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 integer c__1 = 1;
518 /* > \brief \b ZSPTRF */
520 /* =========== DOCUMENTATION =========== */
522 /* Online html documentation available at */
523 /* http://www.netlib.org/lapack/explore-html/ */
526 /* > Download ZSPTRF + dependencies */
527 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/zsptrf.
530 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/zsptrf.
533 /* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/zsptrf.
541 /* SUBROUTINE ZSPTRF( UPLO, N, AP, IPIV, INFO ) */
544 /* INTEGER INFO, N */
545 /* INTEGER IPIV( * ) */
546 /* COMPLEX*16 AP( * ) */
549 /* > \par Purpose: */
554 /* > ZSPTRF computes the factorization of a complex symmetric matrix A */
555 /* > stored in packed format using the Bunch-Kaufman diagonal pivoting */
558 /* > A = U*D*U**T or A = L*D*L**T */
560 /* > where U (or L) is a product of permutation and unit upper (lower) */
561 /* > triangular matrices, and D is symmetric and block diagonal with */
562 /* > 1-by-1 and 2-by-2 diagonal blocks. */
568 /* > \param[in] UPLO */
570 /* > UPLO is CHARACTER*1 */
571 /* > = 'U': Upper triangle of A is stored; */
572 /* > = 'L': Lower triangle of A is stored. */
578 /* > The order of the matrix A. N >= 0. */
581 /* > \param[in,out] AP */
583 /* > AP is COMPLEX*16 array, dimension (N*(N+1)/2) */
584 /* > On entry, the upper or lower triangle of the symmetric matrix */
585 /* > A, packed columnwise in a linear array. The j-th column of A */
586 /* > is stored in the array AP as follows: */
587 /* > if UPLO = 'U', AP(i + (j-1)*j/2) = A(i,j) for 1<=i<=j; */
588 /* > if UPLO = 'L', AP(i + (j-1)*(2n-j)/2) = A(i,j) for j<=i<=n. */
590 /* > On exit, the block diagonal matrix D and the multipliers used */
591 /* > to obtain the factor U or L, stored as a packed triangular */
592 /* > matrix overwriting A (see below for further details). */
595 /* > \param[out] IPIV */
597 /* > IPIV is INTEGER array, dimension (N) */
598 /* > Details of the interchanges and the block structure of D. */
599 /* > If IPIV(k) > 0, then rows and columns k and IPIV(k) were */
600 /* > interchanged and D(k,k) is a 1-by-1 diagonal block. */
601 /* > If UPLO = 'U' and IPIV(k) = IPIV(k-1) < 0, then rows and */
602 /* > columns k-1 and -IPIV(k) were interchanged and D(k-1:k,k-1:k) */
603 /* > is a 2-by-2 diagonal block. If UPLO = 'L' and IPIV(k) = */
604 /* > IPIV(k+1) < 0, then rows and columns k+1 and -IPIV(k) were */
605 /* > interchanged and D(k:k+1,k:k+1) is a 2-by-2 diagonal block. */
608 /* > \param[out] INFO */
610 /* > INFO is INTEGER */
611 /* > = 0: successful exit */
612 /* > < 0: if INFO = -i, the i-th argument had an illegal value */
613 /* > > 0: if INFO = i, D(i,i) is exactly zero. The factorization */
614 /* > has been completed, but the block diagonal matrix D is */
615 /* > exactly singular, and division by zero will occur if it */
616 /* > is used to solve a system of equations. */
622 /* > \author Univ. of Tennessee */
623 /* > \author Univ. of California Berkeley */
624 /* > \author Univ. of Colorado Denver */
625 /* > \author NAG Ltd. */
627 /* > \date December 2016 */
629 /* > \ingroup complex16OTHERcomputational */
631 /* > \par Further Details: */
632 /* ===================== */
636 /* > 5-96 - Based on modifications by J. Lewis, Boeing Computer Services */
639 /* > If UPLO = 'U', then A = U*D*U**T, where */
640 /* > U = P(n)*U(n)* ... *P(k)U(k)* ..., */
641 /* > i.e., U is a product of terms P(k)*U(k), where k decreases from n to */
642 /* > 1 in steps of 1 or 2, and D is a block diagonal matrix with 1-by-1 */
643 /* > and 2-by-2 diagonal blocks D(k). P(k) is a permutation matrix as */
644 /* > defined by IPIV(k), and U(k) is a unit upper triangular matrix, such */
645 /* > that if the diagonal block D(k) is of order s (s = 1 or 2), then */
647 /* > ( I v 0 ) k-s */
648 /* > U(k) = ( 0 I 0 ) s */
649 /* > ( 0 0 I ) n-k */
652 /* > If s = 1, D(k) overwrites A(k,k), and v overwrites A(1:k-1,k). */
653 /* > If s = 2, the upper triangle of D(k) overwrites A(k-1,k-1), A(k-1,k), */
654 /* > and A(k,k), and v overwrites A(1:k-2,k-1:k). */
656 /* > If UPLO = 'L', then A = L*D*L**T, where */
657 /* > L = P(1)*L(1)* ... *P(k)*L(k)* ..., */
658 /* > i.e., L is a product of terms P(k)*L(k), where k increases from 1 to */
659 /* > n in steps of 1 or 2, and D is a block diagonal matrix with 1-by-1 */
660 /* > and 2-by-2 diagonal blocks D(k). P(k) is a permutation matrix as */
661 /* > defined by IPIV(k), and L(k) is a unit lower triangular matrix, such */
662 /* > that if the diagonal block D(k) is of order s (s = 1 or 2), then */
664 /* > ( I 0 0 ) k-1 */
665 /* > L(k) = ( 0 I 0 ) s */
666 /* > ( 0 v I ) n-k-s+1 */
667 /* > k-1 s n-k-s+1 */
669 /* > If s = 1, D(k) overwrites A(k,k), and v overwrites A(k+1:n,k). */
670 /* > If s = 2, the lower triangle of D(k) overwrites A(k,k), A(k+1,k), */
671 /* > and A(k+1,k+1), and v overwrites A(k+2:n,k:k+1). */
674 /* ===================================================================== */
675 /* Subroutine */ int zsptrf_(char *uplo, integer *n, doublecomplex *ap,
676 integer *ipiv, integer *info)
678 /* System generated locals */
679 integer i__1, i__2, i__3, i__4, i__5, i__6;
680 doublereal d__1, d__2, d__3, d__4;
681 doublecomplex z__1, z__2, z__3, z__4;
683 /* Local variables */
685 extern /* Subroutine */ int zspr_(char *, integer *, doublecomplex *,
686 doublecomplex *, integer *, doublecomplex *);
690 extern logical lsame_(char *, char *);
691 extern /* Subroutine */ int zscal_(integer *, doublecomplex *,
692 doublecomplex *, integer *);
696 extern /* Subroutine */ int zswap_(integer *, doublecomplex *, integer *,
697 doublecomplex *, integer *);
698 doublecomplex d11, d12, d21, d22;
703 extern /* Subroutine */ int xerbla_(char *, integer *, ftnlen);
705 extern integer izamax_(integer *, doublecomplex *, integer *);
707 integer knc, kpc, npp;
708 doublecomplex wkm1, wkp1;
711 /* -- LAPACK computational routine (version 3.7.0) -- */
712 /* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
713 /* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
717 /* ===================================================================== */
720 /* Test the input parameters. */
722 /* Parameter adjustments */
728 upper = lsame_(uplo, "U");
729 if (! upper && ! lsame_(uplo, "L")) {
736 xerbla_("ZSPTRF", &i__1, (ftnlen)6);
740 /* Initialize ALPHA for use in choosing pivot block size. */
742 alpha = (sqrt(17.) + 1.) / 8.;
746 /* Factorize A as U*D*U**T using the upper triangle of A */
748 /* K is the main loop index, decreasing from N to 1 in steps of */
752 kc = (*n - 1) * *n / 2 + 1;
756 /* If K < 1, exit from loop */
763 /* Determine rows and columns to be interchanged and whether */
764 /* a 1-by-1 or 2-by-2 pivot block will be used */
767 absakk = (d__1 = ap[i__1].r, abs(d__1)) + (d__2 = d_imag(&ap[kc + k -
770 /* IMAX is the row-index of the largest off-diagonal element in */
771 /* column K, and COLMAX is its absolute value */
775 imax = izamax_(&i__1, &ap[kc], &c__1);
776 i__1 = kc + imax - 1;
777 colmax = (d__1 = ap[i__1].r, abs(d__1)) + (d__2 = d_imag(&ap[kc +
778 imax - 1]), abs(d__2));
783 if (f2cmax(absakk,colmax) == 0.) {
785 /* Column K is zero: set INFO and continue */
792 if (absakk >= alpha * colmax) {
794 /* no interchange, use 1-by-1 pivot block */
801 kx = imax * (imax + 1) / 2 + imax;
803 for (j = imax + 1; j <= i__1; ++j) {
805 if ((d__1 = ap[i__2].r, abs(d__1)) + (d__2 = d_imag(&ap[
806 kx]), abs(d__2)) > rowmax) {
808 rowmax = (d__1 = ap[i__2].r, abs(d__1)) + (d__2 =
809 d_imag(&ap[kx]), abs(d__2));
815 kpc = (imax - 1) * imax / 2 + 1;
818 jmax = izamax_(&i__1, &ap[kpc], &c__1);
820 i__1 = kpc + jmax - 1;
821 d__3 = rowmax, d__4 = (d__1 = ap[i__1].r, abs(d__1)) + (
822 d__2 = d_imag(&ap[kpc + jmax - 1]), abs(d__2));
823 rowmax = f2cmax(d__3,d__4);
826 if (absakk >= alpha * colmax * (colmax / rowmax)) {
828 /* no interchange, use 1-by-1 pivot block */
831 } else /* if(complicated condition) */ {
832 i__1 = kpc + imax - 1;
833 if ((d__1 = ap[i__1].r, abs(d__1)) + (d__2 = d_imag(&ap[
834 kpc + imax - 1]), abs(d__2)) >= alpha * rowmax) {
836 /* interchange rows and columns K and IMAX, use 1-by-1 */
842 /* interchange rows and columns K-1 and IMAX, use 2-by-2 */
857 /* Interchange rows and columns KK and KP in the leading */
858 /* submatrix A(1:k,1:k) */
861 zswap_(&i__1, &ap[knc], &c__1, &ap[kpc], &c__1);
864 for (j = kp + 1; j <= i__1; ++j) {
867 t.r = ap[i__2].r, t.i = ap[i__2].i;
870 ap[i__2].r = ap[i__3].r, ap[i__2].i = ap[i__3].i;
872 ap[i__2].r = t.r, ap[i__2].i = t.i;
876 t.r = ap[i__1].r, t.i = ap[i__1].i;
879 ap[i__1].r = ap[i__2].r, ap[i__1].i = ap[i__2].i;
881 ap[i__1].r = t.r, ap[i__1].i = t.i;
884 t.r = ap[i__1].r, t.i = ap[i__1].i;
887 ap[i__1].r = ap[i__2].r, ap[i__1].i = ap[i__2].i;
889 ap[i__1].r = t.r, ap[i__1].i = t.i;
893 /* Update the leading submatrix */
897 /* 1-by-1 pivot block D(k): column k now holds */
899 /* W(k) = U(k)*D(k) */
901 /* where U(k) is the k-th column of U */
903 /* Perform a rank-1 update of A(1:k-1,1:k-1) as */
905 /* A := A - U(k)*D(k)*U(k)**T = A - W(k)*1/D(k)*W(k)**T */
907 z_div(&z__1, &c_b1, &ap[kc + k - 1]);
908 r1.r = z__1.r, r1.i = z__1.i;
910 z__1.r = -r1.r, z__1.i = -r1.i;
911 zspr_(uplo, &i__1, &z__1, &ap[kc], &c__1, &ap[1]);
913 /* Store U(k) in column k */
916 zscal_(&i__1, &r1, &ap[kc], &c__1);
919 /* 2-by-2 pivot block D(k): columns k and k-1 now hold */
921 /* ( W(k-1) W(k) ) = ( U(k-1) U(k) )*D(k) */
923 /* where U(k) and U(k-1) are the k-th and (k-1)-th columns */
926 /* Perform a rank-2 update of A(1:k-2,1:k-2) as */
928 /* A := A - ( U(k-1) U(k) )*D(k)*( U(k-1) U(k) )**T */
929 /* = A - ( W(k-1) W(k) )*inv(D(k))*( W(k-1) W(k) )**T */
933 i__1 = k - 1 + (k - 1) * k / 2;
934 d12.r = ap[i__1].r, d12.i = ap[i__1].i;
935 z_div(&z__1, &ap[k - 1 + (k - 2) * (k - 1) / 2], &d12);
936 d22.r = z__1.r, d22.i = z__1.i;
937 z_div(&z__1, &ap[k + (k - 1) * k / 2], &d12);
938 d11.r = z__1.r, d11.i = z__1.i;
939 z__3.r = d11.r * d22.r - d11.i * d22.i, z__3.i = d11.r *
940 d22.i + d11.i * d22.r;
941 z__2.r = z__3.r - 1., z__2.i = z__3.i + 0.;
942 z_div(&z__1, &c_b1, &z__2);
943 t.r = z__1.r, t.i = z__1.i;
944 z_div(&z__1, &t, &d12);
945 d12.r = z__1.r, d12.i = z__1.i;
947 for (j = k - 2; j >= 1; --j) {
948 i__1 = j + (k - 2) * (k - 1) / 2;
949 z__3.r = d11.r * ap[i__1].r - d11.i * ap[i__1].i,
950 z__3.i = d11.r * ap[i__1].i + d11.i * ap[i__1]
952 i__2 = j + (k - 1) * k / 2;
953 z__2.r = z__3.r - ap[i__2].r, z__2.i = z__3.i - ap[
955 z__1.r = d12.r * z__2.r - d12.i * z__2.i, z__1.i =
956 d12.r * z__2.i + d12.i * z__2.r;
957 wkm1.r = z__1.r, wkm1.i = z__1.i;
958 i__1 = j + (k - 1) * k / 2;
959 z__3.r = d22.r * ap[i__1].r - d22.i * ap[i__1].i,
960 z__3.i = d22.r * ap[i__1].i + d22.i * ap[i__1]
962 i__2 = j + (k - 2) * (k - 1) / 2;
963 z__2.r = z__3.r - ap[i__2].r, z__2.i = z__3.i - ap[
965 z__1.r = d12.r * z__2.r - d12.i * z__2.i, z__1.i =
966 d12.r * z__2.i + d12.i * z__2.r;
967 wk.r = z__1.r, wk.i = z__1.i;
968 for (i__ = j; i__ >= 1; --i__) {
969 i__1 = i__ + (j - 1) * j / 2;
970 i__2 = i__ + (j - 1) * j / 2;
971 i__3 = i__ + (k - 1) * k / 2;
972 z__3.r = ap[i__3].r * wk.r - ap[i__3].i * wk.i,
973 z__3.i = ap[i__3].r * wk.i + ap[i__3].i *
975 z__2.r = ap[i__2].r - z__3.r, z__2.i = ap[i__2].i
977 i__4 = i__ + (k - 2) * (k - 1) / 2;
978 z__4.r = ap[i__4].r * wkm1.r - ap[i__4].i *
979 wkm1.i, z__4.i = ap[i__4].r * wkm1.i + ap[
981 z__1.r = z__2.r - z__4.r, z__1.i = z__2.i -
983 ap[i__1].r = z__1.r, ap[i__1].i = z__1.i;
986 i__1 = j + (k - 1) * k / 2;
987 ap[i__1].r = wk.r, ap[i__1].i = wk.i;
988 i__1 = j + (k - 2) * (k - 1) / 2;
989 ap[i__1].r = wkm1.r, ap[i__1].i = wkm1.i;
997 /* Store details of the interchanges in IPIV */
1006 /* Decrease K and return to the start of the main loop */
1014 /* Factorize A as L*D*L**T using the lower triangle of A */
1016 /* K is the main loop index, increasing from 1 to N in steps of */
1021 npp = *n * (*n + 1) / 2;
1025 /* If K > N, exit from loop */
1032 /* Determine rows and columns to be interchanged and whether */
1033 /* a 1-by-1 or 2-by-2 pivot block will be used */
1036 absakk = (d__1 = ap[i__1].r, abs(d__1)) + (d__2 = d_imag(&ap[kc]),
1039 /* IMAX is the row-index of the largest off-diagonal element in */
1040 /* column K, and COLMAX is its absolute value */
1044 imax = k + izamax_(&i__1, &ap[kc + 1], &c__1);
1045 i__1 = kc + imax - k;
1046 colmax = (d__1 = ap[i__1].r, abs(d__1)) + (d__2 = d_imag(&ap[kc +
1047 imax - k]), abs(d__2));
1052 if (f2cmax(absakk,colmax) == 0.) {
1054 /* Column K is zero: set INFO and continue */
1061 if (absakk >= alpha * colmax) {
1063 /* no interchange, use 1-by-1 pivot block */
1068 /* JMAX is the column-index of the largest off-diagonal */
1069 /* element in row IMAX, and ROWMAX is its absolute value */
1074 for (j = k; j <= i__1; ++j) {
1076 if ((d__1 = ap[i__2].r, abs(d__1)) + (d__2 = d_imag(&ap[
1077 kx]), abs(d__2)) > rowmax) {
1079 rowmax = (d__1 = ap[i__2].r, abs(d__1)) + (d__2 =
1080 d_imag(&ap[kx]), abs(d__2));
1086 kpc = npp - (*n - imax + 1) * (*n - imax + 2) / 2 + 1;
1089 jmax = imax + izamax_(&i__1, &ap[kpc + 1], &c__1);
1091 i__1 = kpc + jmax - imax;
1092 d__3 = rowmax, d__4 = (d__1 = ap[i__1].r, abs(d__1)) + (
1093 d__2 = d_imag(&ap[kpc + jmax - imax]), abs(d__2));
1094 rowmax = f2cmax(d__3,d__4);
1097 if (absakk >= alpha * colmax * (colmax / rowmax)) {
1099 /* no interchange, use 1-by-1 pivot block */
1102 } else /* if(complicated condition) */ {
1104 if ((d__1 = ap[i__1].r, abs(d__1)) + (d__2 = d_imag(&ap[
1105 kpc]), abs(d__2)) >= alpha * rowmax) {
1107 /* interchange rows and columns K and IMAX, use 1-by-1 */
1113 /* interchange rows and columns K+1 and IMAX, use 2-by-2 */
1124 knc = knc + *n - k + 1;
1128 /* Interchange rows and columns KK and KP in the trailing */
1129 /* submatrix A(k:n,k:n) */
1133 zswap_(&i__1, &ap[knc + kp - kk + 1], &c__1, &ap[kpc + 1],
1138 for (j = kk + 1; j <= i__1; ++j) {
1139 kx = kx + *n - j + 1;
1140 i__2 = knc + j - kk;
1141 t.r = ap[i__2].r, t.i = ap[i__2].i;
1142 i__2 = knc + j - kk;
1144 ap[i__2].r = ap[i__3].r, ap[i__2].i = ap[i__3].i;
1146 ap[i__2].r = t.r, ap[i__2].i = t.i;
1150 t.r = ap[i__1].r, t.i = ap[i__1].i;
1153 ap[i__1].r = ap[i__2].r, ap[i__1].i = ap[i__2].i;
1155 ap[i__1].r = t.r, ap[i__1].i = t.i;
1158 t.r = ap[i__1].r, t.i = ap[i__1].i;
1161 ap[i__1].r = ap[i__2].r, ap[i__1].i = ap[i__2].i;
1163 ap[i__1].r = t.r, ap[i__1].i = t.i;
1167 /* Update the trailing submatrix */
1171 /* 1-by-1 pivot block D(k): column k now holds */
1173 /* W(k) = L(k)*D(k) */
1175 /* where L(k) is the k-th column of L */
1179 /* Perform a rank-1 update of A(k+1:n,k+1:n) as */
1181 /* A := A - L(k)*D(k)*L(k)**T = A - W(k)*(1/D(k))*W(k)**T */
1183 z_div(&z__1, &c_b1, &ap[kc]);
1184 r1.r = z__1.r, r1.i = z__1.i;
1186 z__1.r = -r1.r, z__1.i = -r1.i;
1187 zspr_(uplo, &i__1, &z__1, &ap[kc + 1], &c__1, &ap[kc + *n
1190 /* Store L(k) in column K */
1193 zscal_(&i__1, &r1, &ap[kc + 1], &c__1);
1197 /* 2-by-2 pivot block D(k): columns K and K+1 now hold */
1199 /* ( W(k) W(k+1) ) = ( L(k) L(k+1) )*D(k) */
1201 /* where L(k) and L(k+1) are the k-th and (k+1)-th columns */
1206 /* Perform a rank-2 update of A(k+2:n,k+2:n) as */
1208 /* A := A - ( L(k) L(k+1) )*D(k)*( L(k) L(k+1) )**T */
1209 /* = A - ( W(k) W(k+1) )*inv(D(k))*( W(k) W(k+1) )**T */
1211 /* where L(k) and L(k+1) are the k-th and (k+1)-th */
1214 i__1 = k + 1 + (k - 1) * ((*n << 1) - k) / 2;
1215 d21.r = ap[i__1].r, d21.i = ap[i__1].i;
1216 z_div(&z__1, &ap[k + 1 + k * ((*n << 1) - k - 1) / 2], &
1218 d11.r = z__1.r, d11.i = z__1.i;
1219 z_div(&z__1, &ap[k + (k - 1) * ((*n << 1) - k) / 2], &d21)
1221 d22.r = z__1.r, d22.i = z__1.i;
1222 z__3.r = d11.r * d22.r - d11.i * d22.i, z__3.i = d11.r *
1223 d22.i + d11.i * d22.r;
1224 z__2.r = z__3.r - 1., z__2.i = z__3.i + 0.;
1225 z_div(&z__1, &c_b1, &z__2);
1226 t.r = z__1.r, t.i = z__1.i;
1227 z_div(&z__1, &t, &d21);
1228 d21.r = z__1.r, d21.i = z__1.i;
1231 for (j = k + 2; j <= i__1; ++j) {
1232 i__2 = j + (k - 1) * ((*n << 1) - k) / 2;
1233 z__3.r = d11.r * ap[i__2].r - d11.i * ap[i__2].i,
1234 z__3.i = d11.r * ap[i__2].i + d11.i * ap[i__2]
1236 i__3 = j + k * ((*n << 1) - k - 1) / 2;
1237 z__2.r = z__3.r - ap[i__3].r, z__2.i = z__3.i - ap[
1239 z__1.r = d21.r * z__2.r - d21.i * z__2.i, z__1.i =
1240 d21.r * z__2.i + d21.i * z__2.r;
1241 wk.r = z__1.r, wk.i = z__1.i;
1242 i__2 = j + k * ((*n << 1) - k - 1) / 2;
1243 z__3.r = d22.r * ap[i__2].r - d22.i * ap[i__2].i,
1244 z__3.i = d22.r * ap[i__2].i + d22.i * ap[i__2]
1246 i__3 = j + (k - 1) * ((*n << 1) - k) / 2;
1247 z__2.r = z__3.r - ap[i__3].r, z__2.i = z__3.i - ap[
1249 z__1.r = d21.r * z__2.r - d21.i * z__2.i, z__1.i =
1250 d21.r * z__2.i + d21.i * z__2.r;
1251 wkp1.r = z__1.r, wkp1.i = z__1.i;
1253 for (i__ = j; i__ <= i__2; ++i__) {
1254 i__3 = i__ + (j - 1) * ((*n << 1) - j) / 2;
1255 i__4 = i__ + (j - 1) * ((*n << 1) - j) / 2;
1256 i__5 = i__ + (k - 1) * ((*n << 1) - k) / 2;
1257 z__3.r = ap[i__5].r * wk.r - ap[i__5].i * wk.i,
1258 z__3.i = ap[i__5].r * wk.i + ap[i__5].i *
1260 z__2.r = ap[i__4].r - z__3.r, z__2.i = ap[i__4].i
1262 i__6 = i__ + k * ((*n << 1) - k - 1) / 2;
1263 z__4.r = ap[i__6].r * wkp1.r - ap[i__6].i *
1264 wkp1.i, z__4.i = ap[i__6].r * wkp1.i + ap[
1266 z__1.r = z__2.r - z__4.r, z__1.i = z__2.i -
1268 ap[i__3].r = z__1.r, ap[i__3].i = z__1.i;
1271 i__2 = j + (k - 1) * ((*n << 1) - k) / 2;
1272 ap[i__2].r = wk.r, ap[i__2].i = wk.i;
1273 i__2 = j + k * ((*n << 1) - k - 1) / 2;
1274 ap[i__2].r = wkp1.r, ap[i__2].i = wkp1.i;
1281 /* Store details of the interchanges in IPIV */
1290 /* Increase K and return to the start of the main loop */
1293 kc = knc + *n - k + 2;