column block size). */
607 /* > \param[in,out] A */
609 /* > A is COMPLEX*16 array, dimension (LDA,N) */
613 /* > The elements on and above the diagonal are not used as */
614 /* > input. The elements below the diagonal represent the unit */
615 /* > lower-trapezoidal blocked matrix V computed by ZLATSQR */
616 /* > that defines the input matrices Q_in(k) (ones on the */
617 /* > diagonal are not stored). See ZLATSQR for more details. */
621 /* > The array A contains an M-by-N orthonormal matrix Q_out, */
622 /* > i.e the columns of A are orthogonal unit vectors. */
625 /* > \param[in] LDA */
627 /* > LDA is INTEGER */
628 /* > The leading dimension of the array A. LDA >= f2cmax(1,M). */
633 /* > T is COMPLEX*16 array, */
634 /* > dimension (LDT, N * NIRB) */
635 /* > where NIRB = Number_of_input_row_blocks */
636 /* > = MAX( 1, CEIL((M-N)/(MB-N)) ) */
637 /* > Let NICB = Number_of_input_col_blocks */
640 /* > The upper-triangular block reflectors used to define the */
641 /* > input matrices Q_in(k), k=(1:NIRB*NICB). The block */
642 /* > reflectors are stored in compact form in NIRB block */
643 /* > reflector sequences. Each of the NIRB block reflector */
644 /* > sequences is stored in a larger NB-by-N column block of T */
645 /* > and consists of NICB smaller NB-by-NB upper-triangular */
646 /* > column blocks. See ZLATSQR for more details on the format */
650 /* > \param[in] LDT */
652 /* > LDT is INTEGER */
653 /* > The leading dimension of the array T. */
654 /* > LDT >= f2cmax(1,f2cmin(NB,N)). */
657 /* > \param[out] WORK */
659 /* > (workspace) COMPLEX*16 array, dimension (MAX(1,LWORK)) */
660 /* > On exit, if INFO = 0, WORK(1) returns the optimal LWORK. */
663 /* > \param[in] LWORK */
665 /* > The dimension of the array WORK. */
666 /* > LWORK >= NBLOCAL * MAX(NBLOCAL,(N-NBLOCAL)), */
667 /* > where NBLOCAL=MIN(NB,N). */
668 /* > If LWORK = -1, then a workspace query is assumed. */
669 /* > The routine only calculates the optimal size of the WORK */
670 /* > array, returns this value as the first entry of the WORK */
671 /* > array, and no error message related to LWORK is issued */
675 /* > \param[out] INFO */
677 /* > INFO is INTEGER */
678 /* > = 0: successful exit */
679 /* > < 0: if INFO = -i, the i-th argument had an illegal value */
685 /* > \author Univ. of Tennessee */
686 /* > \author Univ. of California Berkeley */
687 /* > \author Univ. of Colorado Denver */
688 /* > \author NAG Ltd. */
690 /* > \ingroup complex16OTHERcomputational */
692 /* > \par Contributors: */
693 /* ================== */
697 /* > November 2020, Igor Kozachenko, */
698 /* > Computer Science Division, */
699 /* > University of California, Berkeley */
703 /* ===================================================================== */
704 /* Subroutine */ int zungtsqr_row_(integer *m, integer *n, integer *mb,
705 integer *nb, doublecomplex *a, integer *lda, doublecomplex *t,
706 integer *ldt, doublecomplex *work, integer *lwork, integer *info)
708 /* System generated locals */
709 integer a_dim1, a_offset, t_dim1, t_offset, i__1, i__2, i__3, i__4, i__5;
712 /* Local variables */
713 integer jb_t__, itmp, lworkopt;
714 doublecomplex dummy[1] /* was [1][1] */;
715 integer ib_bottom__, ib, kb;
716 extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
718 extern /* Subroutine */ int zlaset_(char *, integer *, integer *,
719 doublecomplex *, doublecomplex *, doublecomplex *, integer *);
720 integer m_plus_one__;
722 integer num_all_row_blocks__, imb, knb, nblocal, kb_last__;
723 extern /* Subroutine */ int zlarfb_gett_(char *, integer *, integer *,
724 integer *, doublecomplex *, integer *, doublecomplex *, integer *,
725 doublecomplex *, integer *, doublecomplex *, integer *);
728 /* -- LAPACK computational routine -- */
729 /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
730 /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
733 /* ===================================================================== */
736 /* Test the input parameters */
738 /* Parameter adjustments */
740 a_offset = 1 + a_dim1 * 1;
743 t_offset = 1 + t_dim1 * 1;
749 lquery = *lwork == -1;
752 } else if (*n < 0 || *m < *n) {
754 } else if (*mb <= *n) {
756 } else if (*nb < 1) {
758 } else if (*lda < f2cmax(1,*m)) {
760 } else /* if(complicated condition) */ {
762 i__1 = 1, i__2 = f2cmin(*nb,*n);
763 if (*ldt < f2cmax(i__1,i__2)) {
765 } else if (*lwork < 1 && ! lquery) {
770 nblocal = f2cmin(*nb,*n);
772 /* Determine the workspace size. */
776 i__1 = nblocal, i__2 = *n - nblocal;
777 lworkopt = nblocal * f2cmax(i__1,i__2);
780 /* Handle error in the input parameters and handle the workspace query. */
784 xerbla_("ZUNGTSQR_ROW", &i__1, (ftnlen)12);
787 z__1.r = (doublereal) lworkopt, z__1.i = 0.;
788 work[1].r = z__1.r, work[1].i = z__1.i;
792 /* Quick return if possible */
794 if (f2cmin(*m,*n) == 0) {
795 z__1.r = (doublereal) lworkopt, z__1.i = 0.;
796 work[1].r = z__1.r, work[1].i = z__1.i;
800 /* (0) Set the upper-triangular part of the matrix A to zero and */
801 /* its diagonal elements to one. */
803 zlaset_("U", m, n, &c_b2, &c_b1, &a[a_offset], lda);
805 /* KB_LAST is the column index of the last column block reflector */
806 /* in the matrices T and V. */
808 kb_last__ = (*n - 1) / nblocal * nblocal + 1;
811 /* (1) Bottom-up loop over row blocks of A, except the top row block. */
812 /* NOTE: If MB>=M, then the loop is never executed. */
816 /* MB2 is the row blocking size for the row blocks before the */
817 /* first top row block in the matrix A. IB is the row index for */
818 /* the row blocks in the matrix A before the first top row block. */
819 /* IB_BOTTOM is the row index for the last bottom row block */
820 /* in the matrix A. JB_T is the column index of the corresponding */
821 /* column block in the matrix T. */
823 /* Initialize variables. */
825 /* NUM_ALL_ROW_BLOCKS is the number of row blocks in the matrix A */
826 /* including the first row block. */
829 m_plus_one__ = *m + 1;
830 itmp = (*m - *mb - 1) / mb2;
831 ib_bottom__ = itmp * mb2 + *mb + 1;
832 num_all_row_blocks__ = itmp + 2;
833 jb_t__ = num_all_row_blocks__ * *n + 1;
837 for (ib = ib_bottom__; i__2 < 0 ? ib >= i__1 : ib <= i__1; ib += i__2)
840 /* Determine the block size IMB for the current row block */
841 /* in the matrix A. */
844 i__3 = m_plus_one__ - ib;
845 imb = f2cmin(i__3,mb2);
847 /* Determine the column index JB_T for the current column block */
848 /* in the matrix T. */
852 /* Apply column blocks of H in the row block from right to left. */
854 /* KB is the column index of the current column block reflector */
855 /* in the matrices T and V. */
858 for (kb = kb_last__; i__3 < 0 ? kb >= 1 : kb <= 1; kb += i__3) {
860 /* Determine the size of the current column block KNB in */
861 /* the matrices T and V. */
864 i__4 = nblocal, i__5 = *n - kb + 1;
865 knb = f2cmin(i__4,i__5);
868 zlarfb_gett_("I", &imb, &i__4, &knb, &t[(jb_t__ + kb - 1) *
869 t_dim1 + 1], ldt, &a[kb + kb * a_dim1], lda, &a[ib +
870 kb * a_dim1], lda, &work[1], &knb);
878 /* (2) Top row block of A. */
879 /* NOTE: If MB>=M, then we have only one row block of A of size M */
880 /* and we work on the entire matrix A. */
882 mb1 = f2cmin(*mb,*m);
884 /* Apply column blocks of H in the top row block from right to left. */
886 /* KB is the column index of the current block reflector in */
887 /* the matrices T and V. */
890 for (kb = kb_last__; i__2 < 0 ? kb >= 1 : kb <= 1; kb += i__2) {
892 /* Determine the size of the current column block KNB in */
893 /* the matrices T and V. */
896 i__1 = nblocal, i__3 = *n - kb + 1;
897 knb = f2cmin(i__1,i__3);
899 if (mb1 - kb - knb + 1 == 0) {
901 /* In SLARFB_GETT parameters, when M=0, then the matrix B */
902 /* does not exist, hence we need to pass a dummy array */
903 /* reference DUMMY(1,1) to B with LDDUMMY=1. */
906 zlarfb_gett_("N", &c__0, &i__1, &knb, &t[kb * t_dim1 + 1], ldt, &
907 a[kb + kb * a_dim1], lda, dummy, &c__1, &work[1], &knb);
909 i__1 = mb1 - kb - knb + 1;
911 zlarfb_gett_("N", &i__1, &i__3, &knb, &t[kb * t_dim1 + 1], ldt, &
912 a[kb + kb * a_dim1], lda, &a[kb + knb + kb * a_dim1], lda,
918 z__1.r = (doublereal) lworkopt, z__1.i = 0.;
919 work[1].r = z__1.r, work[1].i = z__1.i;
922 /* End of ZUNGTSQR_ROW */
924 } /* zungtsqr_row__ */