void magma_zsyrk(
magma_uplo_t uplo, magma_trans_t trans,
magma_int_t n, magma_int_t k,
cuDoubleComplex alpha, cuDoubleComplex const* dA, magma_int_t lda,
cuDoubleComplex beta, cuDoubleComplex* dC, magma_int_t ldc )
{
cublasZsyrk(
cublas_uplo_const( uplo ),
cublas_trans_const( trans ),
n, k,
alpha, dA, lda,
beta, dC, ldc );
}
/** Perform symmetric rank-k update.
\f$ C = \alpha A A^T + \beta C \f$ (trans == MagmaNoTrans), or \n
\f$ C = \alpha A^T A + \beta C \f$ (trans == MagmaTrans), \n
where \f$ C \f$ is symmetric.
@param[in]
uplo Whether the upper or lower triangle of C is referenced.
@param[in]
trans Operation to perform on A.
@param[in]
n Number of rows and columns of C. n >= 0.
@param[in]
k Number of columns of A (for MagmaNoTrans) or rows of A (for MagmaTrans). k >= 0.
@param[in]
alpha Scalar \f$ \alpha \f$
@param[in]
dA COMPLEX_16 array on GPU device.
If trans == MagmaNoTrans, the n-by-k matrix A of dimension (ldda,k), ldda >= max(1,n); \n
otherwise, the k-by-n matrix A of dimension (ldda,n), ldda >= max(1,k).
@param[in]
ldda Leading dimension of dA.
@param[in]
beta Scalar \f$ \beta \f$
@param[in,out]
dC COMPLEX_16 array on GPU device.
The n-by-n symmetric matrix C of dimension (lddc,n), lddc >= max(1,n).
@param[in]
lddc Leading dimension of dC.
@ingroup magma_zblas3
*/
extern "C" void
magma_zsyrk(
magma_uplo_t uplo, magma_trans_t trans,
magma_int_t n, magma_int_t k,
magmaDoubleComplex alpha,
magmaDoubleComplex_const_ptr dA, magma_int_t ldda,
magmaDoubleComplex beta,
magmaDoubleComplex_ptr dC, magma_int_t lddc )
{
cublasZsyrk(
cublas_uplo_const( uplo ),
cublas_trans_const( trans ),
n, k,
alpha, dA, ldda,
beta, dC, lddc );
}
FLA_Error FLA_Syrk_external_gpu( FLA_Uplo uplo, FLA_Trans trans, FLA_Obj alpha, FLA_Obj A, void* A_gpu, FLA_Obj beta, FLA_Obj C, void* C_gpu )
{
FLA_Datatype datatype;
int k_A;
int m_A, n_A;
int m_C;
int ldim_A;
int ldim_C;
char blas_uplo;
char blas_trans;
if ( FLA_Check_error_level() == FLA_FULL_ERROR_CHECKING )
FLA_Syrk_check( uplo, trans, alpha, A, beta, C );
if ( FLA_Obj_has_zero_dim( C ) ) return FLA_SUCCESS;
datatype = FLA_Obj_datatype( A );
m_A = FLA_Obj_length( A );
n_A = FLA_Obj_width( A );
ldim_A = FLA_Obj_length( A );
m_C = FLA_Obj_length( C );
ldim_C = FLA_Obj_length( C );
if ( trans == FLA_NO_TRANSPOSE )
k_A = n_A;
else
k_A = m_A;
FLA_Param_map_flame_to_netlib_uplo( uplo, &blas_uplo );
FLA_Param_map_flame_to_netlib_trans( trans, &blas_trans );
switch( datatype ){
case FLA_FLOAT:
{
float *buff_alpha = ( float * ) FLA_FLOAT_PTR( alpha );
float *buff_beta = ( float * ) FLA_FLOAT_PTR( beta );
cublasSsyrk( blas_uplo,
blas_trans,
m_C,
k_A,
*buff_alpha,
( float * ) A_gpu, ldim_A,
*buff_beta,
( float * ) C_gpu, ldim_C );
break;
}
case FLA_DOUBLE:
{
double *buff_alpha = ( double * ) FLA_DOUBLE_PTR( alpha );
double *buff_beta = ( double * ) FLA_DOUBLE_PTR( beta );
cublasDsyrk( blas_uplo,
blas_trans,
m_C,
k_A,
*buff_alpha,
( double * ) A_gpu, ldim_A,
*buff_beta,
( double * ) C_gpu, ldim_C );
break;
}
case FLA_COMPLEX:
{
cuComplex *buff_alpha = ( cuComplex * ) FLA_COMPLEX_PTR( alpha );
cuComplex *buff_beta = ( cuComplex * ) FLA_COMPLEX_PTR( beta );
cublasCsyrk( blas_uplo,
blas_trans,
m_C,
k_A,
*buff_alpha,
( cuComplex * ) A_gpu, ldim_A,
*buff_beta,
( cuComplex * ) C_gpu, ldim_C );
break;
}
case FLA_DOUBLE_COMPLEX:
{
cuDoubleComplex *buff_alpha = ( cuDoubleComplex * ) FLA_DOUBLE_COMPLEX_PTR( alpha );
cuDoubleComplex *buff_beta = ( cuDoubleComplex * ) FLA_DOUBLE_COMPLEX_PTR( beta );
cublasZsyrk( blas_uplo,
blas_trans,
m_C,
k_A,
*buff_alpha,
( cuDoubleComplex * ) A_gpu, ldim_A,
*buff_beta,
( cuDoubleComplex * ) C_gpu, ldim_C );
break;
}
}
return FLA_SUCCESS;
}
