Mentions légales du service

Skip to content
Snippets Groups Projects
Commit f0dfcdb7 authored by Mathieu Faverge's avatar Mathieu Faverge
Browse files

starpu/codelets/gemmex: add gemmex codelet

parent 20eab8fa
No related branches found
No related tags found
1 merge request!401Add all the architecture to manage multi-precision data descriptors.
Hide whitespace changes
Inline Side-by-side
include/chameleon/tasks.h
+ 7
− 0
View file @ f0dfcdb7
......@@ -107,6 +107,13 @@ void INSERT_TASK_gemm( const RUNTIME_option_t *options,
const CHAM_desc_t *B, int Bm, int Bn,
double beta, const CHAM_desc_t *C, int Cm, int Cn );
void INSERT_TASK_gemmex( const RUNTIME_option_t *options,
cham_trans_t transA, cham_trans_t transB,
int m, int n, int k, int nb,
double alpha, const CHAM_desc_t *A, int Am, int An,
const CHAM_desc_t *B, int Bm, int Bn,
double beta, const CHAM_desc_t *C, int Cm, int Cn );
void INSERT_TASK_hgemm( const RUNTIME_option_t *options,
cham_trans_t transA, cham_trans_t transB,
int m, int n, int k, int nb,
......
runtime/starpu/codelets/codelet_gemmex.c 0 → 100644
+ 278
− 0
View file @ f0dfcdb7
/**
*
* @file starpu/codelet_gemmex.c
*
* @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.
*
***
*
* @brief Chameleon gemmex StarPU codelet
*
* @version 1.3.0
* @author Mathieu Faverge
* @date 2023-07-06
*
*/
#include "chameleon_starpu.h"
#include "runtime_codelets.h"
CHAMELEON_CL_CB( gemmex, cti_handle_get_m(task->handles[2]), cti_handle_get_n(task->handles[2]), cti_handle_get_n(task->handles[0]), 2. *M*N*K) /* If A^t, computation is wrong */
struct cl_gemmex_args_s {
cham_trans_t transA;
cham_trans_t transB;
int m;
int n;
int k;
double alpha;
double beta;
};
#if !defined(CHAMELEON_SIMULATION)
#if defined(CHAMELEON_USE_CUDA)
static void
cl_gemmex_cuda_func( void *descr[], void *cl_arg )
{
struct cl_gemmex_args_s *clargs = (struct cl_gemmex_args_s *)cl_arg;
cublasHandle_t handle = starpu_cublas_get_local_handle();
CHAM_tile_t *tileA;
CHAM_tile_t *tileB;
CHAM_tile_t *tileC;
void *ptrAlpha, *ptrBeta;
switch( tileC->flttype ) {
case ChamRealHalf:
{
CHAMELEON_Real16_t halpha = clargs->alpha;
CHAMELEON_Real16_t hbeta = clargs->beta;
ptrAlpha = &halpha;
ptrBeta = &hbeta;
}
break;
case ChamRealFloat:
{
float salpha = clargs->alpha;
float sbeta = clargs->beta;
ptrAlpha = &salpha;
ptrBeta = &sbeta;
}
break;
case ChamRealDouble:
{
double dalpha = clargs->alpha;
double dbeta = clargs->beta;
ptrAlpha = &dalpha;
ptrBeta = &dbeta;
}
break;
case ChamComplexFloat:
{
CHAMELEON_Complex32_t calpha = clargs->alpha;
CHAMELEON_Complex32_t cbeta = clargs->beta;
ptrAlpha = &calpha;
ptrBeta = &cbeta;
}
break;
case ChamComplexDouble:
{
CHAMELEON_Complex64_t zalpha = clargs->alpha;
CHAMELEON_Complex64_t zbeta = clargs->beta;
ptrAlpha = &zalpha;
ptrBeta = &zbeta;
}
break;
default:
fprintf( stderr, "cl_gemmex: Unknown C datatype\n" );
return;
}
tileA = cti_interface_get(descr[0]);
tileB = cti_interface_get(descr[1]);
tileC = cti_interface_get(descr[2]);
assert( tileA->format & CHAMELEON_TILE_FULLRANK );
assert( tileB->format & CHAMELEON_TILE_FULLRANK );
assert( tileC->format & CHAMELEON_TILE_FULLRANK );
CUDA_gemmex(
clargs->transA, clargs->transB,
clargs->m, clargs->n, clargs->k,
ptrAlpha,
tileA->mat, tileA->ld, tileA->flttype,
tileB->mat, tileB->ld, tileB->flttype,
ptrBeta,
tileC->mat, tileC->ld, tileC->flttype,
handle );
}
#endif /* defined(CHAMELEON_USE_CUDA) */
#endif /* !defined(CHAMELEON_SIMULATION) */
/*
* Codelet definition
*/
CODELETS( gemmex, NULL, cl_gemmex_cuda_func, STARPU_CUDA_ASYNC )
void INSERT_TASK_gemmex_Astat( const RUNTIME_option_t *options,
cham_trans_t transA, cham_trans_t transB,
int m, int n, int k, int nb,
double alpha, const CHAM_desc_t *A, int Am, int An,
const CHAM_desc_t *B, int Bm, int Bn,
double beta, const CHAM_desc_t *C, int Cm, int Cn )
{
/* if ( alpha == 0. ) { */
/* INSERT_TASK_hlascal( options, ChamUpperLower, m, n, nb, */
/* beta, C, Cm, Cn ); */
/* return; */
/* } */
struct cl_gemmex_args_s *clargs = NULL;
void (*callback)(void*);
int accessC;
int exec = 0;
char *cl_name = "gemmex_Astat";
/* Handle cache */
CHAMELEON_BEGIN_ACCESS_DECLARATION;
/* Check A as write, since it will be the owner of the computation */
CHAMELEON_ACCESS_W(A, Am, An);
CHAMELEON_ACCESS_R(B, Bm, Bn);
/* Check C as read, since it will be used in a reduction */
CHAMELEON_ACCESS_R(C, Cm, Cn);
exec = __chameleon_need_exec;
CHAMELEON_END_ACCESS_DECLARATION;
if ( exec ) {
clargs = malloc( sizeof( struct cl_gemmex_args_s ) );
clargs->transA = transA;
clargs->transB = transB;
clargs->m = m;
clargs->n = n;
clargs->k = k;
clargs->alpha = alpha;
clargs->beta = beta;
}
/* Callback for profiling information */
callback = options->profiling ? cl_gemmex_callback : NULL;
/* Reduce the C access if needed */
if ( beta == 0. ) {
accessC = STARPU_W;
}
#if defined(HAVE_STARPU_MPI_REDUX)
else if ( beta == 1. ) {
accessC = STARPU_MPI_REDUX;
}
#endif
else {
accessC = STARPU_RW;
}
/* Refine name */
cl_name = chameleon_codelet_name( cl_name, 3,
A->get_blktile( A, Am, An ),
B->get_blktile( B, Bm, Bn ),
C->get_blktile( C, Cm, Cn ) );
/* Insert the task */
rt_starpu_insert_task(
&cl_gemmex,
/* Task codelet arguments */
STARPU_CL_ARGS, clargs, sizeof(struct cl_gemmex_args_s),
/* Task handles */
STARPU_R, RTBLKADDR(A, ChamRealHalf, Am, An),
STARPU_R, RTBLKADDR(B, ChamRealHalf, Bm, Bn),
accessC, RTBLKADDR(C, ChamRealHalf, Cm, Cn),
/* Common task arguments */
STARPU_PRIORITY, options->priority,
STARPU_CALLBACK, callback,
STARPU_EXECUTE_ON_NODE, A->get_rankof(A, Am, An),
#if defined(CHAMELEON_CODELETS_HAVE_NAME)
STARPU_NAME, cl_name,
#endif
0 );
}
void INSERT_TASK_gemmex( const RUNTIME_option_t *options,
cham_trans_t transA, cham_trans_t transB,
int m, int n, int k, int nb,
double alpha, const CHAM_desc_t *A, int Am, int An,
const CHAM_desc_t *B, int Bm, int Bn,
double beta, const CHAM_desc_t *C, int Cm, int Cn )
{
/* if ( alpha == 0. ) { */
/* INSERT_TASK_zlascal( options, ChamUpperLower, m, n, nb, */
/* beta, C, Cm, Cn ); */
/* return; */
/* } */
struct cl_gemmex_args_s *clargs = NULL;
void (*callback)(void*);
int accessC;
int exec = 0;
char *cl_name = "gemmex";
if ( !(options->withcuda) ) {
/* Fallback to cpu version */
INSERT_TASK_gemm( options, transA, transB, m, n, k, nb,
alpha, A, Am, An, B, Bm, Bn, beta, C, Cm, Cn );
return;
}
/* Handle cache */
CHAMELEON_BEGIN_ACCESS_DECLARATION;
CHAMELEON_ACCESS_R(A, Am, An);
CHAMELEON_ACCESS_R(B, Bm, Bn);
CHAMELEON_ACCESS_RW(C, Cm, Cn);
exec = __chameleon_need_exec;
CHAMELEON_END_ACCESS_DECLARATION;
if ( exec ) {
clargs = malloc( sizeof( struct cl_gemmex_args_s ) );
clargs->transA = transA;
clargs->transB = transB;
clargs->m = m;
clargs->n = n;
clargs->k = k;
clargs->alpha = alpha;
clargs->beta = beta;
}
/* Callback for profiling information */
callback = options->profiling ? cl_gemmex_callback : NULL;
/* Reduce the C access if needed */
accessC = ( beta == 0. ) ? STARPU_W : (STARPU_RW | ((beta == 1.) ? STARPU_COMMUTE : 0));
/* Refine name */
cl_name = chameleon_codelet_name( cl_name, 3,
A->get_blktile( A, Am, An ),
B->get_blktile( B, Bm, Bn ),
C->get_blktile( C, Cm, Cn ) );
/* Insert the task */
rt_starpu_insert_task(
&cl_gemmex,
/* Task codelet arguments */
STARPU_CL_ARGS, clargs, sizeof(struct cl_gemmex_args_s),
/* Task handles */
STARPU_R, RTBLKADDR(A, ChamRealHalf, Am, An),
STARPU_R, RTBLKADDR(B, ChamRealHalf, Bm, Bn),
accessC, RTBLKADDR(C, ChamRealHalf, Cm, Cn),
/* Common task arguments */
STARPU_PRIORITY, options->priority,
STARPU_CALLBACK, callback,
STARPU_EXECUTE_ON_WORKER, options->workerid,
STARPU_POSSIBLY_PARALLEL, options->parallel,
#if defined(CHAMELEON_CODELETS_HAVE_NAME)
STARPU_NAME, cl_name,
#endif
0 );
}
runtime/starpu/include/runtime_codelets.h
+ 1
− 0
View file @ f0dfcdb7
......@@ -152,6 +152,7 @@
CODELETS_HEADER(map);
CODELETS_HEADER(hgemm);
CODELETS_HEADER(gemmex);
struct cl_hgemm_args_s {
cham_trans_t transA;
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment