Newer
Older
* @copyright 2009-2014 The University of Tennessee and The University of
* Tennessee Research Foundation. All rights reserved.
* @copyright 2012-2023 Bordeaux INP, CNRS (LaBRI UMR 5800), Inria,
* Univ. Bordeaux. All rights reserved.
* @author Hatem Ltaief
* @author Jakub Kurzak
* @author Mathieu Faverge
* @author Emmanuel Agullo
* @author Cedric Castagnede
* @author Gwenole Lucas
* @precisions normal z -> c
*

Mathieu Faverge
committed
*/
#include "chameleon_starpu.h"
#include "runtime_codelet_z.h"
struct cl_zherk_args_s {
int n;
int k;
double alpha;
double beta;
};
#if !defined(CHAMELEON_SIMULATION)
static void
cl_zherk_cpu_func(void *descr[], void *cl_arg)
{
struct cl_zherk_args_s *clargs = (struct cl_zherk_args_s *)cl_arg;
CHAM_tile_t *tileA;
CHAM_tile_t *tileC;

Mathieu Faverge
committed
tileA = cti_interface_get(descr[0]);
tileC = cti_interface_get(descr[1]);
TCORE_zherk( clargs->uplo, clargs->trans, clargs->n, clargs->k,
clargs->alpha, tileA, clargs->beta, tileC );
#if defined(CHAMELEON_USE_CUDA)
static void
cl_zherk_cuda_func(void *descr[], void *cl_arg)
cublasHandle_t handle = starpu_cublas_get_local_handle();
struct cl_zherk_args_s *clargs = (struct cl_zherk_args_s *)cl_arg;

Mathieu Faverge
committed
CHAM_tile_t *tileA;
CHAM_tile_t *tileC;

Mathieu Faverge
committed
tileA = cti_interface_get(descr[0]);
tileC = cti_interface_get(descr[1]);

PRUVOST Florent
committed
CUDA_zherk(
clargs->uplo, clargs->trans, clargs->n, clargs->k,
&(clargs->alpha),
tileA->mat, tileA->ld,
&(clargs->beta),
tileC->mat, tileC->ld,
#endif /* defined(CHAMELEON_USE_CUDA) */
#if defined(CHAMELEON_USE_HIP)
static void
cl_zherk_hip_func(void *descr[], void *cl_arg)
{
hipblasHandle_t handle = starpu_hipblas_get_local_handle();
struct cl_zherk_args_s *clargs = (struct cl_zherk_args_s *)cl_arg;
CHAM_tile_t *tileA;
CHAM_tile_t *tileC;
tileA = cti_interface_get(descr[0]);
tileC = cti_interface_get(descr[1]);
HIP_zherk(
clargs->uplo, clargs->trans, clargs->n, clargs->k,
&(clargs->alpha),
tileA->mat, tileA->ld,
&(clargs->beta),
tileC->mat, tileC->ld,
handle );
}
#endif /* defined(CHAMELEON_USE_HIP) */
#endif /* !defined(CHAMELEON_SIMULATION) */
/*
* Codelet definition
*/
#if defined(CHAMELEON_USE_HIP)
CODELETS_GPU( zherk, cl_zherk_cpu_func, cl_zherk_hip_func, STARPU_HIP_ASYNC )
#else
CODELETS( zherk, cl_zherk_cpu_func, cl_zherk_cuda_func, STARPU_CUDA_ASYNC )
void INSERT_TASK_zherk( const RUNTIME_option_t *options,
cham_uplo_t uplo, cham_trans_t trans,
int n, int k, int nb,
double alpha, const CHAM_desc_t *A, int Am, int An,
double beta, const CHAM_desc_t *C, int Cm, int Cn )

Mathieu Faverge
committed
if ( alpha == 0. ) {
INSERT_TASK_zlascal( options, uplo, n, n, nb,
beta, C, Cm, Cn );
return;

Mathieu Faverge
committed
}
struct cl_zherk_args_s *clargs = NULL;
void (*callback)(void*);
int accessC;
char *cl_name = "zherk";
/* Handle cache */
CHAMELEON_BEGIN_ACCESS_DECLARATION;
CHAMELEON_ACCESS_R(A, Am, An);
CHAMELEON_ACCESS_RW(C, Cm, Cn);
exec = __chameleon_need_exec;
if ( exec ) {
clargs = malloc( sizeof( struct cl_zherk_args_s ) );
clargs->uplo = uplo;
clargs->trans = trans;
clargs->n = n;
clargs->k = k;
clargs->alpha = alpha;
clargs->beta = beta;
}
/* Callback fro profiling information */
callback = options->profiling ? cl_zherk_callback : NULL;
/* Reduce the C access if needed */
accessC = ( beta == 0. ) ? STARPU_W : STARPU_RW;

Mathieu Faverge
committed
/* Refine name */
cl_name = chameleon_codelet_name( cl_name, 2,
A->get_blktile( A, Am, An ),
C->get_blktile( C, Cm, Cn ) );
/* Insert the task */

Mathieu Faverge
committed
rt_starpu_insert_task(
&cl_zherk,
/* Task codelet arguments */
STARPU_CL_ARGS, clargs, sizeof(struct cl_zherk_args_s),
STARPU_R, RTBLKADDR(A, ChamComplexDouble, Am, An),
accessC, RTBLKADDR(C, ChamComplexDouble, Cm, Cn),
/* Common task arguments */
STARPU_PRIORITY, options->priority,
STARPU_CALLBACK, callback,
STARPU_EXECUTE_ON_WORKER, options->workerid,
#if defined(CHAMELEON_CODELETS_HAVE_NAME)
STARPU_NAME, cl_name,
0 );
(void)nb;