OpenBLAS/lapack-netlib/SRC/dlasy2.c

1056 lines
30 KiB
C

#include <math.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include <complex.h>
#ifdef complex
#undef complex
#endif
#ifdef I
#undef I
#endif
#if defined(_WIN64)
typedef long long BLASLONG;
typedef unsigned long long BLASULONG;
#else
typedef long BLASLONG;
typedef unsigned long BLASULONG;
#endif
#ifdef LAPACK_ILP64
typedef BLASLONG blasint;
#if defined(_WIN64)
#define blasabs(x) llabs(x)
#else
#define blasabs(x) labs(x)
#endif
#else
typedef int blasint;
#define blasabs(x) abs(x)
#endif
typedef blasint integer;
typedef unsigned int uinteger;
typedef char *address;
typedef short int shortint;
typedef float real;
typedef double doublereal;
typedef struct { real r, i; } complex;
typedef struct { doublereal r, i; } doublecomplex;
#ifdef _MSC_VER
static inline _Fcomplex Cf(complex *z) {_Fcomplex zz={z->r , z->i}; return zz;}
static inline _Dcomplex Cd(doublecomplex *z) {_Dcomplex zz={z->r , z->i};return zz;}
static inline _Fcomplex * _pCf(complex *z) {return (_Fcomplex*)z;}
static inline _Dcomplex * _pCd(doublecomplex *z) {return (_Dcomplex*)z;}
#else
static inline _Complex float Cf(complex *z) {return z->r + z->i*_Complex_I;}
static inline _Complex double Cd(doublecomplex *z) {return z->r + z->i*_Complex_I;}
static inline _Complex float * _pCf(complex *z) {return (_Complex float*)z;}
static inline _Complex double * _pCd(doublecomplex *z) {return (_Complex double*)z;}
#endif
#define pCf(z) (*_pCf(z))
#define pCd(z) (*_pCd(z))
typedef blasint logical;
typedef char logical1;
typedef char integer1;
#define TRUE_ (1)
#define FALSE_ (0)
/* Extern is for use with -E */
#ifndef Extern
#define Extern extern
#endif
/* I/O stuff */
typedef int flag;
typedef int ftnlen;
typedef int ftnint;
/*external read, write*/
typedef struct
{ flag cierr;
ftnint ciunit;
flag ciend;
char *cifmt;
ftnint cirec;
} cilist;
/*internal read, write*/
typedef struct
{ flag icierr;
char *iciunit;
flag iciend;
char *icifmt;
ftnint icirlen;
ftnint icirnum;
} icilist;
/*open*/
typedef struct
{ flag oerr;
ftnint ounit;
char *ofnm;
ftnlen ofnmlen;
char *osta;
char *oacc;
char *ofm;
ftnint orl;
char *oblnk;
} olist;
/*close*/
typedef struct
{ flag cerr;
ftnint cunit;
char *csta;
} cllist;
/*rewind, backspace, endfile*/
typedef struct
{ flag aerr;
ftnint aunit;
} alist;
/* inquire */
typedef struct
{ flag inerr;
ftnint inunit;
char *infile;
ftnlen infilen;
ftnint *inex; /*parameters in standard's order*/
ftnint *inopen;
ftnint *innum;
ftnint *innamed;
char *inname;
ftnlen innamlen;
char *inacc;
ftnlen inacclen;
char *inseq;
ftnlen inseqlen;
char *indir;
ftnlen indirlen;
char *infmt;
ftnlen infmtlen;
char *inform;
ftnint informlen;
char *inunf;
ftnlen inunflen;
ftnint *inrecl;
ftnint *innrec;
char *inblank;
ftnlen inblanklen;
} inlist;
#define VOID void
union Multitype { /* for multiple entry points */
integer1 g;
shortint h;
integer i;
/* longint j; */
real r;
doublereal d;
complex c;
doublecomplex z;
};
typedef union Multitype Multitype;
struct Vardesc { /* for Namelist */
char *name;
char *addr;
ftnlen *dims;
int type;
};
typedef struct Vardesc Vardesc;
struct Namelist {
char *name;
Vardesc **vars;
int nvars;
};
typedef struct Namelist Namelist;
#define abs(x) ((x) >= 0 ? (x) : -(x))
#define dabs(x) (fabs(x))
#define f2cmin(a,b) ((a) <= (b) ? (a) : (b))
#define f2cmax(a,b) ((a) >= (b) ? (a) : (b))
#define dmin(a,b) (f2cmin(a,b))
#define dmax(a,b) (f2cmax(a,b))
#define bit_test(a,b) ((a) >> (b) & 1)
#define bit_clear(a,b) ((a) & ~((uinteger)1 << (b)))
#define bit_set(a,b) ((a) | ((uinteger)1 << (b)))
#define abort_() { sig_die("Fortran abort routine called", 1); }
#define c_abs(z) (cabsf(Cf(z)))
#define c_cos(R,Z) { pCf(R)=ccos(Cf(Z)); }
#ifdef _MSC_VER
#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]);}
#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]);}
#else
#define c_div(c, a, b) {pCf(c) = Cf(a)/Cf(b);}
#define z_div(c, a, b) {pCd(c) = Cd(a)/Cd(b);}
#endif
#define c_exp(R, Z) {pCf(R) = cexpf(Cf(Z));}
#define c_log(R, Z) {pCf(R) = clogf(Cf(Z));}
#define c_sin(R, Z) {pCf(R) = csinf(Cf(Z));}
//#define c_sqrt(R, Z) {*(R) = csqrtf(Cf(Z));}
#define c_sqrt(R, Z) {pCf(R) = csqrtf(Cf(Z));}
#define d_abs(x) (fabs(*(x)))
#define d_acos(x) (acos(*(x)))
#define d_asin(x) (asin(*(x)))
#define d_atan(x) (atan(*(x)))
#define d_atn2(x, y) (atan2(*(x),*(y)))
#define d_cnjg(R, Z) { pCd(R) = conj(Cd(Z)); }
#define r_cnjg(R, Z) { pCf(R) = conjf(Cf(Z)); }
#define d_cos(x) (cos(*(x)))
#define d_cosh(x) (cosh(*(x)))
#define d_dim(__a, __b) ( *(__a) > *(__b) ? *(__a) - *(__b) : 0.0 )
#define d_exp(x) (exp(*(x)))
#define d_imag(z) (cimag(Cd(z)))
#define r_imag(z) (cimagf(Cf(z)))
#define d_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
#define r_int(__x) (*(__x)>0 ? floor(*(__x)) : -floor(- *(__x)))
#define d_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
#define r_lg10(x) ( 0.43429448190325182765 * log(*(x)) )
#define d_log(x) (log(*(x)))
#define d_mod(x, y) (fmod(*(x), *(y)))
#define u_nint(__x) ((__x)>=0 ? floor((__x) + .5) : -floor(.5 - (__x)))
#define d_nint(x) u_nint(*(x))
#define u_sign(__a,__b) ((__b) >= 0 ? ((__a) >= 0 ? (__a) : -(__a)) : -((__a) >= 0 ? (__a) : -(__a)))
#define d_sign(a,b) u_sign(*(a),*(b))
#define r_sign(a,b) u_sign(*(a),*(b))
#define d_sin(x) (sin(*(x)))
#define d_sinh(x) (sinh(*(x)))
#define d_sqrt(x) (sqrt(*(x)))
#define d_tan(x) (tan(*(x)))
#define d_tanh(x) (tanh(*(x)))
#define i_abs(x) abs(*(x))
#define i_dnnt(x) ((integer)u_nint(*(x)))
#define i_len(s, n) (n)
#define i_nint(x) ((integer)u_nint(*(x)))
#define i_sign(a,b) ((integer)u_sign((integer)*(a),(integer)*(b)))
#define pow_dd(ap, bp) ( pow(*(ap), *(bp)))
#define pow_si(B,E) spow_ui(*(B),*(E))
#define pow_ri(B,E) spow_ui(*(B),*(E))
#define pow_di(B,E) dpow_ui(*(B),*(E))
#define pow_zi(p, a, b) {pCd(p) = zpow_ui(Cd(a), *(b));}
#define pow_ci(p, a, b) {pCf(p) = cpow_ui(Cf(a), *(b));}
#define pow_zz(R,A,B) {pCd(R) = cpow(Cd(A),*(B));}
#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++ = ' '; }
#define s_cmp(a,b,c,d) ((integer)strncmp((a),(b),f2cmin((c),(d))))
#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]; }
#define sig_die(s, kill) { exit(1); }
#define s_stop(s, n) {exit(0);}
static char junk[] = "\n@(#)LIBF77 VERSION 19990503\n";
#define z_abs(z) (cabs(Cd(z)))
#define z_exp(R, Z) {pCd(R) = cexp(Cd(Z));}
#define z_sqrt(R, Z) {pCd(R) = csqrt(Cd(Z));}
#define myexit_() break;
#define mycycle() continue;
#define myceiling(w) {ceil(w)}
#define myhuge(w) {HUGE_VAL}
//#define mymaxloc_(w,s,e,n) {if (sizeof(*(w)) == sizeof(double)) dmaxloc_((w),*(s),*(e),n); else dmaxloc_((w),*(s),*(e),n);}
#define mymaxloc(w,s,e,n) {dmaxloc_(w,*(s),*(e),n)}
/* procedure parameter types for -A and -C++ */
#ifdef __cplusplus
typedef logical (*L_fp)(...);
#else
typedef logical (*L_fp)();
#endif
static float spow_ui(float x, integer n) {
float pow=1.0; unsigned long int u;
if(n != 0) {
if(n < 0) n = -n, x = 1/x;
for(u = n; ; ) {
if(u & 01) pow *= x;
if(u >>= 1) x *= x;
else break;
}
}
return pow;
}
static double dpow_ui(double x, integer n) {
double pow=1.0; unsigned long int u;
if(n != 0) {
if(n < 0) n = -n, x = 1/x;
for(u = n; ; ) {
if(u & 01) pow *= x;
if(u >>= 1) x *= x;
else break;
}
}
return pow;
}
#ifdef _MSC_VER
static _Fcomplex cpow_ui(complex x, integer n) {
complex pow={1.0,0.0}; unsigned long int u;
if(n != 0) {
if(n < 0) n = -n, x.r = 1/x.r, x.i=1/x.i;
for(u = n; ; ) {
if(u & 01) pow.r *= x.r, pow.i *= x.i;
if(u >>= 1) x.r *= x.r, x.i *= x.i;
else break;
}
}
_Fcomplex p={pow.r, pow.i};
return p;
}
#else
static _Complex float cpow_ui(_Complex float x, integer n) {
_Complex float pow=1.0; unsigned long int u;
if(n != 0) {
if(n < 0) n = -n, x = 1/x;
for(u = n; ; ) {
if(u & 01) pow *= x;
if(u >>= 1) x *= x;
else break;
}
}
return pow;
}
#endif
#ifdef _MSC_VER
static _Dcomplex zpow_ui(_Dcomplex x, integer n) {
_Dcomplex pow={1.0,0.0}; unsigned long int u;
if(n != 0) {
if(n < 0) n = -n, x._Val[0] = 1/x._Val[0], x._Val[1] =1/x._Val[1];
for(u = n; ; ) {
if(u & 01) pow._Val[0] *= x._Val[0], pow._Val[1] *= x._Val[1];
if(u >>= 1) x._Val[0] *= x._Val[0], x._Val[1] *= x._Val[1];
else break;
}
}
_Dcomplex p = {pow._Val[0], pow._Val[1]};
return p;
}
#else
static _Complex double zpow_ui(_Complex double x, integer n) {
_Complex double pow=1.0; unsigned long int u;
if(n != 0) {
if(n < 0) n = -n, x = 1/x;
for(u = n; ; ) {
if(u & 01) pow *= x;
if(u >>= 1) x *= x;
else break;
}
}
return pow;
}
#endif
static integer pow_ii(integer x, integer n) {
integer pow; unsigned long int u;
if (n <= 0) {
if (n == 0 || x == 1) pow = 1;
else if (x != -1) pow = x == 0 ? 1/x : 0;
else n = -n;
}
if ((n > 0) || !(n == 0 || x == 1 || x != -1)) {
u = n;
for(pow = 1; ; ) {
if(u & 01) pow *= x;
if(u >>= 1) x *= x;
else break;
}
}
return pow;
}
static integer dmaxloc_(double *w, integer s, integer e, integer *n)
{
double m; integer i, mi;
for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
if (w[i-1]>m) mi=i ,m=w[i-1];
return mi-s+1;
}
static integer smaxloc_(float *w, integer s, integer e, integer *n)
{
float m; integer i, mi;
for(m=w[s-1], mi=s, i=s+1; i<=e; i++)
if (w[i-1]>m) mi=i ,m=w[i-1];
return mi-s+1;
}
static inline void cdotc_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
integer n = *n_, incx = *incx_, incy = *incy_, i;
#ifdef _MSC_VER
_Fcomplex zdotc = {0.0, 0.0};
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += conjf(Cf(&x[i]))._Val[0] * Cf(&y[i])._Val[0];
zdotc._Val[1] += conjf(Cf(&x[i]))._Val[1] * Cf(&y[i])._Val[1];
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += conjf(Cf(&x[i*incx]))._Val[0] * Cf(&y[i*incy])._Val[0];
zdotc._Val[1] += conjf(Cf(&x[i*incx]))._Val[1] * Cf(&y[i*incy])._Val[1];
}
}
pCf(z) = zdotc;
}
#else
_Complex float zdotc = 0.0;
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += conjf(Cf(&x[i])) * Cf(&y[i]);
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += conjf(Cf(&x[i*incx])) * Cf(&y[i*incy]);
}
}
pCf(z) = zdotc;
}
#endif
static inline void zdotc_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
integer n = *n_, incx = *incx_, incy = *incy_, i;
#ifdef _MSC_VER
_Dcomplex zdotc = {0.0, 0.0};
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += conj(Cd(&x[i]))._Val[0] * Cd(&y[i])._Val[0];
zdotc._Val[1] += conj(Cd(&x[i]))._Val[1] * Cd(&y[i])._Val[1];
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += conj(Cd(&x[i*incx]))._Val[0] * Cd(&y[i*incy])._Val[0];
zdotc._Val[1] += conj(Cd(&x[i*incx]))._Val[1] * Cd(&y[i*incy])._Val[1];
}
}
pCd(z) = zdotc;
}
#else
_Complex double zdotc = 0.0;
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += conj(Cd(&x[i])) * Cd(&y[i]);
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += conj(Cd(&x[i*incx])) * Cd(&y[i*incy]);
}
}
pCd(z) = zdotc;
}
#endif
static inline void cdotu_(complex *z, integer *n_, complex *x, integer *incx_, complex *y, integer *incy_) {
integer n = *n_, incx = *incx_, incy = *incy_, i;
#ifdef _MSC_VER
_Fcomplex zdotc = {0.0, 0.0};
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += Cf(&x[i])._Val[0] * Cf(&y[i])._Val[0];
zdotc._Val[1] += Cf(&x[i])._Val[1] * Cf(&y[i])._Val[1];
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += Cf(&x[i*incx])._Val[0] * Cf(&y[i*incy])._Val[0];
zdotc._Val[1] += Cf(&x[i*incx])._Val[1] * Cf(&y[i*incy])._Val[1];
}
}
pCf(z) = zdotc;
}
#else
_Complex float zdotc = 0.0;
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += Cf(&x[i]) * Cf(&y[i]);
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += Cf(&x[i*incx]) * Cf(&y[i*incy]);
}
}
pCf(z) = zdotc;
}
#endif
static inline void zdotu_(doublecomplex *z, integer *n_, doublecomplex *x, integer *incx_, doublecomplex *y, integer *incy_) {
integer n = *n_, incx = *incx_, incy = *incy_, i;
#ifdef _MSC_VER
_Dcomplex zdotc = {0.0, 0.0};
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += Cd(&x[i])._Val[0] * Cd(&y[i])._Val[0];
zdotc._Val[1] += Cd(&x[i])._Val[1] * Cd(&y[i])._Val[1];
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc._Val[0] += Cd(&x[i*incx])._Val[0] * Cd(&y[i*incy])._Val[0];
zdotc._Val[1] += Cd(&x[i*incx])._Val[1] * Cd(&y[i*incy])._Val[1];
}
}
pCd(z) = zdotc;
}
#else
_Complex double zdotc = 0.0;
if (incx == 1 && incy == 1) {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += Cd(&x[i]) * Cd(&y[i]);
}
} else {
for (i=0;i<n;i++) { /* zdotc = zdotc + dconjg(x(i))* y(i) */
zdotc += Cd(&x[i*incx]) * Cd(&y[i*incy]);
}
}
pCd(z) = zdotc;
}
#endif
/* -- translated by f2c (version 20000121).
You must link the resulting object file with the libraries:
-lf2c -lm (in that order)
*/
/* Table of constant values */
static integer c__4 = 4;
static integer c__1 = 1;
static integer c__16 = 16;
static integer c__0 = 0;
/* > \brief \b DLASY2 solves the Sylvester matrix equation where the matrices are of order 1 or 2. */
/* =========== DOCUMENTATION =========== */
/* Online html documentation available at */
/* http://www.netlib.org/lapack/explore-html/ */
/* > \htmlonly */
/* > Download DLASY2 + dependencies */
/* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.tgz?format=tgz&filename=/lapack/lapack_routine/dlasy2.
f"> */
/* > [TGZ]</a> */
/* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.zip?format=zip&filename=/lapack/lapack_routine/dlasy2.
f"> */
/* > [ZIP]</a> */
/* > <a href="http://www.netlib.org/cgi-bin/netlibfiles.txt?format=txt&filename=/lapack/lapack_routine/dlasy2.
f"> */
/* > [TXT]</a> */
/* > \endhtmlonly */
/* Definition: */
/* =========== */
/* SUBROUTINE DLASY2( LTRANL, LTRANR, ISGN, N1, N2, TL, LDTL, TR, */
/* LDTR, B, LDB, SCALE, X, LDX, XNORM, INFO ) */
/* LOGICAL LTRANL, LTRANR */
/* INTEGER INFO, ISGN, LDB, LDTL, LDTR, LDX, N1, N2 */
/* DOUBLE PRECISION SCALE, XNORM */
/* DOUBLE PRECISION B( LDB, * ), TL( LDTL, * ), TR( LDTR, * ), */
/* $ X( LDX, * ) */
/* > \par Purpose: */
/* ============= */
/* > */
/* > \verbatim */
/* > */
/* > DLASY2 solves for the N1 by N2 matrix X, 1 <= N1,N2 <= 2, in */
/* > */
/* > op(TL)*X + ISGN*X*op(TR) = SCALE*B, */
/* > */
/* > where TL is N1 by N1, TR is N2 by N2, B is N1 by N2, and ISGN = 1 or */
/* > -1. op(T) = T or T**T, where T**T denotes the transpose of T. */
/* > \endverbatim */
/* Arguments: */
/* ========== */
/* > \param[in] LTRANL */
/* > \verbatim */
/* > LTRANL is LOGICAL */
/* > On entry, LTRANL specifies the op(TL): */
/* > = .FALSE., op(TL) = TL, */
/* > = .TRUE., op(TL) = TL**T. */
/* > \endverbatim */
/* > */
/* > \param[in] LTRANR */
/* > \verbatim */
/* > LTRANR is LOGICAL */
/* > On entry, LTRANR specifies the op(TR): */
/* > = .FALSE., op(TR) = TR, */
/* > = .TRUE., op(TR) = TR**T. */
/* > \endverbatim */
/* > */
/* > \param[in] ISGN */
/* > \verbatim */
/* > ISGN is INTEGER */
/* > On entry, ISGN specifies the sign of the equation */
/* > as described before. ISGN may only be 1 or -1. */
/* > \endverbatim */
/* > */
/* > \param[in] N1 */
/* > \verbatim */
/* > N1 is INTEGER */
/* > On entry, N1 specifies the order of matrix TL. */
/* > N1 may only be 0, 1 or 2. */
/* > \endverbatim */
/* > */
/* > \param[in] N2 */
/* > \verbatim */
/* > N2 is INTEGER */
/* > On entry, N2 specifies the order of matrix TR. */
/* > N2 may only be 0, 1 or 2. */
/* > \endverbatim */
/* > */
/* > \param[in] TL */
/* > \verbatim */
/* > TL is DOUBLE PRECISION array, dimension (LDTL,2) */
/* > On entry, TL contains an N1 by N1 matrix. */
/* > \endverbatim */
/* > */
/* > \param[in] LDTL */
/* > \verbatim */
/* > LDTL is INTEGER */
/* > The leading dimension of the matrix TL. LDTL >= f2cmax(1,N1). */
/* > \endverbatim */
/* > */
/* > \param[in] TR */
/* > \verbatim */
/* > TR is DOUBLE PRECISION array, dimension (LDTR,2) */
/* > On entry, TR contains an N2 by N2 matrix. */
/* > \endverbatim */
/* > */
/* > \param[in] LDTR */
/* > \verbatim */
/* > LDTR is INTEGER */
/* > The leading dimension of the matrix TR. LDTR >= f2cmax(1,N2). */
/* > \endverbatim */
/* > */
/* > \param[in] B */
/* > \verbatim */
/* > B is DOUBLE PRECISION array, dimension (LDB,2) */
/* > On entry, the N1 by N2 matrix B contains the right-hand */
/* > side of the equation. */
/* > \endverbatim */
/* > */
/* > \param[in] LDB */
/* > \verbatim */
/* > LDB is INTEGER */
/* > The leading dimension of the matrix B. LDB >= f2cmax(1,N1). */
/* > \endverbatim */
/* > */
/* > \param[out] SCALE */
/* > \verbatim */
/* > SCALE is DOUBLE PRECISION */
/* > On exit, SCALE contains the scale factor. SCALE is chosen */
/* > less than or equal to 1 to prevent the solution overflowing. */
/* > \endverbatim */
/* > */
/* > \param[out] X */
/* > \verbatim */
/* > X is DOUBLE PRECISION array, dimension (LDX,2) */
/* > On exit, X contains the N1 by N2 solution. */
/* > \endverbatim */
/* > */
/* > \param[in] LDX */
/* > \verbatim */
/* > LDX is INTEGER */
/* > The leading dimension of the matrix X. LDX >= f2cmax(1,N1). */
/* > \endverbatim */
/* > */
/* > \param[out] XNORM */
/* > \verbatim */
/* > XNORM is DOUBLE PRECISION */
/* > On exit, XNORM is the infinity-norm of the solution. */
/* > \endverbatim */
/* > */
/* > \param[out] INFO */
/* > \verbatim */
/* > INFO is INTEGER */
/* > On exit, INFO is set to */
/* > 0: successful exit. */
/* > 1: TL and TR have too close eigenvalues, so TL or */
/* > TR is perturbed to get a nonsingular equation. */
/* > NOTE: In the interests of speed, this routine does not */
/* > check the inputs for errors. */
/* > \endverbatim */
/* Authors: */
/* ======== */
/* > \author Univ. of Tennessee */
/* > \author Univ. of California Berkeley */
/* > \author Univ. of Colorado Denver */
/* > \author NAG Ltd. */
/* > \date June 2016 */
/* > \ingroup doubleSYauxiliary */
/* ===================================================================== */
/* Subroutine */ void dlasy2_(logical *ltranl, logical *ltranr, integer *isgn,
integer *n1, integer *n2, doublereal *tl, integer *ldtl, doublereal *
tr, integer *ldtr, doublereal *b, integer *ldb, doublereal *scale,
doublereal *x, integer *ldx, doublereal *xnorm, integer *info)
{
/* Initialized data */
static integer locu12[4] = { 3,4,1,2 };
static integer locl21[4] = { 2,1,4,3 };
static integer locu22[4] = { 4,3,2,1 };
static logical xswpiv[4] = { FALSE_,FALSE_,TRUE_,TRUE_ };
static logical bswpiv[4] = { FALSE_,TRUE_,FALSE_,TRUE_ };
/* System generated locals */
integer b_dim1, b_offset, tl_dim1, tl_offset, tr_dim1, tr_offset, x_dim1,
x_offset;
doublereal d__1, d__2, d__3, d__4, d__5, d__6, d__7, d__8;
/* Local variables */
doublereal btmp[4], smin;
integer ipiv;
doublereal temp;
integer jpiv[4];
doublereal xmax;
integer ipsv, jpsv, i__, j, k;
logical bswap;
extern /* Subroutine */ void dcopy_(integer *, doublereal *, integer *,
doublereal *, integer *), dswap_(integer *, doublereal *, integer
*, doublereal *, integer *);
logical xswap;
doublereal x2[2], l21, u11, u12;
integer ip, jp;
doublereal u22, t16[16] /* was [4][4] */;
extern doublereal dlamch_(char *);
extern integer idamax_(integer *, doublereal *, integer *);
doublereal smlnum, gam, bet, eps, sgn, tmp[4], tau1;
/* -- LAPACK auxiliary routine (version 3.7.0) -- */
/* -- LAPACK is a software package provided by Univ. of Tennessee, -- */
/* -- Univ. of California Berkeley, Univ. of Colorado Denver and NAG Ltd..-- */
/* June 2016 */
/* ===================================================================== */
/* Parameter adjustments */
tl_dim1 = *ldtl;
tl_offset = 1 + tl_dim1 * 1;
tl -= tl_offset;
tr_dim1 = *ldtr;
tr_offset = 1 + tr_dim1 * 1;
tr -= tr_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1 * 1;
b -= b_offset;
x_dim1 = *ldx;
x_offset = 1 + x_dim1 * 1;
x -= x_offset;
/* Function Body */
/* Do not check the input parameters for errors */
*info = 0;
/* Quick return if possible */
if (*n1 == 0 || *n2 == 0) {
return;
}
/* Set constants to control overflow */
eps = dlamch_("P");
smlnum = dlamch_("S") / eps;
sgn = (doublereal) (*isgn);
k = *n1 + *n1 + *n2 - 2;
switch (k) {
case 1: goto L10;
case 2: goto L20;
case 3: goto L30;
case 4: goto L50;
}
/* 1 by 1: TL11*X + SGN*X*TR11 = B11 */
L10:
tau1 = tl[tl_dim1 + 1] + sgn * tr[tr_dim1 + 1];
bet = abs(tau1);
if (bet <= smlnum) {
tau1 = smlnum;
bet = smlnum;
*info = 1;
}
*scale = 1.;
gam = (d__1 = b[b_dim1 + 1], abs(d__1));
if (smlnum * gam > bet) {
*scale = 1. / gam;
}
x[x_dim1 + 1] = b[b_dim1 + 1] * *scale / tau1;
*xnorm = (d__1 = x[x_dim1 + 1], abs(d__1));
return;
/* 1 by 2: */
/* TL11*[X11 X12] + ISGN*[X11 X12]*op[TR11 TR12] = [B11 B12] */
/* [TR21 TR22] */
L20:
/* Computing MAX */
/* Computing MAX */
d__7 = (d__1 = tl[tl_dim1 + 1], abs(d__1)), d__8 = (d__2 = tr[tr_dim1 + 1]
, abs(d__2)), d__7 = f2cmax(d__7,d__8), d__8 = (d__3 = tr[(tr_dim1 <<
1) + 1], abs(d__3)), d__7 = f2cmax(d__7,d__8), d__8 = (d__4 = tr[
tr_dim1 + 2], abs(d__4)), d__7 = f2cmax(d__7,d__8), d__8 = (d__5 =
tr[(tr_dim1 << 1) + 2], abs(d__5));
d__6 = eps * f2cmax(d__7,d__8);
smin = f2cmax(d__6,smlnum);
tmp[0] = tl[tl_dim1 + 1] + sgn * tr[tr_dim1 + 1];
tmp[3] = tl[tl_dim1 + 1] + sgn * tr[(tr_dim1 << 1) + 2];
if (*ltranr) {
tmp[1] = sgn * tr[tr_dim1 + 2];
tmp[2] = sgn * tr[(tr_dim1 << 1) + 1];
} else {
tmp[1] = sgn * tr[(tr_dim1 << 1) + 1];
tmp[2] = sgn * tr[tr_dim1 + 2];
}
btmp[0] = b[b_dim1 + 1];
btmp[1] = b[(b_dim1 << 1) + 1];
goto L40;
/* 2 by 1: */
/* op[TL11 TL12]*[X11] + ISGN* [X11]*TR11 = [B11] */
/* [TL21 TL22] [X21] [X21] [B21] */
L30:
/* Computing MAX */
/* Computing MAX */
d__7 = (d__1 = tr[tr_dim1 + 1], abs(d__1)), d__8 = (d__2 = tl[tl_dim1 + 1]
, abs(d__2)), d__7 = f2cmax(d__7,d__8), d__8 = (d__3 = tl[(tl_dim1 <<
1) + 1], abs(d__3)), d__7 = f2cmax(d__7,d__8), d__8 = (d__4 = tl[
tl_dim1 + 2], abs(d__4)), d__7 = f2cmax(d__7,d__8), d__8 = (d__5 =
tl[(tl_dim1 << 1) + 2], abs(d__5));
d__6 = eps * f2cmax(d__7,d__8);
smin = f2cmax(d__6,smlnum);
tmp[0] = tl[tl_dim1 + 1] + sgn * tr[tr_dim1 + 1];
tmp[3] = tl[(tl_dim1 << 1) + 2] + sgn * tr[tr_dim1 + 1];
if (*ltranl) {
tmp[1] = tl[(tl_dim1 << 1) + 1];
tmp[2] = tl[tl_dim1 + 2];
} else {
tmp[1] = tl[tl_dim1 + 2];
tmp[2] = tl[(tl_dim1 << 1) + 1];
}
btmp[0] = b[b_dim1 + 1];
btmp[1] = b[b_dim1 + 2];
L40:
/* Solve 2 by 2 system using complete pivoting. */
/* Set pivots less than SMIN to SMIN. */
ipiv = idamax_(&c__4, tmp, &c__1);
u11 = tmp[ipiv - 1];
if (abs(u11) <= smin) {
*info = 1;
u11 = smin;
}
u12 = tmp[locu12[ipiv - 1] - 1];
l21 = tmp[locl21[ipiv - 1] - 1] / u11;
u22 = tmp[locu22[ipiv - 1] - 1] - u12 * l21;
xswap = xswpiv[ipiv - 1];
bswap = bswpiv[ipiv - 1];
if (abs(u22) <= smin) {
*info = 1;
u22 = smin;
}
if (bswap) {
temp = btmp[1];
btmp[1] = btmp[0] - l21 * temp;
btmp[0] = temp;
} else {
btmp[1] -= l21 * btmp[0];
}
*scale = 1.;
if (smlnum * 2. * abs(btmp[1]) > abs(u22) || smlnum * 2. * abs(btmp[0]) >
abs(u11)) {
/* Computing MAX */
d__1 = abs(btmp[0]), d__2 = abs(btmp[1]);
*scale = .5 / f2cmax(d__1,d__2);
btmp[0] *= *scale;
btmp[1] *= *scale;
}
x2[1] = btmp[1] / u22;
x2[0] = btmp[0] / u11 - u12 / u11 * x2[1];
if (xswap) {
temp = x2[1];
x2[1] = x2[0];
x2[0] = temp;
}
x[x_dim1 + 1] = x2[0];
if (*n1 == 1) {
x[(x_dim1 << 1) + 1] = x2[1];
*xnorm = (d__1 = x[x_dim1 + 1], abs(d__1)) + (d__2 = x[(x_dim1 << 1)
+ 1], abs(d__2));
} else {
x[x_dim1 + 2] = x2[1];
/* Computing MAX */
d__3 = (d__1 = x[x_dim1 + 1], abs(d__1)), d__4 = (d__2 = x[x_dim1 + 2]
, abs(d__2));
*xnorm = f2cmax(d__3,d__4);
}
return;
/* 2 by 2: */
/* op[TL11 TL12]*[X11 X12] +ISGN* [X11 X12]*op[TR11 TR12] = [B11 B12] */
/* [TL21 TL22] [X21 X22] [X21 X22] [TR21 TR22] [B21 B22] */
/* Solve equivalent 4 by 4 system using complete pivoting. */
/* Set pivots less than SMIN to SMIN. */
L50:
/* Computing MAX */
d__5 = (d__1 = tr[tr_dim1 + 1], abs(d__1)), d__6 = (d__2 = tr[(tr_dim1 <<
1) + 1], abs(d__2)), d__5 = f2cmax(d__5,d__6), d__6 = (d__3 = tr[
tr_dim1 + 2], abs(d__3)), d__5 = f2cmax(d__5,d__6), d__6 = (d__4 =
tr[(tr_dim1 << 1) + 2], abs(d__4));
smin = f2cmax(d__5,d__6);
/* Computing MAX */
d__5 = smin, d__6 = (d__1 = tl[tl_dim1 + 1], abs(d__1)), d__5 = f2cmax(d__5,
d__6), d__6 = (d__2 = tl[(tl_dim1 << 1) + 1], abs(d__2)), d__5 =
f2cmax(d__5,d__6), d__6 = (d__3 = tl[tl_dim1 + 2], abs(d__3)), d__5 =
f2cmax(d__5,d__6), d__6 = (d__4 = tl[(tl_dim1 << 1) + 2], abs(d__4))
;
smin = f2cmax(d__5,d__6);
/* Computing MAX */
d__1 = eps * smin;
smin = f2cmax(d__1,smlnum);
btmp[0] = 0.;
dcopy_(&c__16, btmp, &c__0, t16, &c__1);
t16[0] = tl[tl_dim1 + 1] + sgn * tr[tr_dim1 + 1];
t16[5] = tl[(tl_dim1 << 1) + 2] + sgn * tr[tr_dim1 + 1];
t16[10] = tl[tl_dim1 + 1] + sgn * tr[(tr_dim1 << 1) + 2];
t16[15] = tl[(tl_dim1 << 1) + 2] + sgn * tr[(tr_dim1 << 1) + 2];
if (*ltranl) {
t16[4] = tl[tl_dim1 + 2];
t16[1] = tl[(tl_dim1 << 1) + 1];
t16[14] = tl[tl_dim1 + 2];
t16[11] = tl[(tl_dim1 << 1) + 1];
} else {
t16[4] = tl[(tl_dim1 << 1) + 1];
t16[1] = tl[tl_dim1 + 2];
t16[14] = tl[(tl_dim1 << 1) + 1];
t16[11] = tl[tl_dim1 + 2];
}
if (*ltranr) {
t16[8] = sgn * tr[(tr_dim1 << 1) + 1];
t16[13] = sgn * tr[(tr_dim1 << 1) + 1];
t16[2] = sgn * tr[tr_dim1 + 2];
t16[7] = sgn * tr[tr_dim1 + 2];
} else {
t16[8] = sgn * tr[tr_dim1 + 2];
t16[13] = sgn * tr[tr_dim1 + 2];
t16[2] = sgn * tr[(tr_dim1 << 1) + 1];
t16[7] = sgn * tr[(tr_dim1 << 1) + 1];
}
btmp[0] = b[b_dim1 + 1];
btmp[1] = b[b_dim1 + 2];
btmp[2] = b[(b_dim1 << 1) + 1];
btmp[3] = b[(b_dim1 << 1) + 2];
/* Perform elimination */
for (i__ = 1; i__ <= 3; ++i__) {
xmax = 0.;
for (ip = i__; ip <= 4; ++ip) {
for (jp = i__; jp <= 4; ++jp) {
if ((d__1 = t16[ip + (jp << 2) - 5], abs(d__1)) >= xmax) {
xmax = (d__1 = t16[ip + (jp << 2) - 5], abs(d__1));
ipsv = ip;
jpsv = jp;
}
/* L60: */
}
/* L70: */
}
if (ipsv != i__) {
dswap_(&c__4, &t16[ipsv - 1], &c__4, &t16[i__ - 1], &c__4);
temp = btmp[i__ - 1];
btmp[i__ - 1] = btmp[ipsv - 1];
btmp[ipsv - 1] = temp;
}
if (jpsv != i__) {
dswap_(&c__4, &t16[(jpsv << 2) - 4], &c__1, &t16[(i__ << 2) - 4],
&c__1);
}
jpiv[i__ - 1] = jpsv;
if ((d__1 = t16[i__ + (i__ << 2) - 5], abs(d__1)) < smin) {
*info = 1;
t16[i__ + (i__ << 2) - 5] = smin;
}
for (j = i__ + 1; j <= 4; ++j) {
t16[j + (i__ << 2) - 5] /= t16[i__ + (i__ << 2) - 5];
btmp[j - 1] -= t16[j + (i__ << 2) - 5] * btmp[i__ - 1];
for (k = i__ + 1; k <= 4; ++k) {
t16[j + (k << 2) - 5] -= t16[j + (i__ << 2) - 5] * t16[i__ + (
k << 2) - 5];
/* L80: */
}
/* L90: */
}
/* L100: */
}
if (abs(t16[15]) < smin) {
*info = 1;
t16[15] = smin;
}
*scale = 1.;
if (smlnum * 8. * abs(btmp[0]) > abs(t16[0]) || smlnum * 8. * abs(btmp[1])
> abs(t16[5]) || smlnum * 8. * abs(btmp[2]) > abs(t16[10]) ||
smlnum * 8. * abs(btmp[3]) > abs(t16[15])) {
/* Computing MAX */
d__1 = abs(btmp[0]), d__2 = abs(btmp[1]), d__1 = f2cmax(d__1,d__2), d__2
= abs(btmp[2]), d__1 = f2cmax(d__1,d__2), d__2 = abs(btmp[3]);
*scale = .125 / f2cmax(d__1,d__2);
btmp[0] *= *scale;
btmp[1] *= *scale;
btmp[2] *= *scale;
btmp[3] *= *scale;
}
for (i__ = 1; i__ <= 4; ++i__) {
k = 5 - i__;
temp = 1. / t16[k + (k << 2) - 5];
tmp[k - 1] = btmp[k - 1] * temp;
for (j = k + 1; j <= 4; ++j) {
tmp[k - 1] -= temp * t16[k + (j << 2) - 5] * tmp[j - 1];
/* L110: */
}
/* L120: */
}
for (i__ = 1; i__ <= 3; ++i__) {
if (jpiv[4 - i__ - 1] != 4 - i__) {
temp = tmp[4 - i__ - 1];
tmp[4 - i__ - 1] = tmp[jpiv[4 - i__ - 1] - 1];
tmp[jpiv[4 - i__ - 1] - 1] = temp;
}
/* L130: */
}
x[x_dim1 + 1] = tmp[0];
x[x_dim1 + 2] = tmp[1];
x[(x_dim1 << 1) + 1] = tmp[2];
x[(x_dim1 << 1) + 2] = tmp[3];
/* Computing MAX */
d__1 = abs(tmp[0]) + abs(tmp[2]), d__2 = abs(tmp[1]) + abs(tmp[3]);
*xnorm = f2cmax(d__1,d__2);
return;
/* End of DLASY2 */
} /* dlasy2_ */