void F77_zgerc(int *layout, int *m, int *n, CBLAS_TEST_ZOMPLEX *alpha,
CBLAS_TEST_ZOMPLEX *x, int *incx, CBLAS_TEST_ZOMPLEX *y, int *incy,
CBLAS_TEST_ZOMPLEX *a, int *lda) {
CBLAS_TEST_ZOMPLEX *A;
int i,j,LDA;
if (*layout == TEST_ROW_MJR) {
LDA = *n+1;
A=(CBLAS_TEST_ZOMPLEX* )malloc((*m)*LDA*sizeof(CBLAS_TEST_ZOMPLEX ) );
for( i=0; i<*m; i++ )
for( j=0; j<*n; j++ ) {
A[ LDA*i+j ].real=a[ (*lda)*j+i ].real;
A[ LDA*i+j ].imag=a[ (*lda)*j+i ].imag;
}
cblas_zgerc( CblasRowMajor, *m, *n, alpha, x, *incx, y, *incy, A, LDA );
for( i=0; i<*m; i++ )
for( j=0; j<*n; j++ ) {
a[ (*lda)*j+i ].real=A[ LDA*i+j ].real;
a[ (*lda)*j+i ].imag=A[ LDA*i+j ].imag;
}
free(A);
}
else if (*layout == TEST_COL_MJR)
cblas_zgerc( CblasColMajor, *m, *n, alpha, x, *incx, y, *incy, a, *lda );
else
cblas_zgerc( UNDEFINED, *m, *n, alpha, x, *incx, y, *incy, a, *lda );
}
inline
void gerc (CBLAS_ORDER const Order, int const M, int const N,
traits::complex_d const& alpha,
traits::complex_d const* X, int const incX,
traits::complex_d const* Y, int const incY,
traits::complex_d* A, int const lda)
{
cblas_zgerc (Order, M, N,
static_cast<void const*> (&alpha),
static_cast<void const*> (X), incX,
static_cast<void const*> (Y), incY,
static_cast<void*> (A), lda);
}
void phi_gerc(const int N,
const Complex *alpha, const Complex *X,
const Complex *Y, Complex *A){
#ifndef NOBLAS
#ifdef SINGLEPRECISION
cblas_cgerc(CblasRowMajor,N,N,alpha,X,1,Y,1,A,N);
#else
cblas_zgerc(CblasRowMajor,N,N,alpha,X,1,Y,1,A,N);
#endif
#else
int i,j;
for(i = 0; i < N; ++i){
for(j = 0; j < N; ++j){
A[i*N+j] = (*alpha)*X[i]*Y[j];
}
}
#endif
}
int CORE_zttlqt(int M, int N, int IB,
PLASMA_Complex64_t *A1, int LDA1,
PLASMA_Complex64_t *A2, int LDA2,
PLASMA_Complex64_t *T, int LDT,
PLASMA_Complex64_t *TAU, PLASMA_Complex64_t *WORK)
{
static PLASMA_Complex64_t zone = 1.0;
static PLASMA_Complex64_t zzero = 0.0;
#ifdef COMPLEX
static int ione = 1;
#endif
PLASMA_Complex64_t alpha;
int i, j, l, ii, sb, mi, ni;
/* Check input arguments */
if (M < 0) {
coreblas_error(1, "Illegal value of M");
return -1;
}
if (N < 0) {
coreblas_error(2, "Illegal value of N");
return -2;
}
if (IB < 0) {
coreblas_error(3, "Illegal value of IB");
return -3;
}
if ((LDA2 < max(1,M)) && (M > 0)) {
coreblas_error(7, "Illegal value of LDA2");
return -7;
}
/* Quick return */
if ((M == 0) || (N == 0) || (IB == 0))
return PLASMA_SUCCESS;
/* TODO: Need to check why some cases require
* this to not have uninitialized values */
CORE_zlaset( PlasmaUpperLower, IB, N,
0., 0., T, LDT);
for(ii = 0; ii < M; ii += IB) {
sb = min(M-ii, IB);
for(i = 0; i < sb; i++) {
j = ii + i;
mi = sb-i-1;
ni = min( j + 1, N);
/*
* Generate elementary reflector H( II*IB+I ) to annihilate A( II*IB+I, II*IB+I:M ).
*/
#ifdef COMPLEX
LAPACKE_zlacgv_work(ni, &A2[j], LDA2);
LAPACKE_zlacgv_work(ione, &A1[LDA1*j+j], LDA1);
#endif
LAPACKE_zlarfg_work(ni+1, &A1[LDA1*j+j], &A2[j], LDA2, &TAU[j]);
if (mi > 0) {
/*
* Apply H( j-1 ) to A( j:II+IB-1, j-1:M ) from the right.
*/
cblas_zcopy(
mi,
&A1[LDA1*j+(j+1)], 1,
WORK, 1);
cblas_zgemv(
CblasColMajor, (CBLAS_TRANSPOSE)PlasmaNoTrans,
mi, ni,
CBLAS_SADDR(zone), &A2[j+1], LDA2,
&A2[j], LDA2,
CBLAS_SADDR(zone), WORK, 1);
alpha = -(TAU[j]);
cblas_zaxpy(
mi, CBLAS_SADDR(alpha),
WORK, 1,
&A1[LDA1*j+j+1], 1);
cblas_zgerc(
CblasColMajor, mi, ni,
CBLAS_SADDR(alpha), WORK, 1,
&A2[j], LDA2,
&A2[j+1], LDA2);
}
/*
* Calculate T.
*/
if (i > 0 ) {
l = min(i, max(0, N-ii));
alpha = -(TAU[j]);
CORE_zpemv(
PlasmaNoTrans, PlasmaRowwise,
i , min(j, N), l,
alpha, &A2[ii], LDA2,
&A2[j], LDA2,
zzero, &T[LDT*j], 1,
WORK);
/* T(0:i-1, j) = T(0:i-1, ii:j-1) * T(0:i-1, j) */
cblas_ztrmv(
CblasColMajor, (CBLAS_UPLO)PlasmaUpper,
(CBLAS_TRANSPOSE)PlasmaNoTrans,
(CBLAS_DIAG)PlasmaNonUnit,
i, &T[LDT*ii], LDT,
&T[LDT*j], 1);
}
#ifdef COMPLEX
LAPACKE_zlacgv_work(ni, &A2[j], LDA2 );
LAPACKE_zlacgv_work(ione, &A1[LDA1*j+j], LDA1 );
#endif
T[LDT*j+i] = TAU[j];
}
/* Apply Q to the rest of the matrix to the right */
if (M > ii+sb) {
mi = M-(ii+sb);
ni = min(ii+sb, N);
l = min(sb, max(0, ni-ii));
CORE_zparfb(
PlasmaRight, PlasmaNoTrans,
PlasmaForward, PlasmaRowwise,
mi, IB, mi, ni, sb, l,
&A1[LDA1*ii+ii+sb], LDA1,
&A2[ii+sb], LDA2,
&A2[ii], LDA2,
&T[LDT*ii], LDT,
WORK, M);
}
}
return PLASMA_SUCCESS;
}
int CORE_zttqrt(int M, int N, int IB,
PLASMA_Complex64_t *A1, int LDA1,
PLASMA_Complex64_t *A2, int LDA2,
PLASMA_Complex64_t *T, int LDT,
PLASMA_Complex64_t *TAU, PLASMA_Complex64_t *WORK)
{
static PLASMA_Complex64_t zone = 1.0;
static PLASMA_Complex64_t zzero = 0.0;
static int ione = 1;
PLASMA_Complex64_t alpha;
int i, j, ii, sb, mi, ni;
/* Check input arguments */
if (M < 0) {
coreblas_error(1, "Illegal value of M");
return -1;
}
if (N < 0) {
coreblas_error(2, "Illegal value of N");
return -2;
}
if (IB < 0) {
coreblas_error(3, "Illegal value of IB");
return -3;
}
if ((LDA2 < max(1,M)) && (M > 0)) {
coreblas_error(7, "Illegal value of LDA2");
return -7;
}
/* Quick return */
if ((M == 0) || (N == 0) || (IB == 0))
return PLASMA_SUCCESS;
for(ii = 0; ii < N; ii += IB) {
sb = min(N-ii, IB);
for(i = 0; i < sb; i++) {
/*
* Generate elementary reflector H( II*IB+I ) to annihilate
* A( II*IB+I:mi, II*IB+I ).
*/
mi = ii + i + 1;
LAPACKE_zlarfg_work(mi+1, &A1[LDA1*(ii+i)+ii+i], &A2[LDA2*(ii+i)], ione, &TAU[ii+i]);
if (sb-i-1>0) {
/*
* Apply H( II*IB+I ) to A( II*IB+I:M, II*IB+I+1:II*IB+IB ) from the left.
*/
ni = sb-i-1;
cblas_zcopy(
ni,
&A1[LDA1*(ii+i+1)+(ii+i)], LDA1,
WORK, 1);
#ifdef COMPLEX
LAPACKE_zlacgv_work(ni, WORK, ione);
#endif
cblas_zgemv(
CblasColMajor, (CBLAS_TRANSPOSE)PlasmaConjTrans,
mi, ni,
CBLAS_SADDR(zone), &A2[LDA2*(ii+i+1)], LDA2,
&A2[LDA2*(ii+i)], 1,
CBLAS_SADDR(zone), WORK, 1);
#ifdef COMPLEX
LAPACKE_zlacgv_work(ni, WORK, ione);
#endif
alpha = -conj(TAU[ii+i]);
cblas_zaxpy(
ni, CBLAS_SADDR(alpha),
WORK, 1,
&A1[LDA1*(ii+i+1)+ii+i], LDA1);
#ifdef COMPLEX
LAPACKE_zlacgv_work(ni, WORK, ione);
#endif
cblas_zgerc(
CblasColMajor, mi, ni,
CBLAS_SADDR(alpha), &A2[LDA2*(ii+i)], 1,
WORK, 1,
&A2[LDA2*(ii+i+1)], LDA2);
}
/*
* Calculate T.
*/
if (i > 0 ) {
cblas_zcopy(i, &A2[LDA2*(ii+i)+ii], 1, &WORK[ii], 1);
cblas_ztrmv(
CblasColMajor, (CBLAS_UPLO)PlasmaUpper,
(CBLAS_TRANSPOSE)PlasmaConjTrans, (CBLAS_DIAG)PlasmaNonUnit,
i, &A2[LDA2*ii+ii], LDA2,
&WORK[ii], 1);
alpha = -(TAU[ii+i]);
for(j = 0; j < i; j++) {
WORK[ii+j] = alpha * WORK[ii+j];
}
if (ii > 0) {
cblas_zgemv(
CblasColMajor, (CBLAS_TRANSPOSE)PlasmaConjTrans, ii, i,
CBLAS_SADDR(alpha), &A2[LDA2*ii], LDA2,
&A2[LDA2*(ii+i)], 1,
CBLAS_SADDR(zzero), WORK, 1);
cblas_zaxpy(i, CBLAS_SADDR(zone), &WORK[ii], 1, WORK, 1);
}
cblas_zcopy(i, WORK, 1, &T[LDT*(ii+i)], 1);
cblas_ztrmv(
CblasColMajor, (CBLAS_UPLO)PlasmaUpper,
(CBLAS_TRANSPOSE)PlasmaNoTrans, (CBLAS_DIAG)PlasmaNonUnit,
i, &T[LDT*ii], LDT,
&T[LDT*(ii+i)], 1);
}
T[LDT*(ii+i)+i] = TAU[ii+i];
}
/* Apply Q' to the rest of the matrix to the left */
if (N > ii+sb) {
CORE_zttrfb(
PlasmaLeft, PlasmaConjTrans,
PlasmaForward, PlasmaColumnwise,
sb, N-(ii+sb), ii+sb, N-(ii+sb), sb,
&A1[LDA1*(ii+sb)+ii], LDA1,
&A2[LDA2*(ii+sb)], LDA2,
&A2[LDA2*ii], LDA2,
&T[LDT*ii], LDT,
WORK, sb);
}
}
return PLASMA_SUCCESS;
}
int CORE_ztsqrt(int M, int N, int IB,
PLASMA_Complex64_t *A1, int LDA1,
PLASMA_Complex64_t *A2, int LDA2,
PLASMA_Complex64_t *T, int LDT,
PLASMA_Complex64_t *TAU, PLASMA_Complex64_t *WORK)
{
static PLASMA_Complex64_t zone = 1.0;
static PLASMA_Complex64_t zzero = 0.0;
PLASMA_Complex64_t alpha;
int i, ii, sb;
/* Check input arguments */
if (M < 0) {
coreblas_error(1, "Illegal value of M");
return -1;
}
if (N < 0) {
coreblas_error(2, "Illegal value of N");
return -2;
}
if (IB < 0) {
coreblas_error(3, "Illegal value of IB");
return -3;
}
if ((LDA2 < max(1,M)) && (M > 0)) {
coreblas_error(8, "Illegal value of LDA2");
return -8;
}
/* Quick return */
if ((M == 0) || (N == 0) || (IB == 0))
return PLASMA_SUCCESS;
for(ii = 0; ii < N; ii += IB) {
sb = min(N-ii, IB);
for(i = 0; i < sb; i++) {
/*
* Generate elementary reflector H( II*IB+I ) to annihilate
* A( II*IB+I:M, II*IB+I )
*/
LAPACKE_zlarfg_work(M+1, &A1[LDA1*(ii+i)+ii+i], &A2[LDA2*(ii+i)], 1, &TAU[ii+i]);
if (ii+i+1 < N) {
/*
* Apply H( II*IB+I ) to A( II*IB+I:M, II*IB+I+1:II*IB+IB ) from the left
*/
alpha = -conj(TAU[ii+i]);
cblas_zcopy(
sb-i-1,
&A1[LDA1*(ii+i+1)+(ii+i)], LDA1,
WORK, 1);
#ifdef COMPLEX
LAPACKE_zlacgv_work(sb-i-1, WORK, 1);
#endif
cblas_zgemv(
CblasColMajor, (CBLAS_TRANSPOSE)PlasmaConjTrans,
M, sb-i-1,
CBLAS_SADDR(zone), &A2[LDA2*(ii+i+1)], LDA2,
&A2[LDA2*(ii+i)], 1,
CBLAS_SADDR(zone), WORK, 1);
#ifdef COMPLEX
LAPACKE_zlacgv_work(sb-i-1, WORK, 1 );
#endif
cblas_zaxpy(
sb-i-1, CBLAS_SADDR(alpha),
WORK, 1,
&A1[LDA1*(ii+i+1)+ii+i], LDA1);
#ifdef COMPLEX
LAPACKE_zlacgv_work(sb-i-1, WORK, 1 );
#endif
cblas_zgerc(
CblasColMajor, M, sb-i-1, CBLAS_SADDR(alpha),
&A2[LDA2*(ii+i)], 1,
WORK, 1,
&A2[LDA2*(ii+i+1)], LDA2);
}
/*
* Calculate T
*/
alpha = -TAU[ii+i];
cblas_zgemv(
CblasColMajor, (CBLAS_TRANSPOSE)PlasmaConjTrans, M, i,
CBLAS_SADDR(alpha), &A2[LDA2*ii], LDA2,
&A2[LDA2*(ii+i)], 1,
CBLAS_SADDR(zzero), &T[LDT*(ii+i)], 1);
cblas_ztrmv(
CblasColMajor, (CBLAS_UPLO)PlasmaUpper,
(CBLAS_TRANSPOSE)PlasmaNoTrans, (CBLAS_DIAG)PlasmaNonUnit, i,
&T[LDT*ii], LDT,
&T[LDT*(ii+i)], 1);
T[LDT*(ii+i)+i] = TAU[ii+i];
}
if (N > ii+sb) {
CORE_ztsmqr(
PlasmaLeft, PlasmaConjTrans,
sb, N-(ii+sb), M, N-(ii+sb), IB, IB,
&A1[LDA1*(ii+sb)+ii], LDA1,
&A2[LDA2*(ii+sb)], LDA2,
&A2[LDA2*ii], LDA2,
&T[LDT*ii], LDT,
WORK, sb);
}
}
return PLASMA_SUCCESS;
}
void
test_ger (void) {
const double flteps = 1e-4, dbleps = 1e-6;
{
int order = 101;
int M = 1;
int N = 1;
int lda = 1;
float alpha = 1.0f;
float A[] = { -0.515f };
float X[] = { 0.611f };
int incX = -1;
float Y[] = { -0.082f };
int incY = -1;
float A_expected[] = { -0.565102f };
cblas_sger(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[i], A_expected[i], flteps, "sger(case 1390)");
}
};
};
{
int order = 102;
int M = 1;
int N = 1;
int lda = 1;
float alpha = 1.0f;
float A[] = { -0.515f };
float X[] = { 0.611f };
int incX = -1;
float Y[] = { -0.082f };
int incY = -1;
float A_expected[] = { -0.565102f };
cblas_sger(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[i], A_expected[i], flteps, "sger(case 1391)");
}
};
};
{
int order = 101;
int M = 1;
int N = 1;
int lda = 1;
double alpha = 1;
double A[] = { -0.809 };
double X[] = { -0.652 };
int incX = -1;
double Y[] = { 0.712 };
int incY = -1;
double A_expected[] = { -1.273224 };
cblas_dger(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[i], A_expected[i], dbleps, "dger(case 1392)");
}
};
};
{
int order = 102;
int M = 1;
int N = 1;
int lda = 1;
double alpha = 1;
double A[] = { -0.809 };
double X[] = { -0.652 };
int incX = -1;
double Y[] = { 0.712 };
int incY = -1;
double A_expected[] = { -1.273224 };
cblas_dger(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[i], A_expected[i], dbleps, "dger(case 1393)");
}
};
};
{
int order = 101;
int M = 1;
int N = 1;
int lda = 1;
float alpha[2] = {0.0f, 0.0f};
float A[] = { -0.651f, 0.856f };
float X[] = { -0.38f, -0.235f };
int incX = -1;
float Y[] = { -0.627f, 0.757f };
int incY = -1;
float A_expected[] = { -0.651f, 0.856f };
cblas_cgeru(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], flteps, "cgeru(case 1394) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], flteps, "cgeru(case 1394) imag");
};
};
};
{
int order = 101;
int M = 1;
int N = 1;
int lda = 1;
float alpha[2] = {0.0f, 0.0f};
float A[] = { -0.651f, 0.856f };
float X[] = { -0.38f, -0.235f };
int incX = -1;
float Y[] = { -0.627f, 0.757f };
int incY = -1;
float A_expected[] = { -0.651f, 0.856f };
cblas_cgerc(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], flteps, "cgerc(case 1395) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], flteps, "cgerc(case 1395) imag");
};
};
};
{
int order = 102;
int M = 1;
int N = 1;
int lda = 1;
float alpha[2] = {0.0f, 0.0f};
float A[] = { -0.651f, 0.856f };
float X[] = { -0.38f, -0.235f };
int incX = -1;
float Y[] = { -0.627f, 0.757f };
int incY = -1;
float A_expected[] = { -0.651f, 0.856f };
cblas_cgeru(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], flteps, "cgeru(case 1396) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], flteps, "cgeru(case 1396) imag");
};
};
};
{
int order = 102;
int M = 1;
int N = 1;
int lda = 1;
float alpha[2] = {0.0f, 0.0f};
float A[] = { -0.651f, 0.856f };
float X[] = { -0.38f, -0.235f };
int incX = -1;
float Y[] = { -0.627f, 0.757f };
int incY = -1;
float A_expected[] = { -0.651f, 0.856f };
cblas_cgerc(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], flteps, "cgerc(case 1397) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], flteps, "cgerc(case 1397) imag");
};
};
};
{
int order = 101;
int M = 1;
int N = 1;
int lda = 1;
double alpha[2] = {-1, 0};
double A[] = { -0.426, 0.757 };
double X[] = { -0.579, -0.155 };
int incX = -1;
double Y[] = { 0.831, 0.035 };
int incY = -1;
double A_expected[] = { 0.049724, 0.90607 };
cblas_zgeru(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], dbleps, "zgeru(case 1398) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], dbleps, "zgeru(case 1398) imag");
};
};
};
{
int order = 101;
int M = 1;
int N = 1;
int lda = 1;
double alpha[2] = {-1, 0};
double A[] = { -0.426, 0.757 };
double X[] = { -0.579, -0.155 };
int incX = -1;
double Y[] = { 0.831, 0.035 };
int incY = -1;
double A_expected[] = { 0.060574, 0.86554 };
cblas_zgerc(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], dbleps, "zgerc(case 1399) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], dbleps, "zgerc(case 1399) imag");
};
};
};
{
int order = 102;
int M = 1;
int N = 1;
int lda = 1;
double alpha[2] = {-1, 0};
double A[] = { -0.426, 0.757 };
double X[] = { -0.579, -0.155 };
int incX = -1;
double Y[] = { 0.831, 0.035 };
int incY = -1;
double A_expected[] = { 0.049724, 0.90607 };
cblas_zgeru(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], dbleps, "zgeru(case 1400) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], dbleps, "zgeru(case 1400) imag");
};
};
};
{
int order = 102;
int M = 1;
int N = 1;
int lda = 1;
double alpha[2] = {-1, 0};
double A[] = { -0.426, 0.757 };
double X[] = { -0.579, -0.155 };
int incX = -1;
double Y[] = { 0.831, 0.035 };
int incY = -1;
double A_expected[] = { 0.060574, 0.86554 };
cblas_zgerc(order, M, N, alpha, X, incX, Y, incY, A, lda);
{
int i;
for (i = 0; i < 1; i++) {
gsl_test_rel(A[2*i], A_expected[2*i], dbleps, "zgerc(case 1401) real");
gsl_test_rel(A[2*i+1], A_expected[2*i+1], dbleps, "zgerc(case 1401) imag");
};
};
};
}