/*! _zhematrix*zhematrix operator */
inline _zgematrix operator*(const _zhematrix& matA, const zhematrix& matB)
{
#ifdef CPPL_VERBOSE
std::cerr << "# [MARK] operator*(const _zhematrix&, const zhematrix&)"
<< std::endl;
#endif//CPPL_VERBOSE
#ifdef CPPL_DEBUG
if(matA.N!=matB.N){
std::cerr << "[ERROR] operator*(_zhematrix&, zhematrix&)" << std::endl
<< "These two matrises can not make a product." << std::endl
<< "Your input was (" << matA.N << "x" << matA.N << ") * ("
<< matB.N << "x" << matB.N << ")." << std::endl;
exit(1);
}
#endif//CPPL_DEBUG
matB.complete();
zgematrix newmat( matA.N, matB.N );
zhemm_( 'L', 'L', matA.N, matB.N, std::complex<double>(1.0,0.0),
matA.Array, matA.N, matB.Array, matB.N,
std::complex<double>(0.0,0.0), newmat.array, newmat.m );
matA.destroy();
return _(newmat);
}
/*! zgematrix*=zhematrix operator */
inline zgematrix& zgematrix::operator*=(const zhematrix& mat)
{
#ifdef CPPL_VERBOSE
std::cerr << "# [MARK] zgematrix::operator*=(const zhematrix&)"
<< std::endl;
#endif//CPPL_VERBOSE
#ifdef CPPL_DEBUG
if(N!=mat.N){
std::cerr << "[ERROR] zgematrix::operator*=(zhematrix&)" << std::endl
<< "These two matrises can not make a product." << std::endl
<< "Your input was (" << M << "x" << N << ") *= ("
<< mat.N << "x" << mat.N << ")." << std::endl;
exit(1);
}
#endif//CPPL_DEBUG
zgematrix newmat( M, mat.N );
zhemm_( 'R', 'L', mat.N, N, std::complex<double>(1.0,0.0), mat.Array, mat.N,
Array, M, std::complex<double>(0.0,0.0), newmat.array, newmat.m );
swap(*this,newmat);
return *this;
}
/*
* transform ket, s2 to label AO symmetry
*/
int RIhalfmmm_r_s2_ket(double complex *vout, double complex *vin,
struct _AO2MOEnvs *envs, int seekdim)
{
switch (seekdim) {
case 1: return envs->nao * envs->ket_count;
case 2: return envs->nao * (envs->nao+1) / 2;
}
const double complex Z0 = 0;
const double complex Z1 = 1;
const char SIDE_L = 'L';
const char UPLO_U = 'U';
int n2c = envs->nao;
int j_start = envs->ket_start;
int j_count = envs->ket_count;
double complex *mo_coeff = envs->mo_coeff;
double complex *buf = malloc(sizeof(double complex)*n2c*j_count);
int i, j;
zhemm_(&SIDE_L, &UPLO_U, &n2c, &j_count,
&Z1, vin, &n2c, mo_coeff+j_start*n2c, &n2c,
&Z0, buf, &n2c);
for (j = 0; j < n2c; j++) {
for (i = 0; i < j_count; i++) {
vout[i] = buf[i*n2c+j];
}
vout += j_count;
}
free(buf);
return 0;
}
int
f2c_zhemm(char* side, char* uplo, integer* M, integer* N,
doublecomplex* alpha,
doublecomplex* A, integer* lda,
doublecomplex* B, integer* ldb,
doublecomplex* beta,
doublecomplex* C, integer* ldc)
{
zhemm_(side, uplo, M, N,
alpha, A, lda, B, ldb, beta, C, ldc);
return 0;
}
/*! zhematrix*zgematrix operator */
inline _zgematrix operator*(const zhematrix& matA, const zgematrix& matB)
{VERBOSE_REPORT;
#ifdef CPPL_DEBUG
if(matA.n!=matB.m){
ERROR_REPORT;
std::cerr << "These two matrises can not make a product." << std::endl
<< "Your input was (" << matA.n << "x" << matA.n << ") * (" << matB.m << "x" << matB.n << ")." << std::endl;
exit(1);
}
#endif//CPPL_DEBUG
zgematrix newmat( matA.n, matB.n );
zhemm_( 'l', 'l', matA.n, matB.n, comple(1.0,0.0),
matA.array, matA.n, matB.array, matB.m,
comple(0.0,0.0), newmat.array, newmat.m );
return _(newmat);
}
/*! zgematrix*=_zhematrix operator */
inline zgematrix& zgematrix::operator*=(const _zhematrix& mat)
{VERBOSE_REPORT;
#ifdef CPPL_DEBUG
if(n!=mat.n){
ERROR_REPORT;
std::cerr << "These two matrises can not make a product." << std::endl
<< "Your input was (" << m << "x" << n << ") *= (" << mat.n << "x" << mat.n << ")." << std::endl;
exit(1);
}
#endif//CPPL_DEBUG
zgematrix newmat( m, mat.n );
zhemm_( 'R', 'l', mat.n, n, comple(1.0,0.0), mat.array, mat.n,
array, m, comple(0.0,0.0), newmat.array, newmat.m );
swap(*this,newmat);
mat.destroy();
return *this;
}
/*
* transform bra (without doing conj(mo)), v_{iq} = C_{pi} v_{pq}
* s2 to label AO symmetry
*/
int RIhalfmmm_r_s2_bra_noconj(double complex *vout, double complex *vin,
struct _AO2MOEnvs *envs, int seekdim)
{
switch (seekdim) {
case 1: return envs->bra_count * envs->nao;
case 2: return envs->nao * (envs->nao+1) / 2;
}
const double complex Z0 = 0;
const double complex Z1 = 1;
const char SIDE_L = 'L';
const char UPLO_U = 'U';
int n2c = envs->nao;
int i_start = envs->bra_start;
int i_count = envs->bra_count;
double complex *mo_coeff = envs->mo_coeff;
zhemm_(&SIDE_L, &UPLO_U, &n2c, &i_count,
&Z1, vin, &n2c, mo_coeff+i_start*n2c, &n2c,
&Z0, vout, &n2c);
return 0;
}
void
zhemm(char side, char uplo, int m, int n, doublecomplex *alpha, doublecomplex *a, int lda, doublecomplex *b, int ldb, doublecomplex *beta, doublecomplex *c, int ldc)
{
zhemm_( &side, &uplo, &m, &n, alpha, a, &lda, b, &ldb, beta, c, &ldc);
}
int main( int argc, char** argv )
{
obj_t a, b, c;
obj_t c_save;
obj_t alpha, beta;
dim_t m, n;
dim_t p;
dim_t p_begin, p_end, p_inc;
int m_input, n_input;
num_t dt;
int r, n_repeats;
side_t side;
uplo_t uploa;
f77_char f77_side;
f77_char f77_uploa;
double dtime;
double dtime_save;
double gflops;
bli_init();
//bli_error_checking_level_set( BLIS_NO_ERROR_CHECKING );
n_repeats = 3;
#ifndef PRINT
p_begin = 200;
p_end = 2000;
p_inc = 200;
m_input = -1;
n_input = -1;
#else
p_begin = 16;
p_end = 16;
p_inc = 1;
m_input = 4;
n_input = 4;
#endif
#if 1
//dt = BLIS_FLOAT;
dt = BLIS_DOUBLE;
#else
//dt = BLIS_SCOMPLEX;
dt = BLIS_DCOMPLEX;
#endif
side = BLIS_LEFT;
//side = BLIS_RIGHT;
uploa = BLIS_LOWER;
//uploa = BLIS_UPPER;
bli_param_map_blis_to_netlib_side( side, &f77_side );
bli_param_map_blis_to_netlib_uplo( uploa, &f77_uploa );
for ( p = p_begin; p <= p_end; p += p_inc )
{
if ( m_input < 0 ) m = p * ( dim_t )abs(m_input);
else m = ( dim_t ) m_input;
if ( n_input < 0 ) n = p * ( dim_t )abs(n_input);
else n = ( dim_t ) n_input;
bli_obj_create( dt, 1, 1, 0, 0, &alpha );
bli_obj_create( dt, 1, 1, 0, 0, &beta );
if ( bli_is_left( side ) )
bli_obj_create( dt, m, m, 0, 0, &a );
else
bli_obj_create( dt, n, n, 0, 0, &a );
bli_obj_create( dt, m, n, 0, 0, &b );
bli_obj_create( dt, m, n, 0, 0, &c );
bli_obj_create( dt, m, n, 0, 0, &c_save );
bli_randm( &a );
bli_randm( &b );
bli_randm( &c );
bli_obj_set_struc( BLIS_HERMITIAN, a );
bli_obj_set_uplo( uploa, a );
// Randomize A, make it densely Hermitian, and zero the unstored
// triangle to ensure the implementation reads only from the stored
// region.
bli_randm( &a );
bli_mkherm( &a );
bli_mktrim( &a );
/*
bli_obj_toggle_uplo( a );
bli_obj_inc_diag_off( 1, a );
bli_setm( &BLIS_ZERO, &a );
bli_obj_inc_diag_off( -1, a );
bli_obj_toggle_uplo( a );
bli_obj_set_diag( BLIS_NONUNIT_DIAG, a );
bli_scalm( &BLIS_TWO, &a );
bli_scalm( &BLIS_TWO, &a );
*/
bli_setsc( (2.0/1.0), 1.0, &alpha );
bli_setsc( -(1.0/1.0), 0.0, &beta );
bli_copym( &c, &c_save );
dtime_save = 1.0e9;
for ( r = 0; r < n_repeats; ++r )
{
bli_copym( &c_save, &c );
dtime = bli_clock();
#ifdef PRINT
bli_printm( "a", &a, "%4.1f", "" );
bli_printm( "b", &b, "%4.1f", "" );
bli_printm( "c", &c, "%4.1f", "" );
#endif
#ifdef BLIS
bli_hemm( side,
&alpha,
&a,
&b,
&beta,
&c );
#else
if ( bli_is_float( dt ) )
{
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 );
float* alphap = bli_obj_buffer( alpha );
float* ap = bli_obj_buffer( a );
float* bp = bli_obj_buffer( b );
float* betap = bli_obj_buffer( beta );
float* cp = bli_obj_buffer( c );
ssymm_( &f77_side,
&f77_uploa,
&mm,
&nn,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
else if ( bli_is_double( dt ) )
{
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_( &f77_side,
&f77_uploa,
&mm,
&nn,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
else if ( bli_is_scomplex( dt ) )
{
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 );
scomplex* alphap = bli_obj_buffer( alpha );
scomplex* ap = bli_obj_buffer( a );
scomplex* bp = bli_obj_buffer( b );
scomplex* betap = bli_obj_buffer( beta );
scomplex* cp = bli_obj_buffer( c );
chemm_( &f77_side,
&f77_uploa,
&mm,
&nn,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
else if ( bli_is_dcomplex( dt ) )
{
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_( &f77_side,
&f77_uploa,
&mm,
&nn,
alphap,
ap, &lda,
bp, &ldb,
betap,
cp, &ldc );
}
#endif
#ifdef PRINT
bli_printm( "c after", &c, "%9.5f", "" );
exit(1);
#endif
dtime_save = bli_clock_min_diff( dtime_save, dtime );
}
if ( bli_is_left( side ) )
gflops = ( 2.0 * m * m * n ) / ( dtime_save * 1.0e9 );
else
gflops = ( 2.0 * m * n * n ) / ( dtime_save * 1.0e9 );
if ( bli_is_complex( dt ) ) gflops *= 4.0;
#ifdef BLIS
printf( "data_hemm_blis" );
#else
printf( "data_hemm_%s", BLAS );
#endif
printf( "( %2lu, 1:4 ) = [ %4lu %4lu %10.3e %6.3f ];\n",
( unsigned long )(p - p_begin + 1)/p_inc + 1,
( unsigned long )m,
( unsigned long )n, dtime_save, gflops );
bli_obj_free( &alpha );
bli_obj_free( &beta );
bli_obj_free( &a );
bli_obj_free( &b );
bli_obj_free( &c );
bli_obj_free( &c_save );
}
bli_finalize();
return 0;
}
/* Subroutine */ int zlarhs_(char *path, char *xtype, char *uplo, char *trans,
integer *m, integer *n, integer *kl, integer *ku, integer *nrhs,
doublecomplex *a, integer *lda, doublecomplex *x, integer *ldx,
doublecomplex *b, integer *ldb, integer *iseed, integer *info)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, x_dim1, x_offset, i__1;
/* Local variables */
integer j;
char c1[1], c2[2];
integer mb, nx;
logical gen, tri, qrs, sym, band;
char diag[1];
logical tran;
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 *), zgbmv_(char *, integer *, integer *,
integer *, integer *, doublecomplex *, doublecomplex *, integer *,
doublecomplex *, integer *, doublecomplex *, doublecomplex *,
integer *), zhbmv_(char *, integer *, integer *,
doublecomplex *, doublecomplex *, integer *, doublecomplex *,
integer *, doublecomplex *, doublecomplex *, integer *),
zsbmv_(char *, integer *, integer *, doublecomplex *,
doublecomplex *, integer *, doublecomplex *, integer *,
doublecomplex *, doublecomplex *, integer *), ztbmv_(char
*, char *, char *, integer *, integer *, doublecomplex *, integer
*, doublecomplex *, integer *), zhpmv_(
char *, integer *, doublecomplex *, doublecomplex *,
doublecomplex *, integer *, doublecomplex *, doublecomplex *,
integer *), ztrmm_(char *, char *, char *, char *,
integer *, integer *, doublecomplex *, doublecomplex *, integer *,
doublecomplex *, integer *),
zspmv_(char *, integer *, doublecomplex *, doublecomplex *,
doublecomplex *, integer *, doublecomplex *, doublecomplex *,
integer *), zsymm_(char *, char *, integer *, integer *,
doublecomplex *, doublecomplex *, integer *, doublecomplex *,
integer *, doublecomplex *, doublecomplex *, integer *), ztpmv_(char *, char *, char *, integer *, doublecomplex *
, doublecomplex *, integer *), xerbla_(
char *, integer *);
extern logical lsamen_(integer *, char *, char *);
logical notran;
extern /* Subroutine */ int zlacpy_(char *, integer *, integer *,
doublecomplex *, integer *, doublecomplex *, integer *),
zlarnv_(integer *, integer *, integer *, doublecomplex *);
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* ZLARHS chooses a set of NRHS random solution vectors and sets */
/* up the right hand sides for the linear system */
/* op( A ) * X = B, */
/* where op( A ) may be A, A**T (transpose of A), or A**H (conjugate */
/* transpose of A). */
/* Arguments */
/* ========= */
/* PATH (input) CHARACTER*3 */
/* The type of the complex matrix A. PATH may be given in any */
/* combination of upper and lower case. Valid paths include */
/* xGE: General m x n matrix */
/* xGB: General banded matrix */
/* xPO: Hermitian positive definite, 2-D storage */
/* xPP: Hermitian positive definite packed */
/* xPB: Hermitian positive definite banded */
/* xHE: Hermitian indefinite, 2-D storage */
/* xHP: Hermitian indefinite packed */
/* xHB: Hermitian indefinite banded */
/* xSY: Symmetric indefinite, 2-D storage */
/* xSP: Symmetric indefinite packed */
/* xSB: Symmetric indefinite banded */
/* xTR: Triangular */
/* xTP: Triangular packed */
/* xTB: Triangular banded */
/* xQR: General m x n matrix */
/* xLQ: General m x n matrix */
/* xQL: General m x n matrix */
/* xRQ: General m x n matrix */
/* where the leading character indicates the precision. */
/* XTYPE (input) CHARACTER*1 */
/* Specifies how the exact solution X will be determined: */
/* = 'N': New solution; generate a random X. */
/* = 'C': Computed; use value of X on entry. */
/* UPLO (input) CHARACTER*1 */
/* Used only if A is symmetric or triangular; specifies whether */
/* the upper or lower triangular part of the matrix A is stored. */
/* = 'U': Upper triangular */
/* = 'L': Lower triangular */
/* TRANS (input) CHARACTER*1 */
/* Used only if A is nonsymmetric; specifies the operation */
/* applied to the matrix A. */
/* = 'N': B := A * X */
/* = 'T': B := A**T * X */
/* = 'C': B := A**H * X */
/* M (input) INTEGER */
/* The number of rows of the matrix A. M >= 0. */
/* N (input) INTEGER */
/* The number of columns of the matrix A. N >= 0. */
/* KL (input) INTEGER */
/* Used only if A is a band matrix; specifies the number of */
/* subdiagonals of A if A is a general band matrix or if A is */
/* symmetric or triangular and UPLO = 'L'; specifies the number */
/* of superdiagonals of A if A is symmetric or triangular and */
/* UPLO = 'U'. 0 <= KL <= M-1. */
/* KU (input) INTEGER */
/* Used only if A is a general band matrix or if A is */
/* triangular. */
/* If PATH = xGB, specifies the number of superdiagonals of A, */
/* and 0 <= KU <= N-1. */
/* If PATH = xTR, xTP, or xTB, specifies whether or not the */
/* matrix has unit diagonal: */
/* = 1: matrix has non-unit diagonal (default) */
/* = 2: matrix has unit diagonal */
/* NRHS (input) INTEGER */
/* The number of right hand side vectors in the system A*X = B. */
/* A (input) COMPLEX*16 array, dimension (LDA,N) */
/* The test matrix whose type is given by PATH. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. */
/* If PATH = xGB, LDA >= KL+KU+1. */
/* If PATH = xPB, xSB, xHB, or xTB, LDA >= KL+1. */
/* Otherwise, LDA >= max(1,M). */
/* X (input or output) COMPLEX*16 array, dimension (LDX,NRHS) */
/* On entry, if XTYPE = 'C' (for 'Computed'), then X contains */
/* the exact solution to the system of linear equations. */
/* On exit, if XTYPE = 'N' (for 'New'), then X is initialized */
/* with random values. */
/* LDX (input) INTEGER */
/* The leading dimension of the array X. If TRANS = 'N', */
/* LDX >= max(1,N); if TRANS = 'T', LDX >= max(1,M). */
/* B (output) COMPLEX*16 array, dimension (LDB,NRHS) */
/* The right hand side vector(s) for the system of equations, */
/* computed from B = op(A) * X, where op(A) is determined by */
/* TRANS. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. If TRANS = 'N', */
/* LDB >= max(1,M); if TRANS = 'T', LDB >= max(1,N). */
/* ISEED (input/output) INTEGER array, dimension (4) */
/* The seed vector for the random number generator (used in */
/* ZLATMS). Modified on exit. */
/* INFO (output) INTEGER */
/* = 0: successful exit */
/* < 0: if INFO = -i, the i-th argument had an illegal value */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Test the input parameters. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
x_dim1 = *ldx;
x_offset = 1 + x_dim1;
x -= x_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--iseed;
/* Function Body */
*info = 0;
*(unsigned char *)c1 = *(unsigned char *)path;
s_copy(c2, path + 1, (ftnlen)2, (ftnlen)2);
tran = lsame_(trans, "T") || lsame_(trans, "C");
notran = ! tran;
gen = lsame_(path + 1, "G");
qrs = lsame_(path + 1, "Q") || lsame_(path + 2,
"Q");
sym = lsame_(path + 1, "P") || lsame_(path + 1,
"S") || lsame_(path + 1, "H");
tri = lsame_(path + 1, "T");
band = lsame_(path + 2, "B");
if (! lsame_(c1, "Zomplex precision")) {
*info = -1;
} else if (! (lsame_(xtype, "N") || lsame_(xtype,
"C"))) {
*info = -2;
} else if ((sym || tri) && ! (lsame_(uplo, "U") ||
lsame_(uplo, "L"))) {
*info = -3;
} else if ((gen || qrs) && ! (tran || lsame_(trans, "N"))) {
*info = -4;
} else if (*m < 0) {
*info = -5;
} else if (*n < 0) {
*info = -6;
} else if (band && *kl < 0) {
*info = -7;
} else if (band && *ku < 0) {
*info = -8;
} else if (*nrhs < 0) {
*info = -9;
} else if (! band && *lda < max(1,*m) || band && (sym || tri) && *lda < *
kl + 1 || band && gen && *lda < *kl + *ku + 1) {
*info = -11;
} else if (notran && *ldx < max(1,*n) || tran && *ldx < max(1,*m)) {
*info = -13;
} else if (notran && *ldb < max(1,*m) || tran && *ldb < max(1,*n)) {
*info = -15;
}
if (*info != 0) {
i__1 = -(*info);
xerbla_("ZLARHS", &i__1);
return 0;
}
/* Initialize X to NRHS random vectors unless XTYPE = 'C'. */
if (tran) {
nx = *m;
mb = *n;
} else {
nx = *n;
mb = *m;
}
if (! lsame_(xtype, "C")) {
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
zlarnv_(&c__2, &iseed[1], n, &x[j * x_dim1 + 1]);
/* L10: */
}
}
/* Multiply X by op( A ) using an appropriate */
/* matrix multiply routine. */
if (lsamen_(&c__2, c2, "GE") || lsamen_(&c__2, c2,
"QR") || lsamen_(&c__2, c2, "LQ") || lsamen_(&c__2, c2, "QL") ||
lsamen_(&c__2, c2, "RQ")) {
/* General matrix */
zgemm_(trans, "N", &mb, nrhs, &nx, &c_b1, &a[a_offset], lda, &x[
x_offset], ldx, &c_b2, &b[b_offset], ldb);
} else if (lsamen_(&c__2, c2, "PO") || lsamen_(&
c__2, c2, "HE")) {
/* Hermitian matrix, 2-D storage */
zhemm_("Left", uplo, n, nrhs, &c_b1, &a[a_offset], lda, &x[x_offset],
ldx, &c_b2, &b[b_offset], ldb);
} else if (lsamen_(&c__2, c2, "SY")) {
/* Symmetric matrix, 2-D storage */
zsymm_("Left", uplo, n, nrhs, &c_b1, &a[a_offset], lda, &x[x_offset],
ldx, &c_b2, &b[b_offset], ldb);
} else if (lsamen_(&c__2, c2, "GB")) {
/* General matrix, band storage */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
zgbmv_(trans, m, n, kl, ku, &c_b1, &a[a_offset], lda, &x[j *
x_dim1 + 1], &c__1, &c_b2, &b[j * b_dim1 + 1], &c__1);
/* L20: */
}
} else if (lsamen_(&c__2, c2, "PB") || lsamen_(&
c__2, c2, "HB")) {
/* Hermitian matrix, band storage */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
zhbmv_(uplo, n, kl, &c_b1, &a[a_offset], lda, &x[j * x_dim1 + 1],
&c__1, &c_b2, &b[j * b_dim1 + 1], &c__1);
/* L30: */
}
} else if (lsamen_(&c__2, c2, "SB")) {
/* Symmetric matrix, band storage */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
zsbmv_(uplo, n, kl, &c_b1, &a[a_offset], lda, &x[j * x_dim1 + 1],
&c__1, &c_b2, &b[j * b_dim1 + 1], &c__1);
/* L40: */
}
} else if (lsamen_(&c__2, c2, "PP") || lsamen_(&
c__2, c2, "HP")) {
/* Hermitian matrix, packed storage */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
zhpmv_(uplo, n, &c_b1, &a[a_offset], &x[j * x_dim1 + 1], &c__1, &
c_b2, &b[j * b_dim1 + 1], &c__1);
/* L50: */
}
} else if (lsamen_(&c__2, c2, "SP")) {
/* Symmetric matrix, packed storage */
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
zspmv_(uplo, n, &c_b1, &a[a_offset], &x[j * x_dim1 + 1], &c__1, &
c_b2, &b[j * b_dim1 + 1], &c__1);
/* L60: */
}
} else if (lsamen_(&c__2, c2, "TR")) {
/* Triangular matrix. Note that for triangular matrices, */
/* KU = 1 => non-unit triangular */
/* KU = 2 => unit triangular */
zlacpy_("Full", n, nrhs, &x[x_offset], ldx, &b[b_offset], ldb);
if (*ku == 2) {
*(unsigned char *)diag = 'U';
} else {
*(unsigned char *)diag = 'N';
}
ztrmm_("Left", uplo, trans, diag, n, nrhs, &c_b1, &a[a_offset], lda, &
b[b_offset], ldb);
} else if (lsamen_(&c__2, c2, "TP")) {
/* Triangular matrix, packed storage */
zlacpy_("Full", n, nrhs, &x[x_offset], ldx, &b[b_offset], ldb);
if (*ku == 2) {
*(unsigned char *)diag = 'U';
} else {
*(unsigned char *)diag = 'N';
}
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
ztpmv_(uplo, trans, diag, n, &a[a_offset], &b[j * b_dim1 + 1], &
c__1);
/* L70: */
}
} else if (lsamen_(&c__2, c2, "TB")) {
/* Triangular matrix, banded storage */
zlacpy_("Full", n, nrhs, &x[x_offset], ldx, &b[b_offset], ldb);
if (*ku == 2) {
*(unsigned char *)diag = 'U';
} else {
*(unsigned char *)diag = 'N';
}
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
ztbmv_(uplo, trans, diag, n, kl, &a[a_offset], lda, &b[j * b_dim1
+ 1], &c__1);
/* L80: */
}
} else {
/* If none of the above, set INFO = -1 and return */
*info = -1;
i__1 = -(*info);
xerbla_("ZLARHS", &i__1);
}
return 0;
/* End of ZLARHS */
} /* zlarhs_ */
/* 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_ */
/* Subroutine */ int zpot02_(char *uplo, integer *n, integer *nrhs,
doublecomplex *a, integer *lda, doublecomplex *x, integer *ldx,
doublecomplex *b, integer *ldb, doublereal *rwork, doublereal *resid)
{
/* System generated locals */
integer a_dim1, a_offset, b_dim1, b_offset, x_dim1, x_offset, i__1;
doublereal d__1, d__2;
doublecomplex z__1;
/* Local variables */
integer j;
doublereal eps, anorm, bnorm;
extern /* Subroutine */ int zhemm_(char *, char *, integer *, integer *,
doublecomplex *, doublecomplex *, integer *, doublecomplex *,
integer *, doublecomplex *, doublecomplex *, integer *);
doublereal xnorm;
extern doublereal dlamch_(char *), zlanhe_(char *, char *,
integer *, doublecomplex *, integer *, doublereal *), dzasum_(integer *, doublecomplex *, integer *);
/* -- LAPACK test routine (version 3.1) -- */
/* Univ. of Tennessee, Univ. of California Berkeley and NAG Ltd.. */
/* November 2006 */
/* .. Scalar Arguments .. */
/* .. */
/* .. Array Arguments .. */
/* .. */
/* Purpose */
/* ======= */
/* ZPOT02 computes the residual for the solution of a Hermitian system */
/* of linear equations A*x = b: */
/* RESID = norm(B - A*X) / ( norm(A) * norm(X) * EPS ), */
/* where EPS is the machine epsilon. */
/* Arguments */
/* ========= */
/* UPLO (input) CHARACTER*1 */
/* Specifies whether the upper or lower triangular part of the */
/* Hermitian matrix A is stored: */
/* = 'U': Upper triangular */
/* = 'L': Lower triangular */
/* N (input) INTEGER */
/* The number of rows and columns of the matrix A. N >= 0. */
/* NRHS (input) INTEGER */
/* The number of columns of B, the matrix of right hand sides. */
/* NRHS >= 0. */
/* A (input) COMPLEX*16 array, dimension (LDA,N) */
/* The original Hermitian matrix A. */
/* LDA (input) INTEGER */
/* The leading dimension of the array A. LDA >= max(1,N) */
/* X (input) COMPLEX*16 array, dimension (LDX,NRHS) */
/* The computed solution vectors for the system of linear */
/* equations. */
/* LDX (input) INTEGER */
/* The leading dimension of the array X. LDX >= max(1,N). */
/* B (input/output) COMPLEX*16 array, dimension (LDB,NRHS) */
/* On entry, the right hand side vectors for the system of */
/* linear equations. */
/* On exit, B is overwritten with the difference B - A*X. */
/* LDB (input) INTEGER */
/* The leading dimension of the array B. LDB >= max(1,N). */
/* RWORK (workspace) DOUBLE PRECISION array, dimension (N) */
/* RESID (output) DOUBLE PRECISION */
/* The maximum over the number of right hand sides of */
/* norm(B - A*X) / ( norm(A) * norm(X) * EPS ). */
/* ===================================================================== */
/* .. Parameters .. */
/* .. */
/* .. Local Scalars .. */
/* .. */
/* .. External Functions .. */
/* .. */
/* .. External Subroutines .. */
/* .. */
/* .. Intrinsic Functions .. */
/* .. */
/* .. Executable Statements .. */
/* Quick exit if N = 0 or NRHS = 0. */
/* Parameter adjustments */
a_dim1 = *lda;
a_offset = 1 + a_dim1;
a -= a_offset;
x_dim1 = *ldx;
x_offset = 1 + x_dim1;
x -= x_offset;
b_dim1 = *ldb;
b_offset = 1 + b_dim1;
b -= b_offset;
--rwork;
/* Function Body */
if (*n <= 0 || *nrhs <= 0) {
*resid = 0.;
return 0;
}
/* Exit with RESID = 1/EPS if ANORM = 0. */
eps = dlamch_("Epsilon");
anorm = zlanhe_("1", uplo, n, &a[a_offset], lda, &rwork[1]);
if (anorm <= 0.) {
*resid = 1. / eps;
return 0;
}
/* Compute B - A*X */
z__1.r = -1., z__1.i = -0.;
zhemm_("Left", uplo, n, nrhs, &z__1, &a[a_offset], lda, &x[x_offset], ldx,
&c_b1, &b[b_offset], ldb);
/* Compute the maximum over the number of right hand sides of */
/* norm( B - A*X ) / ( norm(A) * norm(X) * EPS ) . */
*resid = 0.;
i__1 = *nrhs;
for (j = 1; j <= i__1; ++j) {
bnorm = dzasum_(n, &b[j * b_dim1 + 1], &c__1);
xnorm = dzasum_(n, &x[j * x_dim1 + 1], &c__1);
if (xnorm <= 0.) {
*resid = 1. / eps;
} else {
/* Computing MAX */
d__1 = *resid, d__2 = bnorm / anorm / xnorm / eps;
*resid = max(d__1,d__2);
}
/* L10: */
}
return 0;
/* End of ZPOT02 */
} /* zpot02_ */
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_ */
