PyObject* r2k(PyObject *self, PyObject *args)
{
Py_complex alpha;
PyArrayObject* a;
PyArrayObject* b;
double beta;
PyArrayObject* c;
if (!PyArg_ParseTuple(args, "DOOdO", &alpha, &a, &b, &beta, &c))
return NULL;
int n = PyArray_DIMS(a)[0];
int k = PyArray_DIMS(a)[1];
for (int d = 2; d < PyArray_NDIM(a); d++)
k *= PyArray_DIMS(a)[d];
int ldc = PyArray_STRIDES(c)[0] / PyArray_STRIDES(c)[1];
if (PyArray_DESCR(a)->type_num == NPY_DOUBLE)
dsyr2k_("u", "t", &n, &k,
(double*)(&alpha), DOUBLEP(a), &k,
DOUBLEP(b), &k, &beta,
DOUBLEP(c), &ldc);
else
zher2k_("u", "c", &n, &k,
(void*)(&alpha), (void*)COMPLEXP(a), &k,
(void*)COMPLEXP(b), &k, &beta,
(void*)COMPLEXP(c), &ldc);
Py_RETURN_NONE;
}
int
f2c_zher2k(char* uplo, char* trans, integer* N, integer* K,
doublecomplex* alpha,
doublecomplex* A, integer* lda,
doublecomplex* B, integer* ldb,
doublereal* beta,
doublecomplex* C, integer* ldc)
{
zher2k_(uplo, trans, N, K,
alpha, A, lda, B, ldb, beta, C, ldc);
return 0;
}
void
zher2k(char uplo, char transa, int n, int k, doublecomplex *alpha, doublecomplex *a, int lda, doublecomplex *b, int ldb, double beta, doublecomplex *c, int ldc)
{
zher2k_( &uplo, &transa, &n, &k, alpha, a, &lda, b, &ldb, &beta, c, &ldc);
}
/* Subroutine */ int ztimb3_(char *line, integer *nm, integer *mval, integer *
nn, integer *nval, integer *nk, integer *kval, integer *nlda, integer
*ldaval, doublereal *timmin, doublecomplex *a, doublecomplex *b,
doublecomplex *c__, doublereal *reslts, integer *ldr1, integer *ldr2,
integer *nout, ftnlen line_len)
{
/* Initialized data */
static char names[6*9] = "ZGEMM " "ZHEMM " "ZSYMM " "ZHERK " "ZHER2K"
"ZSYRK " "ZSYR2K" "ZTRMM " "ZTRSM ";
static char trans[1*3] = "N" "T" "C";
static char sides[1*2] = "L" "R";
static char uplos[1*2] = "U" "L";
/* Format strings */
static char fmt_9999[] = "(1x,a6,\002 timing run not attempted\002,/)";
static char fmt_9998[] = "(/\002 *** Speed of \002,a6,\002 in megaflops "
"***\002)";
static char fmt_9997[] = "(5x,\002with LDA = \002,i5)";
static char fmt_9996[] = "(5x,\002line \002,i2,\002 with LDA = \002,i5)";
static char fmt_9995[] = "(/1x,\002ZGEMM with TRANSA = '\002,a1,\002', "
"TRANSB = '\002,a1,\002'\002)";
static char fmt_9994[] = "(/1x,\002K = \002,i4,/)";
static char fmt_9993[] = "(/1x,a6,\002 with SIDE = '\002,a1,\002', UPLO "
"= '\002,a1,\002'\002,/)";
static char fmt_9992[] = "(/1x,a6,\002 with UPLO = '\002,a1,\002', TRANS"
" = '\002,a1,\002'\002,/)";
static char fmt_9991[] = "(/1x,a6,\002 with SIDE = '\002,a1,\002', UPLO "
"= '\002,a1,\002',\002,\002 TRANS = '\002,a1,\002'\002,/)";
static char fmt_9990[] = "(/////)";
/* System generated locals */
integer reslts_dim1, reslts_dim2, reslts_offset, i__1, i__2, i__3, i__4;
/* Builtin functions
Subroutine */ int s_copy(char *, char *, ftnlen, ftnlen);
integer s_wsfe(cilist *), do_fio(integer *, char *, ftnlen), e_wsfe(void),
s_cmp(char *, char *, ftnlen, ftnlen);
/* Local variables */
static integer ilda;
static char side[1];
static integer imat, info;
static char path[3];
static doublereal time;
static integer isub;
static char uplo[1];
static integer i__, k, m, n;
static char cname[6];
static integer iside;
extern logical lsame_(char *, char *);
extern /* Subroutine */ int zgemm_(char *, char *, integer *, integer *,
integer *, doublecomplex *, doublecomplex *, integer *,
doublecomplex *, integer *, doublecomplex *, doublecomplex *,
integer *), zhemm_(char *, char *, integer *,
integer *, doublecomplex *, doublecomplex *, integer *,
doublecomplex *, integer *, doublecomplex *, doublecomplex *,
integer *), zherk_(char *, char *, integer *,
integer *, doublereal *, doublecomplex *, integer *, doublereal *,
doublecomplex *, integer *);
static integer iuplo;
static doublereal s1, s2;
extern /* Subroutine */ int ztrmm_(char *, char *, char *, char *,
integer *, integer *, doublecomplex *, doublecomplex *, integer *,
doublecomplex *, integer *),
zsymm_(char *, char *, integer *, integer *, doublecomplex *,
doublecomplex *, integer *, doublecomplex *, integer *,
doublecomplex *, doublecomplex *, integer *),
ztrsm_(char *, char *, char *, char *, integer *, integer *,
doublecomplex *, doublecomplex *, integer *, doublecomplex *,
integer *), zsyrk_(char *, char *,
integer *, integer *, doublecomplex *, doublecomplex *, integer *
, doublecomplex *, doublecomplex *, integer *);
extern doublereal dopbl3_(char *, integer *, integer *, integer *)
;
extern /* Subroutine */ int zher2k_(char *, char *, integer *, integer *,
doublecomplex *, doublecomplex *, integer *, doublecomplex *,
integer *, doublereal *, doublecomplex *, integer *);
static integer ic, ik, im, in;
extern doublereal dsecnd_(void);
extern /* Subroutine */ int zsyr2k_(char *, char *, integer *, integer *,
doublecomplex *, doublecomplex *, integer *, doublecomplex *,
integer *, doublecomplex *, doublecomplex *, integer *), atimck_(integer *, char *, integer *, integer *, integer
*, integer *, integer *, integer *, ftnlen);
extern doublereal dmflop_(doublereal *, doublereal *, integer *);
extern /* Subroutine */ int atimin_(char *, char *, integer *, char *,
logical *, integer *, integer *, ftnlen, ftnlen, ftnlen), dprtbl_(
char *, char *, integer *, integer *, integer *, integer *,
integer *, doublereal *, integer *, integer *, integer *, ftnlen,
ftnlen);
static char transa[1], transb[1];
static doublereal untime;
static logical timsub[9];
extern /* Subroutine */ int ztimmg_(integer *, integer *, integer *,
doublecomplex *, integer *, integer *, integer *);
static integer lda, icl, ita, itb;
static doublereal ops;
/* Fortran I/O blocks */
static cilist io___9 = { 0, 0, 0, fmt_9999, 0 };
static cilist io___11 = { 0, 0, 0, fmt_9998, 0 };
static cilist io___12 = { 0, 0, 0, fmt_9997, 0 };
static cilist io___14 = { 0, 0, 0, fmt_9996, 0 };
static cilist io___34 = { 0, 0, 0, fmt_9995, 0 };
static cilist io___35 = { 0, 0, 0, fmt_9994, 0 };
static cilist io___41 = { 0, 0, 0, fmt_9993, 0 };
static cilist io___42 = { 0, 0, 0, fmt_9993, 0 };
static cilist io___43 = { 0, 0, 0, fmt_9992, 0 };
static cilist io___44 = { 0, 0, 0, fmt_9992, 0 };
static cilist io___45 = { 0, 0, 0, fmt_9992, 0 };
static cilist io___46 = { 0, 0, 0, fmt_9992, 0 };
static cilist io___47 = { 0, 0, 0, fmt_9991, 0 };
static cilist io___48 = { 0, 0, 0, fmt_9991, 0 };
static cilist io___49 = { 0, 0, 0, fmt_9990, 0 };
#define names_ref(a_0,a_1) &names[(a_1)*6 + a_0 - 6]
#define reslts_ref(a_1,a_2,a_3) reslts[((a_3)*reslts_dim2 + (a_2))*\
reslts_dim1 + a_1]
/* -- LAPACK timing routine (version 3.0) --
Univ. of Tennessee, Univ. of California Berkeley, NAG Ltd.,
Courant Institute, Argonne National Lab, and Rice University
March 31, 1993
Purpose
=======
ZTIMB3 times the Level 3 BLAS routines.
Arguments
=========
LINE (input) CHARACTER*80
The input line that requested this routine. The first six
characters contain either the name of a subroutine or a
generic path name. The remaining characters may be used to
specify the individual routines to be timed. See ATIMIN for
a full description of the format of the input line.
NM (input) INTEGER
The number of values of M contained in the vector MVAL.
MVAL (input) INTEGER array, dimension (NM)
The values of the matrix row dimension M.
NN (input) INTEGER
The number of values of N contained in the vector NVAL.
NVAL (input) INTEGER array, dimension (NN)
The values of the matrix column dimension N.
NK (input) INTEGER
The number of values of K contained in the vector KVAL.
KVAL (input) INTEGER array, dimension (NK)
The values of K. K is used as the intermediate matrix
dimension for ZGEMM (the product of an M x K matrix and a
K x N matrix) and as the dimension of the rank-K update in
ZHERK and ZSYRK.
NLDA (input) INTEGER
The number of values of LDA contained in the vector LDAVAL.
LDAVAL (input) INTEGER array, dimension (NLDA)
The values of the leading dimension of the array A.
TIMMIN (input) DOUBLE PRECISION
The minimum time a subroutine will be timed.
A (workspace) COMPLEX*16 array, dimension (LDAMAX*NMAX)
where LDAMAX and NMAX are the maximum values permitted
for LDA and N.
B (workspace) COMPLEX*16 array, dimension (LDAMAX*NMAX)
C (workspace) COMPLEX*16 array, dimension (LDAMAX*NMAX)
RESLTS (output) DOUBLE PRECISION array, dimension (LDR1,LDR2,NLDA)
The timing results for each subroutine over the relevant
values of M, N, K, and LDA.
LDR1 (input) INTEGER
The first dimension of RESLTS. LDR1 >= max(1,NM,NK).
LDR2 (input) INTEGER
The second dimension of RESLTS. LDR2 >= max(1,NN).
NOUT (input) INTEGER
The unit number for output.
=====================================================================
Parameter adjustments */
--mval;
--nval;
--kval;
--ldaval;
--a;
--b;
--c__;
reslts_dim1 = *ldr1;
reslts_dim2 = *ldr2;
reslts_offset = 1 + reslts_dim1 * (1 + reslts_dim2 * 1);
reslts -= reslts_offset;
/* Function Body
Extract the timing request from the input line. */
s_copy(path, "Zomplex precision", (ftnlen)1, (ftnlen)17);
s_copy(path + 1, "B3", (ftnlen)2, (ftnlen)2);
atimin_(path, line, &c__9, names, timsub, nout, &info, (ftnlen)3,
line_len, (ftnlen)6);
if (info != 0) {
goto L690;
}
/* Check that M <= LDA. */
s_copy(cname, line, (ftnlen)6, (ftnlen)6);
atimck_(&c__1, cname, nm, &mval[1], nlda, &ldaval[1], nout, &info, (
ftnlen)6);
if (info > 0) {
io___9.ciunit = *nout;
s_wsfe(&io___9);
do_fio(&c__1, cname, (ftnlen)6);
e_wsfe();
goto L690;
}
/* Time each routine. */
for (isub = 1; isub <= 9; ++isub) {
if (! timsub[isub - 1]) {
goto L680;
}
/* Print header. */
s_copy(cname, names_ref(0, isub), (ftnlen)6, (ftnlen)6);
io___11.ciunit = *nout;
s_wsfe(&io___11);
do_fio(&c__1, cname, (ftnlen)6);
e_wsfe();
if (*nlda == 1) {
io___12.ciunit = *nout;
s_wsfe(&io___12);
do_fio(&c__1, (char *)&ldaval[1], (ftnlen)sizeof(integer));
e_wsfe();
} else {
i__1 = *nlda;
for (i__ = 1; i__ <= i__1; ++i__) {
io___14.ciunit = *nout;
s_wsfe(&io___14);
do_fio(&c__1, (char *)&i__, (ftnlen)sizeof(integer));
do_fio(&c__1, (char *)&ldaval[i__], (ftnlen)sizeof(integer));
e_wsfe();
/* L10: */
}
}
/* Time ZGEMM */
if (s_cmp(cname, "ZGEMM ", (ftnlen)6, (ftnlen)6) == 0) {
for (ita = 1; ita <= 3; ++ita) {
*(unsigned char *)transa = *(unsigned char *)&trans[ita - 1];
for (itb = 1; itb <= 3; ++itb) {
*(unsigned char *)transb = *(unsigned char *)&trans[itb -
1];
i__1 = *nk;
for (ik = 1; ik <= i__1; ++ik) {
k = kval[ik];
i__2 = *nlda;
for (ilda = 1; ilda <= i__2; ++ilda) {
lda = ldaval[ilda];
i__3 = *nm;
for (im = 1; im <= i__3; ++im) {
m = mval[im];
i__4 = *nn;
for (in = 1; in <= i__4; ++in) {
n = nval[in];
if (*(unsigned char *)transa == 'N') {
ztimmg_(&c__1, &m, &k, &a[1], &lda, &
c__0, &c__0);
} else {
ztimmg_(&c__1, &k, &m, &a[1], &lda, &
c__0, &c__0);
}
if (*(unsigned char *)transb == 'N') {
ztimmg_(&c__0, &k, &n, &b[1], &lda, &
c__0, &c__0);
} else {
ztimmg_(&c__0, &n, &k, &b[1], &lda, &
c__0, &c__0);
}
ztimmg_(&c__1, &m, &n, &c__[1], &lda, &
c__0, &c__0);
ic = 0;
s1 = dsecnd_();
L20:
zgemm_(transa, transb, &m, &n, &k, &c_b1,
&a[1], &lda, &b[1], &lda, &c_b1, &
c__[1], &lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&c__1, &m, &n, &c__[1], &lda,
&c__0, &c__0);
goto L20;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L30:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&c__1, &m, &n, &c__[1], &lda,
&c__0, &c__0);
goto L30;
}
time = (time - untime) / (doublereal) ic;
ops = dopbl3_(cname, &m, &n, &k);
reslts_ref(im, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L40: */
}
/* L50: */
}
/* L60: */
}
if (ik == 1) {
io___34.ciunit = *nout;
s_wsfe(&io___34);
do_fio(&c__1, transa, (ftnlen)1);
do_fio(&c__1, transb, (ftnlen)1);
e_wsfe();
}
io___35.ciunit = *nout;
s_wsfe(&io___35);
do_fio(&c__1, (char *)&kval[ik], (ftnlen)sizeof(
integer));
e_wsfe();
dprtbl_("M", "N", nm, &mval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (
ftnlen)1, (ftnlen)1);
/* L70: */
}
/* L80: */
}
/* L90: */
}
/* Time ZHEMM */
} else if (s_cmp(cname, "ZHEMM ", (ftnlen)6, (ftnlen)6) == 0) {
for (iside = 1; iside <= 2; ++iside) {
*(unsigned char *)side = *(unsigned char *)&sides[iside - 1];
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo -
1];
if (lsame_(uplo, "U")) {
imat = 6;
} else {
imat = -6;
}
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nm;
for (im = 1; im <= i__2; ++im) {
m = mval[im];
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
if (iside == 1) {
ztimmg_(&imat, &m, &m, &a[1], &lda, &c__0,
&c__0);
ztimmg_(&c__0, &m, &n, &b[1], &lda, &c__0,
&c__0);
} else {
ztimmg_(&c__0, &m, &n, &b[1], &lda, &c__0,
&c__0);
ztimmg_(&imat, &n, &n, &a[1], &lda, &c__0,
&c__0);
}
ztimmg_(&c__1, &m, &n, &c__[1], &lda, &c__0, &
c__0);
ic = 0;
s1 = dsecnd_();
L100:
zhemm_(side, uplo, &m, &n, &c_b1, &a[1], &lda,
&b[1], &lda, &c_b1, &c__[1], &lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&c__1, &m, &n, &c__[1], &lda, &
c__0, &c__0);
goto L100;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L110:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&c__1, &m, &n, &c__[1], &lda, &
c__0, &c__0);
goto L110;
}
time = (time - untime) / (doublereal) ic;
i__4 = iside - 1;
ops = dopbl3_(cname, &m, &n, &i__4)
;
reslts_ref(im, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L120: */
}
/* L130: */
}
/* L140: */
}
io___41.ciunit = *nout;
s_wsfe(&io___41);
do_fio(&c__1, "ZHEMM ", (ftnlen)6);
do_fio(&c__1, side, (ftnlen)1);
do_fio(&c__1, uplo, (ftnlen)1);
e_wsfe();
dprtbl_("M", "N", nm, &mval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (ftnlen)
1, (ftnlen)1);
/* L150: */
}
/* L160: */
}
/* Time ZSYMM */
} else if (s_cmp(cname, "ZSYMM ", (ftnlen)6, (ftnlen)6) == 0) {
for (iside = 1; iside <= 2; ++iside) {
*(unsigned char *)side = *(unsigned char *)&sides[iside - 1];
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo -
1];
if (lsame_(uplo, "U")) {
imat = 8;
} else {
imat = -8;
}
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nm;
for (im = 1; im <= i__2; ++im) {
m = mval[im];
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
if (iside == 1) {
ztimmg_(&imat, &m, &m, &a[1], &lda, &c__0,
&c__0);
ztimmg_(&c__0, &m, &n, &b[1], &lda, &c__0,
&c__0);
} else {
ztimmg_(&c__0, &m, &n, &b[1], &lda, &c__0,
&c__0);
ztimmg_(&imat, &n, &n, &a[1], &lda, &c__0,
&c__0);
}
ztimmg_(&c__1, &m, &n, &c__[1], &lda, &c__0, &
c__0);
ic = 0;
s1 = dsecnd_();
L170:
zsymm_(side, uplo, &m, &n, &c_b1, &a[1], &lda,
&b[1], &lda, &c_b1, &c__[1], &lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&c__1, &m, &n, &c__[1], &lda, &
c__0, &c__0);
goto L170;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L180:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&c__1, &m, &n, &c__[1], &lda, &
c__0, &c__0);
goto L180;
}
time = (time - untime) / (doublereal) ic;
i__4 = iside - 1;
ops = dopbl3_(cname, &m, &n, &i__4)
;
reslts_ref(im, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L190: */
}
/* L200: */
}
/* L210: */
}
io___42.ciunit = *nout;
s_wsfe(&io___42);
do_fio(&c__1, "ZSYMM ", (ftnlen)6);
do_fio(&c__1, side, (ftnlen)1);
do_fio(&c__1, uplo, (ftnlen)1);
e_wsfe();
dprtbl_("M", "N", nm, &mval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (ftnlen)
1, (ftnlen)1);
/* L220: */
}
/* L230: */
}
/* Time ZHERK */
} else if (s_cmp(cname, "ZHERK ", (ftnlen)6, (ftnlen)6) == 0) {
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo - 1];
if (lsame_(uplo, "U")) {
imat = 6;
} else {
imat = -6;
}
for (ita = 1; ita <= 3; ++ita) {
*(unsigned char *)transa = *(unsigned char *)&trans[ita -
1];
if (*(unsigned char *)transa != 'T') {
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nk;
for (ik = 1; ik <= i__2; ++ik) {
k = kval[ik];
if (*(unsigned char *)transa == 'N') {
ztimmg_(&c__1, &n, &k, &a[1], &lda, &c__0,
&c__0);
} else {
ztimmg_(&c__1, &k, &n, &a[1], &lda, &c__0,
&c__0);
}
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
ztimmg_(&imat, &n, &n, &c__[1], &lda, &
c__0, &c__0);
ic = 0;
s1 = dsecnd_();
L240:
zherk_(uplo, transa, &n, &k, &c_b156, &a[
1], &lda, &c_b156, &c__[1], &lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L240;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L250:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L250;
}
time = (time - untime) / (doublereal) ic;
ops = dopbl3_(cname, &n, &n, &k);
reslts_ref(ik, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L260: */
}
/* L270: */
}
/* L280: */
}
io___43.ciunit = *nout;
s_wsfe(&io___43);
do_fio(&c__1, cname, (ftnlen)6);
do_fio(&c__1, uplo, (ftnlen)1);
do_fio(&c__1, transa, (ftnlen)1);
e_wsfe();
dprtbl_("K", "N", nk, &kval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (
ftnlen)1, (ftnlen)1);
}
/* L290: */
}
/* L300: */
}
/* Time ZHER2K */
} else if (s_cmp(cname, "ZHER2K", (ftnlen)6, (ftnlen)6) == 0) {
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo - 1];
if (lsame_(uplo, "U")) {
imat = 6;
} else {
imat = -6;
}
for (itb = 1; itb <= 3; ++itb) {
*(unsigned char *)transb = *(unsigned char *)&trans[itb -
1];
if (*(unsigned char *)transb != 'T') {
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nk;
for (ik = 1; ik <= i__2; ++ik) {
k = kval[ik];
if (*(unsigned char *)transb == 'N') {
ztimmg_(&c__1, &n, &k, &a[1], &lda, &c__0,
&c__0);
ztimmg_(&c__0, &n, &k, &b[1], &lda, &c__0,
&c__0);
} else {
ztimmg_(&c__1, &k, &n, &a[1], &lda, &c__0,
&c__0);
ztimmg_(&c__0, &k, &n, &b[1], &lda, &c__0,
&c__0);
}
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
ztimmg_(&imat, &n, &n, &c__[1], &lda, &
c__0, &c__0);
ic = 0;
s1 = dsecnd_();
L310:
zher2k_(uplo, transb, &n, &k, &c_b1, &a[1]
, &lda, &b[1], &lda, &c_b156, &
c__[1], &lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L310;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L320:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L320;
}
time = (time - untime) / (doublereal) ic;
ops = dopbl3_(cname, &n, &n, &k);
reslts_ref(ik, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L330: */
}
/* L340: */
}
/* L350: */
}
io___44.ciunit = *nout;
s_wsfe(&io___44);
do_fio(&c__1, cname, (ftnlen)6);
do_fio(&c__1, uplo, (ftnlen)1);
do_fio(&c__1, transb, (ftnlen)1);
e_wsfe();
dprtbl_("K", "N", nk, &kval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (
ftnlen)1, (ftnlen)1);
}
/* L360: */
}
/* L370: */
}
/* Time ZSYRK */
} else if (s_cmp(cname, "ZSYRK ", (ftnlen)6, (ftnlen)6) == 0) {
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo - 1];
if (lsame_(uplo, "U")) {
imat = 8;
} else {
imat = -8;
}
for (ita = 1; ita <= 3; ++ita) {
*(unsigned char *)transa = *(unsigned char *)&trans[ita -
1];
if (*(unsigned char *)transa != 'C') {
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nk;
for (ik = 1; ik <= i__2; ++ik) {
k = kval[ik];
if (*(unsigned char *)transa == 'N') {
ztimmg_(&c__1, &n, &k, &a[1], &lda, &c__0,
&c__0);
} else {
ztimmg_(&c__1, &k, &n, &a[1], &lda, &c__0,
&c__0);
}
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
ztimmg_(&imat, &n, &n, &c__[1], &lda, &
c__0, &c__0);
ic = 0;
s1 = dsecnd_();
L380:
zsyrk_(uplo, transa, &n, &k, &c_b1, &a[1],
&lda, &c_b1, &c__[1], &lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L380;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L390:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L390;
}
time = (time - untime) / (doublereal) ic;
ops = dopbl3_(cname, &n, &n, &k);
reslts_ref(ik, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L400: */
}
/* L410: */
}
/* L420: */
}
io___45.ciunit = *nout;
s_wsfe(&io___45);
do_fio(&c__1, cname, (ftnlen)6);
do_fio(&c__1, uplo, (ftnlen)1);
do_fio(&c__1, transa, (ftnlen)1);
e_wsfe();
dprtbl_("K", "N", nk, &kval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (
ftnlen)1, (ftnlen)1);
}
/* L430: */
}
/* L440: */
}
/* Time ZSYR2K */
} else if (s_cmp(cname, "ZSYR2K", (ftnlen)6, (ftnlen)6) == 0) {
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo - 1];
if (lsame_(uplo, "U")) {
imat = 8;
} else {
imat = -8;
}
for (itb = 1; itb <= 3; ++itb) {
*(unsigned char *)transb = *(unsigned char *)&trans[itb -
1];
if (*(unsigned char *)transb != 'C') {
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nk;
for (ik = 1; ik <= i__2; ++ik) {
k = kval[ik];
if (*(unsigned char *)transb == 'N') {
ztimmg_(&c__1, &n, &k, &a[1], &lda, &c__0,
&c__0);
ztimmg_(&c__0, &n, &k, &b[1], &lda, &c__0,
&c__0);
} else {
ztimmg_(&c__1, &k, &n, &a[1], &lda, &c__0,
&c__0);
ztimmg_(&c__0, &k, &n, &b[1], &lda, &c__0,
&c__0);
}
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
ztimmg_(&imat, &n, &n, &c__[1], &lda, &
c__0, &c__0);
ic = 0;
s1 = dsecnd_();
L450:
zsyr2k_(uplo, transb, &n, &k, &c_b1, &a[1]
, &lda, &b[1], &lda, &c_b1, &c__[
1], &lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L450;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L460:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&imat, &n, &n, &c__[1], &lda,
&c__0, &c__0);
goto L460;
}
time = (time - untime) / (doublereal) ic;
ops = dopbl3_(cname, &n, &n, &k);
reslts_ref(ik, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L470: */
}
/* L480: */
}
/* L490: */
}
io___46.ciunit = *nout;
s_wsfe(&io___46);
do_fio(&c__1, cname, (ftnlen)6);
do_fio(&c__1, uplo, (ftnlen)1);
do_fio(&c__1, transb, (ftnlen)1);
e_wsfe();
dprtbl_("K", "N", nk, &kval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (
ftnlen)1, (ftnlen)1);
}
/* L500: */
}
/* L510: */
}
/* Time ZTRMM */
} else if (s_cmp(cname, "ZTRMM ", (ftnlen)6, (ftnlen)6) == 0) {
for (iside = 1; iside <= 2; ++iside) {
*(unsigned char *)side = *(unsigned char *)&sides[iside - 1];
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo -
1];
if (lsame_(uplo, "U")) {
imat = 11;
} else {
imat = -11;
}
for (ita = 1; ita <= 3; ++ita) {
*(unsigned char *)transa = *(unsigned char *)&trans[
ita - 1];
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nm;
for (im = 1; im <= i__2; ++im) {
m = mval[im];
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
if (iside == 1) {
ztimmg_(&imat, &m, &m, &a[1], &lda, &
c__0, &c__0);
} else {
ztimmg_(&imat, &n, &n, &a[1], &lda, &
c__0, &c__0);
}
ztimmg_(&c__0, &m, &n, &b[1], &lda, &c__0,
&c__0);
ic = 0;
s1 = dsecnd_();
L520:
ztrmm_(side, uplo, transa, "Non-unit", &m,
&n, &c_b1, &a[1], &lda, &b[1], &
lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&c__0, &m, &n, &b[1], &lda, &
c__0, &c__0);
goto L520;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L530:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&c__0, &m, &n, &b[1], &lda, &
c__0, &c__0);
goto L530;
}
time = (time - untime) / (doublereal) ic;
i__4 = iside - 1;
ops = dopbl3_(cname, &m, &n, &i__4);
reslts_ref(im, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L540: */
}
/* L550: */
}
/* L560: */
}
io___47.ciunit = *nout;
s_wsfe(&io___47);
do_fio(&c__1, cname, (ftnlen)6);
do_fio(&c__1, side, (ftnlen)1);
do_fio(&c__1, uplo, (ftnlen)1);
do_fio(&c__1, transa, (ftnlen)1);
e_wsfe();
dprtbl_("M", "N", nm, &mval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (
ftnlen)1, (ftnlen)1);
/* L570: */
}
/* L580: */
}
/* L590: */
}
/* Time ZTRSM */
} else if (s_cmp(cname, "ZTRSM ", (ftnlen)6, (ftnlen)6) == 0) {
for (iside = 1; iside <= 2; ++iside) {
*(unsigned char *)side = *(unsigned char *)&sides[iside - 1];
for (iuplo = 1; iuplo <= 2; ++iuplo) {
*(unsigned char *)uplo = *(unsigned char *)&uplos[iuplo -
1];
if (lsame_(uplo, "U")) {
imat = 11;
} else {
imat = -11;
}
for (ita = 1; ita <= 3; ++ita) {
*(unsigned char *)transa = *(unsigned char *)&trans[
ita - 1];
i__1 = *nlda;
for (ilda = 1; ilda <= i__1; ++ilda) {
lda = ldaval[ilda];
i__2 = *nm;
for (im = 1; im <= i__2; ++im) {
m = mval[im];
i__3 = *nn;
for (in = 1; in <= i__3; ++in) {
n = nval[in];
if (iside == 1) {
ztimmg_(&imat, &m, &m, &a[1], &lda, &
c__0, &c__0);
} else {
ztimmg_(&imat, &n, &n, &a[1], &lda, &
c__0, &c__0);
}
ztimmg_(&c__0, &m, &n, &b[1], &lda, &c__0,
&c__0);
ic = 0;
s1 = dsecnd_();
L600:
ztrsm_(side, uplo, transa, "Non-unit", &m,
&n, &c_b1, &a[1], &lda, &b[1], &
lda);
s2 = dsecnd_();
time = s2 - s1;
++ic;
if (time < *timmin) {
ztimmg_(&c__0, &m, &n, &b[1], &lda, &
c__0, &c__0);
goto L600;
}
/* Subtract the time used in ZTIMMG. */
icl = 1;
s1 = dsecnd_();
L610:
s2 = dsecnd_();
untime = s2 - s1;
++icl;
if (icl <= ic) {
ztimmg_(&c__0, &m, &n, &b[1], &lda, &
c__0, &c__0);
goto L610;
}
time = (time - untime) / (doublereal) ic;
i__4 = iside - 1;
ops = dopbl3_(cname, &m, &n, &i__4);
reslts_ref(im, in, ilda) = dmflop_(&ops, &
time, &c__0);
/* L620: */
}
/* L630: */
}
/* L640: */
}
io___48.ciunit = *nout;
s_wsfe(&io___48);
do_fio(&c__1, cname, (ftnlen)6);
do_fio(&c__1, side, (ftnlen)1);
do_fio(&c__1, uplo, (ftnlen)1);
do_fio(&c__1, transa, (ftnlen)1);
e_wsfe();
dprtbl_("M", "N", nm, &mval[1], nn, &nval[1], nlda, &
reslts[reslts_offset], ldr1, ldr2, nout, (
ftnlen)1, (ftnlen)1);
/* L650: */
}
/* L660: */
}
/* L670: */
}
}
io___49.ciunit = *nout;
s_wsfe(&io___49);
e_wsfe();
L680:
;
}
L690:
return 0;
/* End of ZTIMB3 */
} /* ztimb3_ */
int main( int argc, char** argv )
{
obj_t a, b, c;
obj_t x, y;
obj_t alpha, beta;
dim_t m;
num_t dt_a, dt_b, dt_c;
num_t dt_alpha, dt_beta;
int ii;
#ifdef NBLIS
bli_init();
#endif
m = 4000;
dt_a = BLIS_DOUBLE;
dt_b = BLIS_DOUBLE;
dt_c = BLIS_DOUBLE;
dt_alpha = BLIS_DOUBLE;
dt_beta = BLIS_DOUBLE;
{
#ifdef NBLIS
bli_obj_create( dt_alpha, 1, 1, 0, 0, &alpha );
bli_obj_create( dt_beta, 1, 1, 0, 0, &beta );
bli_obj_create( dt_a, m, 1, 0, 0, &x );
bli_obj_create( dt_a, m, 1, 0, 0, &y );
bli_obj_create( dt_a, m, m, 0, 0, &a );
bli_obj_create( dt_b, m, m, 0, 0, &b );
bli_obj_create( dt_c, m, m, 0, 0, &c );
bli_randm( &a );
bli_randm( &b );
bli_randm( &c );
bli_setsc( (2.0/1.0), 0.0, &alpha );
bli_setsc( -(1.0/1.0), 0.0, &beta );
#endif
#ifdef NBLAS
x.buffer = malloc( m * 1 * sizeof( double ) );
y.buffer = malloc( m * 1 * sizeof( double ) );
alpha.buffer = malloc( 1 * sizeof( double ) );
beta.buffer = malloc( 1 * sizeof( double ) );
a.buffer = malloc( m * m * sizeof( double ) );
a.m = m;
a.n = m;
a.cs = m;
b.buffer = malloc( m * m * sizeof( double ) );
b.m = m;
b.n = m;
b.cs = m;
c.buffer = malloc( m * m * sizeof( double ) );
c.m = m;
c.n = m;
c.cs = m;
*((double*)alpha.buffer) = 2.0;
*((double*)beta.buffer) = -1.0;
#endif
#ifdef NBLIS
#if NBLIS >= 1
for ( ii = 0; ii < 2000000000; ++ii )
{
bli_gemm( &BLIS_ONE,
&a,
&b,
&BLIS_ONE,
&c );
}
#endif
#if NBLIS >= 2
{
bli_hemm( BLIS_LEFT,
&BLIS_ONE,
&a,
&b,
&BLIS_ONE,
&c );
}
#endif
#if NBLIS >= 3
{
bli_herk( &BLIS_ONE,
&a,
&BLIS_ONE,
&c );
}
#endif
#if NBLIS >= 4
{
bli_her2k( &BLIS_ONE,
&a,
&b,
&BLIS_ONE,
&c );
}
#endif
#if NBLIS >= 5
{
bli_trmm( BLIS_LEFT,
&BLIS_ONE,
&a,
&c );
}
#endif
#if NBLIS >= 6
{
bli_trsm( BLIS_LEFT,
&BLIS_ONE,
&a,
&c );
}
#endif
#endif
#ifdef NBLAS
#if NBLAS >= 1
for ( ii = 0; ii < 2000000000; ++ii )
{
f77_char transa = 'N';
f77_char transb = 'N';
f77_int mm = bli_obj_length( c );
f77_int kk = bli_obj_width_after_trans( a );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldb = bli_obj_col_stride( b );
f77_int ldc = bli_obj_col_stride( c );
double* alphap = bli_obj_buffer( alpha );
double* ap = bli_obj_buffer( a );
double* bp = bli_obj_buffer( b );
double* betap = bli_obj_buffer( beta );
double* cp = bli_obj_buffer( c );
dgemm_( &transa,
&transb,
&mm,
&nn,
&kk,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 2
{
f77_char side = 'L';
f77_char uplo = 'L';
f77_int mm = bli_obj_length( c );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldb = bli_obj_col_stride( b );
f77_int ldc = bli_obj_col_stride( c );
double* alphap = bli_obj_buffer( alpha );
double* ap = bli_obj_buffer( a );
double* bp = bli_obj_buffer( b );
double* betap = bli_obj_buffer( beta );
double* cp = bli_obj_buffer( c );
dsymm_( &side,
&uplo,
&mm,
&nn,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 3
{
f77_char uplo = 'L';
f77_char trans = 'N';
f77_int mm = bli_obj_length( c );
f77_int kk = bli_obj_width( a );
f77_int lda = bli_obj_col_stride( a );
f77_int ldc = bli_obj_col_stride( c );
double* alphap = bli_obj_buffer( alpha );
double* ap = bli_obj_buffer( a );
double* betap = bli_obj_buffer( beta );
double* cp = bli_obj_buffer( c );
dsyrk_( &uplo,
&trans,
&mm,
&kk,
alphap,
ap, &lda,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 4
{
f77_char uplo = 'L';
f77_char trans = 'N';
f77_int mm = bli_obj_length( c );
f77_int kk = bli_obj_width( a );
f77_int lda = bli_obj_col_stride( a );
f77_int ldb = bli_obj_col_stride( b );
f77_int ldc = bli_obj_col_stride( c );
double* alphap = bli_obj_buffer( alpha );
double* ap = bli_obj_buffer( a );
double* bp = bli_obj_buffer( b );
double* betap = bli_obj_buffer( beta );
double* cp = bli_obj_buffer( c );
dsyr2k_( &uplo,
&trans,
&mm,
&kk,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 5
{
f77_char side = 'L';
f77_char uplo = 'L';
f77_char trans = 'N';
f77_char diag = 'N';
f77_int mm = bli_obj_length( c );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldc = bli_obj_col_stride( c );
double* alphap = bli_obj_buffer( alpha );
double* ap = bli_obj_buffer( a );
double* cp = bli_obj_buffer( c );
dtrmm_( &side,
&uplo,
&trans,
&diag,
&mm,
&nn,
alphap,
ap, &lda,
cp, &ldc );
}
#endif
#if NBLAS >= 6
{
f77_char side = 'L';
f77_char uplo = 'L';
f77_char trans = 'N';
f77_char diag = 'N';
f77_int mm = bli_obj_length( c );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldc = bli_obj_col_stride( c );
double* alphap = bli_obj_buffer( alpha );
double* ap = bli_obj_buffer( a );
double* cp = bli_obj_buffer( c );
dtrsm_( &side,
&uplo,
&trans,
&diag,
&mm,
&nn,
alphap,
ap, &lda,
cp, &ldc );
}
#endif
#if NBLAS >= 7
{
f77_char transa = 'N';
f77_char transb = 'N';
f77_int mm = bli_obj_length( c );
f77_int kk = bli_obj_width_after_trans( a );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldb = bli_obj_col_stride( b );
f77_int ldc = bli_obj_col_stride( c );
dcomplex* alphap = bli_obj_buffer( alpha );
dcomplex* ap = bli_obj_buffer( a );
dcomplex* bp = bli_obj_buffer( b );
dcomplex* betap = bli_obj_buffer( beta );
dcomplex* cp = bli_obj_buffer( c );
zgemm_( &transa,
&transb,
&mm,
&nn,
&kk,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 8
{
f77_char side = 'L';
f77_char uplo = 'L';
f77_int mm = bli_obj_length( c );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldb = bli_obj_col_stride( b );
f77_int ldc = bli_obj_col_stride( c );
dcomplex* alphap = bli_obj_buffer( alpha );
dcomplex* ap = bli_obj_buffer( a );
dcomplex* bp = bli_obj_buffer( b );
dcomplex* betap = bli_obj_buffer( beta );
dcomplex* cp = bli_obj_buffer( c );
zhemm_( &side,
&uplo,
&mm,
&nn,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 9
{
f77_char uplo = 'L';
f77_char trans = 'N';
f77_int mm = bli_obj_length( c );
f77_int kk = bli_obj_width( a );
f77_int lda = bli_obj_col_stride( a );
f77_int ldc = bli_obj_col_stride( c );
double* alphap = bli_obj_buffer( alpha );
dcomplex* ap = bli_obj_buffer( a );
double* betap = bli_obj_buffer( beta );
dcomplex* cp = bli_obj_buffer( c );
zherk_( &uplo,
&trans,
&mm,
&kk,
alphap,
ap, &lda,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 10
{
f77_char uplo = 'L';
f77_char trans = 'N';
f77_int mm = bli_obj_length( c );
f77_int kk = bli_obj_width( a );
f77_int lda = bli_obj_col_stride( a );
f77_int ldb = bli_obj_col_stride( b );
f77_int ldc = bli_obj_col_stride( c );
dcomplex* alphap = bli_obj_buffer( alpha );
dcomplex* ap = bli_obj_buffer( a );
dcomplex* bp = bli_obj_buffer( b );
double* betap = bli_obj_buffer( beta );
dcomplex* cp = bli_obj_buffer( c );
zher2k_( &uplo,
&trans,
&mm,
&kk,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
#endif
#if NBLAS >= 11
{
f77_char side = 'L';
f77_char uplo = 'L';
f77_char trans = 'N';
f77_char diag = 'N';
f77_int mm = bli_obj_length( c );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldc = bli_obj_col_stride( c );
dcomplex* alphap = bli_obj_buffer( alpha );
dcomplex* ap = bli_obj_buffer( a );
dcomplex* cp = bli_obj_buffer( c );
ztrmm_( &side,
&uplo,
&trans,
&diag,
&mm,
&nn,
alphap,
ap, &lda,
cp, &ldc );
}
#endif
#if NBLAS >= 12
{
f77_char side = 'L';
f77_char uplo = 'L';
f77_char trans = 'N';
f77_char diag = 'N';
f77_int mm = bli_obj_length( c );
f77_int nn = bli_obj_width( c );
f77_int lda = bli_obj_col_stride( a );
f77_int ldc = bli_obj_col_stride( c );
dcomplex* alphap = bli_obj_buffer( alpha );
dcomplex* ap = bli_obj_buffer( a );
dcomplex* cp = bli_obj_buffer( c );
ztrsm_( &side,
&uplo,
&trans,
&diag,
&mm,
&nn,
alphap,
ap, &lda,
cp, &ldc );
}
#endif
#endif
#ifdef NBLIS
bli_obj_free( &x );
bli_obj_free( &y );
bli_obj_free( &alpha );
bli_obj_free( &beta );
bli_obj_free( &a );
bli_obj_free( &b );
bli_obj_free( &c );
#endif
#ifdef NBLAS
free( x.buffer );
free( y.buffer );
free( alpha.buffer );
free( beta.buffer );
free( a.buffer );
free( b.buffer );
free( c.buffer );
#endif
}
#ifdef NBLIS
bli_finalize();
#endif
return 0;
}
int zhegst_(int *itype, char *uplo, int *n,
doublecomplex *a, int *lda, doublecomplex *b, int *ldb,
int *info)
{
/* System generated locals */
int a_dim1, a_offset, b_dim1, b_offset, i__1, i__2, i__3;
doublecomplex z__1;
/* Local variables */
int k, kb, nb;
extern int lsame_(char *, char *);
extern int zhemm_(char *, char *, int *, int *,
doublecomplex *, doublecomplex *, int *, doublecomplex *,
int *, doublecomplex *, doublecomplex *, int *);
int upper;
extern int ztrmm_(char *, char *, char *, char *,
int *, int *, doublecomplex *, doublecomplex *, int *,
doublecomplex *, int *),
ztrsm_(char *, char *, char *, char *, int *, int *,
doublecomplex *, doublecomplex *, int *, doublecomplex *,
int *), zhegs2_(int *,
char *, int *, doublecomplex *, int *, doublecomplex *,
int *, int *), zher2k_(char *, char *, int *,
int *, doublecomplex *, doublecomplex *, int *,
doublecomplex *, int *, double *, doublecomplex *,
int *), xerbla_(char *, int *);
extern int ilaenv_(int *, char *, char *, int *, int *,
int *, int *);
/* -- LAPACK routine (version 3.2) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* ZHEGST reduces a complex Hermitian-definite generalized */
/* eigenproblem to standard form. */
/* If ITYPE = 1, the problem is A*x = lambda*B*x, */
/* and A is overwritten by inv(U**H)*A*inv(U) or inv(L)*A*inv(L**H) */
/* If ITYPE = 2 or 3, the problem is A*B*x = lambda*x or */
/* B*A*x = lambda*x, and A is overwritten by U*A*U**H or L**H*A*L. */
/* B must have been previously factorized as U**H*U or L*L**H by ZPOTRF. */
/* Arguments */
/* ========= */
/* ITYPE (input) INTEGER */
/* = 1: compute inv(U**H)*A*inv(U) or inv(L)*A*inv(L**H); */
/* = 2 or 3: compute U*A*U**H or L**H*A*L. */
/* UPLO (input) CHARACTER*1 */
/* = 'U': Upper triangle of A is stored and B is factored as */
/* U**H*U; */
/* = 'L': Lower triangle of A is stored and B is factored as */
/* L*L**H. */
/* N (input) INTEGER */
/* The order of the matrices A and B. N >= 0. */
/* A (input/output) COMPLEX*16 array, dimension (LDA,N) */
/* On entry, the Hermitian matrix A. If UPLO = 'U', the leading */
/* N-by-N upper triangular part of A contains the upper */
/* triangular part of the matrix A, and the strictly lower */
/* triangular part of A is not referenced. If UPLO = 'L', the */
/* leading N-by-N lower triangular part of A contains the lower */
/* triangular part of the matrix A, and the strictly upper */
/* triangular part of A is not referenced. */
/* On exit, if INFO = 0, the transformed matrix, stored in the */
/* same format as A. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= MAX(1,N). */
/* B (input) COMPLEX*16 array, dimension (LDB,N) */
/* The triangular factor from the Cholesky factorization of B, */
/* as returned by ZPOTRF. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= MAX(1,N). */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
/* Function Body */
*info = 0;
upper = lsame_(uplo, "U");
if (*itype < 1 || *itype > 3) {
*info = -1;
} else if (! upper && ! lsame_(uplo, "L")) {
*info = -2;
} else if (*n < 0) {
*info = -3;
} else if (*lda < MAX(1,*n)) {
*info = -5;
} else if (*ldb < MAX(1,*n)) {
*info = -7;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("ZHEGST", &i__1);
return 0;
}
/* Quick return if possible */
if (*n == 0) {
return 0;
}
/* Determine the block size for this environment. */
nb = ilaenv_(&c__1, "ZHEGST", uplo, n, &c_n1, &c_n1, &c_n1);
if (nb <= 1 || nb >= *n) {
/* Use unblocked code */
zhegs2_(itype, uplo, n, &a[a_offset], lda, &b[b_offset], ldb, info);
} else {
/* Use blocked code */
if (*itype == 1) {
if (upper) {
/* Compute inv(U')*A*inv(U) */
i__1 = *n;
i__2 = nb;
for (k = 1; i__2 < 0 ? k >= i__1 : k <= i__1; k += i__2) {
/* Computing MIN */
i__3 = *n - k + 1;
kb = MIN(i__3,nb);
/* Update the upper triangle of A(k:n,k:n) */
zhegs2_(itype, uplo, &kb, &a[k + k * a_dim1], lda, &b[k +
k * b_dim1], ldb, info);
if (k + kb <= *n) {
i__3 = *n - k - kb + 1;
ztrsm_("Left", uplo, "Conjugate transpose", "Non-unit"
, &kb, &i__3, &c_b1, &b[k + k * b_dim1], ldb,
&a[k + (k + kb) * a_dim1], lda);
i__3 = *n - k - kb + 1;
z__1.r = -.5, z__1.i = -0.;
zhemm_("Left", uplo, &kb, &i__3, &z__1, &a[k + k *
a_dim1], lda, &b[k + (k + kb) * b_dim1], ldb,
&c_b1, &a[k + (k + kb) * a_dim1], lda);
i__3 = *n - k - kb + 1;
z__1.r = -1., z__1.i = -0.;
zher2k_(uplo, "Conjugate transpose", &i__3, &kb, &
z__1, &a[k + (k + kb) * a_dim1], lda, &b[k + (
k + kb) * b_dim1], ldb, &c_b18, &a[k + kb + (
k + kb) * a_dim1], lda)
;
i__3 = *n - k - kb + 1;
z__1.r = -.5, z__1.i = -0.;
zhemm_("Left", uplo, &kb, &i__3, &z__1, &a[k + k *
a_dim1], lda, &b[k + (k + kb) * b_dim1], ldb,
&c_b1, &a[k + (k + kb) * a_dim1], lda);
i__3 = *n - k - kb + 1;
ztrsm_("Right", uplo, "No transpose", "Non-unit", &kb,
&i__3, &c_b1, &b[k + kb + (k + kb) * b_dim1],
ldb, &a[k + (k + kb) * a_dim1], lda);
}
/* L10: */
}
} else {
/* Compute inv(L)*A*inv(L') */
i__2 = *n;
i__1 = nb;
for (k = 1; i__1 < 0 ? k >= i__2 : k <= i__2; k += i__1) {
/* Computing MIN */
i__3 = *n - k + 1;
kb = MIN(i__3,nb);
/* Update the lower triangle of A(k:n,k:n) */
zhegs2_(itype, uplo, &kb, &a[k + k * a_dim1], lda, &b[k +
k * b_dim1], ldb, info);
if (k + kb <= *n) {
i__3 = *n - k - kb + 1;
ztrsm_("Right", uplo, "Conjugate transpose", "Non-un"
"it", &i__3, &kb, &c_b1, &b[k + k * b_dim1],
ldb, &a[k + kb + k * a_dim1], lda);
i__3 = *n - k - kb + 1;
z__1.r = -.5, z__1.i = -0.;
zhemm_("Right", uplo, &i__3, &kb, &z__1, &a[k + k *
a_dim1], lda, &b[k + kb + k * b_dim1], ldb, &
c_b1, &a[k + kb + k * a_dim1], lda);
i__3 = *n - k - kb + 1;
z__1.r = -1., z__1.i = -0.;
zher2k_(uplo, "No transpose", &i__3, &kb, &z__1, &a[k
+ kb + k * a_dim1], lda, &b[k + kb + k *
b_dim1], ldb, &c_b18, &a[k + kb + (k + kb) *
a_dim1], lda);
i__3 = *n - k - kb + 1;
z__1.r = -.5, z__1.i = -0.;
zhemm_("Right", uplo, &i__3, &kb, &z__1, &a[k + k *
a_dim1], lda, &b[k + kb + k * b_dim1], ldb, &
c_b1, &a[k + kb + k * a_dim1], lda);
i__3 = *n - k - kb + 1;
ztrsm_("Left", uplo, "No transpose", "Non-unit", &
i__3, &kb, &c_b1, &b[k + kb + (k + kb) *
b_dim1], ldb, &a[k + kb + k * a_dim1], lda);
}
/* L20: */
}
}
} else {
if (upper) {
/* Compute U*A*U' */
i__1 = *n;
i__2 = nb;
for (k = 1; i__2 < 0 ? k >= i__1 : k <= i__1; k += i__2) {
/* Computing MIN */
i__3 = *n - k + 1;
kb = MIN(i__3,nb);
/* Update the upper triangle of A(1:k+kb-1,1:k+kb-1) */
i__3 = k - 1;
ztrmm_("Left", uplo, "No transpose", "Non-unit", &i__3, &
kb, &c_b1, &b[b_offset], ldb, &a[k * a_dim1 + 1],
lda);
i__3 = k - 1;
zhemm_("Right", uplo, &i__3, &kb, &c_b2, &a[k + k *
a_dim1], lda, &b[k * b_dim1 + 1], ldb, &c_b1, &a[
k * a_dim1 + 1], lda);
i__3 = k - 1;
zher2k_(uplo, "No transpose", &i__3, &kb, &c_b1, &a[k *
a_dim1 + 1], lda, &b[k * b_dim1 + 1], ldb, &c_b18,
&a[a_offset], lda);
i__3 = k - 1;
zhemm_("Right", uplo, &i__3, &kb, &c_b2, &a[k + k *
a_dim1], lda, &b[k * b_dim1 + 1], ldb, &c_b1, &a[
k * a_dim1 + 1], lda);
i__3 = k - 1;
ztrmm_("Right", uplo, "Conjugate transpose", "Non-unit", &
i__3, &kb, &c_b1, &b[k + k * b_dim1], ldb, &a[k *
a_dim1 + 1], lda);
zhegs2_(itype, uplo, &kb, &a[k + k * a_dim1], lda, &b[k +
k * b_dim1], ldb, info);
/* L30: */
}
} else {
/* Compute L'*A*L */
i__2 = *n;
i__1 = nb;
for (k = 1; i__1 < 0 ? k >= i__2 : k <= i__2; k += i__1) {
/* Computing MIN */
i__3 = *n - k + 1;
kb = MIN(i__3,nb);
/* Update the lower triangle of A(1:k+kb-1,1:k+kb-1) */
i__3 = k - 1;
ztrmm_("Right", uplo, "No transpose", "Non-unit", &kb, &
i__3, &c_b1, &b[b_offset], ldb, &a[k + a_dim1],
lda);
i__3 = k - 1;
zhemm_("Left", uplo, &kb, &i__3, &c_b2, &a[k + k * a_dim1]
, lda, &b[k + b_dim1], ldb, &c_b1, &a[k + a_dim1],
lda);
i__3 = k - 1;
zher2k_(uplo, "Conjugate transpose", &i__3, &kb, &c_b1, &
a[k + a_dim1], lda, &b[k + b_dim1], ldb, &c_b18, &
a[a_offset], lda);
i__3 = k - 1;
zhemm_("Left", uplo, &kb, &i__3, &c_b2, &a[k + k * a_dim1]
, lda, &b[k + b_dim1], ldb, &c_b1, &a[k + a_dim1],
lda);
i__3 = k - 1;
ztrmm_("Left", uplo, "Conjugate transpose", "Non-unit", &
kb, &i__3, &c_b1, &b[k + k * b_dim1], ldb, &a[k +
a_dim1], lda);
zhegs2_(itype, uplo, &kb, &a[k + k * a_dim1], lda, &b[k +
k * b_dim1], ldb, info);
/* L40: */
}
}
}
}
return 0;
/* End of ZHEGST */
} /* zhegst_ */
