diff --git a/compute/pzgetrf.c b/compute/pzgetrf.c index 91724fd6a5ca4212598a5219b7e441bf200b21f5..a229b7f3348a178136d558a69f672b56944ef49d 100644 --- a/compute/pzgetrf.c +++ b/compute/pzgetrf.c @@ -398,7 +398,7 @@ void chameleon_pzgetrf( struct chameleon_pzgetrf_s *ws, if ( (ws->alg == ChamGetrfNoPivPerColumn) || (ws->alg == ChamGetrfNoPiv ) ) { - RUNTIME_ipiv_init( IPIV ); + RUNTIME_ipiv_init( sequence, IPIV ); } RUNTIME_options_finalize( &options, chamctxt ); diff --git a/control/control.c b/control/control.c index 3c9ba6a9259f83f1857449e6f6c429a2ecaa396e..a43f9d5712915459fd59d9d6134e7749eca80f5c 100644 --- a/control/control.c +++ b/control/control.c @@ -35,7 +35,8 @@ * * @ingroup Control * - * @brief Initialize CHAMELEON. + * @brief Initialize CHAMELEON with number of cpus and gpus (using + * MPI_COMM_WORLD). * ****************************************************************************** * @@ -59,7 +60,8 @@ int __chameleon_init(int cores, int gpus) * * @ingroup Control * - * @brief Initialize CHAMELEON. + * @brief Initialize CHAMELEON with number of cpus and gpus and threads per + * worker (using MPI_COMM_WORLD). * ****************************************************************************** * @@ -78,6 +80,37 @@ int __chameleon_init(int cores, int gpus) * */ int __chameleon_initpar(int ncpus, int ngpus, int nthreads_per_worker) +{ + return __chameleon_initparcomm( ncpus, ngpus, nthreads_per_worker, MPI_COMM_WORLD ); +} + +/** + * + * @ingroup Control + * + * @brief Initialize CHAMELEON with number of cpus and gpus and threads per + * worker and using a given MPI communicator. + * + ****************************************************************************** + * + * @param[in] ncpus + * Number of cores to use. + * + * @param[in] ngpus + * Number of cuda devices to use. + * + * @param[in] nthreads_per_worker + * Number of threads per worker (cpu, cuda device). + * + * @param[in] comm + * The MPI communicator. + * + ****************************************************************************** + * + * @retval CHAMELEON_SUCCESS successful exit + * + */ +int __chameleon_initparcomm(int ncpus, int ngpus, int nthreads_per_worker, MPI_Comm comm) { CHAM_context_t *chamctxt; @@ -124,6 +157,7 @@ int __chameleon_initpar(int ncpus, int ngpus, int nthreads_per_worker) #endif chamctxt->ncudas = ngpus; + chamctxt->comm = comm; return RUNTIME_init( chamctxt, ncpus, ngpus, nthreads_per_worker ); } @@ -145,7 +179,7 @@ int __chameleon_finalize(void) chameleon_error("CHAMELEON_Finalize", "CHAMELEON not initialized"); return CHAMELEON_ERR_NOT_INITIALIZED; } - RUNTIME_flush(); + RUNTIME_flush(chamctxt); # if !defined(CHAMELEON_SIMULATION) RUNTIME_barrier(chamctxt); # endif diff --git a/coreblas/compute/CMakeLists.txt b/coreblas/compute/CMakeLists.txt index bec6c5aaf474aae765f84ac8f60e6d00c950fc62..d6345ffca541627a14c33eff1c42df6349235fa4 100644 --- a/coreblas/compute/CMakeLists.txt +++ b/coreblas/compute/CMakeLists.txt @@ -164,6 +164,9 @@ endif() target_link_libraries(coreblas PRIVATE MORSE::LAPACKE) target_link_libraries(coreblas PRIVATE MORSE::CBLAS) target_link_libraries(coreblas PUBLIC MORSE::M) +if (CHAMELEON_USE_MPI) + target_link_libraries(coreblas PUBLIC MPI::MPI_C) +endif() # export target coreblas install(EXPORT coreblasTargets diff --git a/example/lapack_to_chameleon/CTestLists.cmake b/example/lapack_to_chameleon/CTestLists.cmake index 774e12e2c34fac67f48a66524da80d674542928e..2d3e8613317aa17b3b9898cf367149abd88ae9d0 100644 --- a/example/lapack_to_chameleon/CTestLists.cmake +++ b/example/lapack_to_chameleon/CTestLists.cmake @@ -9,10 +9,19 @@ set(TESTLIST step3 step4 step5 - step6 - step7 ) foreach(test ${TESTLIST}) - add_test(example_ltm_${test} ./${prec}${test}) + add_test(example_ltm_${test} ./${test}) endforeach() + +if (CHAMELEON_USE_MPI AND MPI_C_FOUND) + set(MPITESTLIST + step6 + step7 + ) + set ( MPICMD mpiexec --bind-to none -n 2 ) + foreach(test ${MPITESTLIST}) + add_test(example_ltm_${test} ${MPICMD} ./${test}) + endforeach() +endif() diff --git a/example/lapack_to_chameleon/step6.c b/example/lapack_to_chameleon/step6.c index fcfad2e230720bd198018b342b437d00021a2805..479b89e9c53e63010e2c1b09183542971f0305da 100644 --- a/example/lapack_to_chameleon/step6.c +++ b/example/lapack_to_chameleon/step6.c @@ -84,8 +84,19 @@ int main(int argc, char *argv[]) { NCPU = iparam[IPARAM_THRDNBR]; NGPU = iparam[IPARAM_NCUDAS]; - /* Initialize CHAMELEON with main parameters */ - int rc = CHAMELEON_Init( NCPU, NGPU ); + /* Example of Chameleon init with a given MPI communicator */ +#if defined(CHAMELEON_USE_MPI) + MPI_Init(NULL, NULL); + int rank; + MPI_Comm_rank(MPI_COMM_WORLD, &rank); + MPI_Comm comm; + MPI_Comm_split(MPI_COMM_WORLD, rank, rank, &comm); +#else + MPI_Comm comm = 0; +#endif + + /* Initialize CHAMELEON with main parameters */ + int rc = CHAMELEON_InitParComm( NCPU, NGPU, 1, comm ); if (rc != CHAMELEON_SUCCESS) { goto finalize; } @@ -235,5 +246,10 @@ finalize: /* Finalize CHAMELEON */ CHAMELEON_Finalize(); +#if defined(CHAMELEON_USE_MPI) + MPI_Comm_free(&comm); + MPI_Finalize(); +#endif + return EXIT_SUCCESS; } diff --git a/example/lapack_to_chameleon/step6.h b/example/lapack_to_chameleon/step6.h index b5876ea519e3422b8b840abbcb7b9891ec6283dc..d93bfcf2c1f4e7c736a0bc45ca9e855de4337e2f 100644 --- a/example/lapack_to_chameleon/step6.h +++ b/example/lapack_to_chameleon/step6.h @@ -26,15 +26,15 @@ /* Specific includes for step 6 */ #include <coreblas/lapacke.h> #include <chameleon.h> -// #if defined(CHAMELEON_USE_MPI) -// #include <mpi.h> -// #endif +#if defined(CHAMELEON_USE_MPI) +#include <mpi.h> +#endif /* Integer parameters for step6 */ enum iparam_step6 { IPARAM_THRDNBR, /* Number of cores */ IPARAM_NCUDAS, /* Number of cuda devices */ - IPARAM_NMPI, /* Number of cuda devices */ + IPARAM_NMPI, /* Number of MPI PROCS */ IPARAM_N, /* Number of columns of the matrix */ IPARAM_NB, /* Number of columns in a tile */ IPARAM_IB, /* Inner-blocking size */ diff --git a/include/chameleon.h b/include/chameleon.h index 03bf38843c710608143e67ffabe44b3fe504aede..bb7bee008a6fcf2fd7320a2b49be5d0d1d4b16f7 100644 --- a/include/chameleon.h +++ b/include/chameleon.h @@ -103,6 +103,7 @@ int CHAMELEON_Initialized (void); int CHAMELEON_My_Mpi_Rank (void) __attribute__((deprecated)); int __chameleon_init (int nworkers, int ncudas); int __chameleon_initpar (int nworkers, int ncudas, int nthreads_per_worker); +int __chameleon_initparcomm (int nworkers, int ncudas, int nthreads_per_worker, MPI_Comm comm); int __chameleon_finalize (void); int CHAMELEON_Pause (void); int CHAMELEON_Resume (void); @@ -223,20 +224,27 @@ void CHAMELEON_Ipiv_Print ( const CHAM_ipiv_t *ipiv ); * */ #if defined(CHAMELEON_SCHED_OPENMP) -#define CHAMELEON_Init( _nworkers_, _ncudas_ ) \ - __chameleon_init( (_nworkers_), (_ncudas_) ); \ - _Pragma("omp parallel") \ - _Pragma("omp master") \ + +#define CHAMELEON_Init( _nworkers_, _ncudas_ ) \ + __chameleon_init( (_nworkers_), (_ncudas_) ); \ + _Pragma("omp parallel") \ + _Pragma("omp master") \ { #define CHAMELEON_InitPar( _nworkers_, _ncudas_, _nthreads_per_worker_ ) \ __chameleon_initpar( (_nworkers_), (_ncudas_), (_nthreads_per_worker_) ); \ - _Pragma("omp parallel")\ - _Pragma("omp master")\ + _Pragma("omp parallel") \ + _Pragma("omp master") \ { -#define CHAMELEON_Finalize() \ - } \ +#define CHAMELEON_InitParComm( _nworkers_, _ncudas_, _nthreads_per_worker_, _comm_ ) \ + __chameleon_initparcomm( (_nworkers_), (_ncudas_), (_nthreads_per_worker_), (_comm_) ); \ + _Pragma("omp parallel") \ + _Pragma("omp master") \ + { + +#define CHAMELEON_Finalize() \ + } \ __chameleon_finalize(); #else @@ -245,7 +253,10 @@ void CHAMELEON_Ipiv_Print ( const CHAM_ipiv_t *ipiv ); __chameleon_init( (_nworkers_), (_ncudas_) ); #define CHAMELEON_InitPar( _nworkers_, _ncudas_, _nthreads_per_worker_ ) \ - __chameleon_initpar( (_nworkers_), (_ncudas_), (_nthreads_per_worker_) ); + __chameleon_initpar( (_nworkers_), (_ncudas_), (_nthreads_per_worker_), MPI_COMM_WORLD ); + +#define CHAMELEON_InitParComm( _nworkers_, _ncudas_, _nthreads_per_worker_, _comm_ ) \ + __chameleon_initparcomm( (_nworkers_), (_ncudas_), (_nthreads_per_worker_), (_comm_) ); #define CHAMELEON_Finalize() \ __chameleon_finalize(); diff --git a/include/chameleon/runtime.h b/include/chameleon/runtime.h index 2655ef34669642404a79a566db7f0020323f9c65..eb41a43501d9ba40cb2d2ec5b65490b59245afad 100644 --- a/include/chameleon/runtime.h +++ b/include/chameleon/runtime.h @@ -480,7 +480,7 @@ RUNTIME_desc_flush( const CHAM_desc_t *desc, * This function flushes all data from the distributed cache of the runtime system. */ void -RUNTIME_flush( ); +RUNTIME_flush( CHAM_context_t *chamctxt ); /** * @brief Flush a single piece of data. @@ -707,7 +707,8 @@ void RUNTIME_sdisplay_oneprofile (cham_tasktype_t task); void RUNTIME_ipiv_create ( CHAM_ipiv_t *ipiv ); void RUNTIME_ipiv_destroy( CHAM_ipiv_t *ipiv ); -void RUNTIME_ipiv_init ( CHAM_ipiv_t *ipiv ); +void RUNTIME_ipiv_init ( const RUNTIME_sequence_t *sequence, + CHAM_ipiv_t *ipiv ); void RUNTIME_ipiv_gather ( CHAM_ipiv_t *desc, int *ipiv, int node ); void *RUNTIME_ipiv_getaddr ( const CHAM_ipiv_t *ipiv, int m ); diff --git a/include/chameleon/runtime_struct.h b/include/chameleon/runtime_struct.h index 4d345a0fdbc2be301a6af7d91a4da28be8bd4def..88ff1a61d84642de1ebd09ee4777cfc8aaa8dabd 100644 --- a/include/chameleon/runtime_struct.h +++ b/include/chameleon/runtime_struct.h @@ -23,6 +23,17 @@ #ifndef _chameleon_runtime_struct_h_ #define _chameleon_runtime_struct_h_ +#if defined(CHAMELEON_USE_MPI) +#include <mpi.h> +#else +#ifndef MPI_Comm +typedef uintptr_t MPI_Comm; +#endif +#ifndef MPI_COMM_WORLD +#define MPI_COMM_WORLD 0 +#endif +#endif + BEGIN_C_DECLS /** @@ -70,6 +81,7 @@ typedef struct runtime_sequence_s { int status; /**< Return status registered by the tasks for the request */ RUNTIME_request_t *request; /**< Pointer to the request that failed if any, NULL otherwise */ void *schedopt; /**< Specific runtime data pointer to handle the sequence */ + MPI_Comm comm; /**< MPI communicator */ } RUNTIME_sequence_t; /** diff --git a/include/chameleon/struct.h b/include/chameleon/struct.h index bff3788a53e5a59b9d67dba01c8577c356d80b7c..6f3e47662f31eb96eb81da062ed7d08e6c086350 100644 --- a/include/chameleon/struct.h +++ b/include/chameleon/struct.h @@ -30,6 +30,17 @@ #include "chameleon/constants.h" #include "chameleon/runtime_struct.h" +#if defined(CHAMELEON_USE_MPI) +#include <mpi.h> +#else +#ifndef MPI_Comm +typedef uintptr_t MPI_Comm; +#endif +#ifndef MPI_COMM_WORLD +#define MPI_COMM_WORLD 0 +#endif +#endif + BEGIN_C_DECLS #define CHAMELEON_TILE_FULLRANK (1 << 0) @@ -191,6 +202,7 @@ typedef struct chameleon_context_s { int lookahead; // depth of the look ahead in algorithms void *schedopt; // structure for runtimes int mpi_outer_init; // MPI has been initialized outside our functions + MPI_Comm comm; // MPI communicator } CHAM_context_t; static inline void * diff --git a/runtime/openmp/control/runtime_descriptor.c b/runtime/openmp/control/runtime_descriptor.c index ecec4def24f49034cccf3b0eacf38e2ab0a6018e..383b35a70a2c4df49cbea756ae188a7e7a63783b 100644 --- a/runtime/openmp/control/runtime_descriptor.c +++ b/runtime/openmp/control/runtime_descriptor.c @@ -69,8 +69,9 @@ RUNTIME_desc_flush( const CHAM_desc_t *desc, void -RUNTIME_flush( ) +RUNTIME_flush( CHAM_context_t *chamctxt ) { + (void)chamctxt; return; } diff --git a/runtime/openmp/control/runtime_descriptor_ipiv.c b/runtime/openmp/control/runtime_descriptor_ipiv.c index f10c4156d83f3e50d4b523f3942b0757475b913f..c9fe8dfd05972eedc9e987f0780ee7842ca44616 100644 --- a/runtime/openmp/control/runtime_descriptor_ipiv.c +++ b/runtime/openmp/control/runtime_descriptor_ipiv.c @@ -107,9 +107,11 @@ void RUNTIME_ipiv_reducek( const RUNTIME_option_t *options, (void)h; } -void RUNTIME_ipiv_init( CHAM_ipiv_t *ipiv ) +void RUNTIME_ipiv_init( const RUNTIME_sequence_t *sequence, + CHAM_ipiv_t *ipiv ) { assert( 0 ); + (void)sequence; (void)ipiv; } diff --git a/runtime/parsec/control/runtime_descriptor.c b/runtime/parsec/control/runtime_descriptor.c index 84fecb36063093e3770b4ae55cff2eb9015994dd..fc926df200eec889b4e3d4362ce286b4e0399080 100644 --- a/runtime/parsec/control/runtime_descriptor.c +++ b/runtime/parsec/control/runtime_descriptor.c @@ -345,8 +345,10 @@ int RUNTIME_desc_release( const CHAM_desc_t *desc ) /** * Flush cached data */ -void RUNTIME_flush() +void RUNTIME_flush( CHAM_context_t *chamctxt ) { + (void)chamctxt; + return; } void RUNTIME_desc_flush( const CHAM_desc_t *desc, diff --git a/runtime/parsec/control/runtime_descriptor_ipiv.c b/runtime/parsec/control/runtime_descriptor_ipiv.c index fefb42abf9aaa65f98e2959bf09ca24779c95a7d..b0657f5ba7818ed9109645702010757ea60829b5 100644 --- a/runtime/parsec/control/runtime_descriptor_ipiv.c +++ b/runtime/parsec/control/runtime_descriptor_ipiv.c @@ -107,9 +107,11 @@ void RUNTIME_ipiv_reducek( const RUNTIME_option_t *options, (void)h; } -void RUNTIME_ipiv_init( CHAM_ipiv_t *ipiv ) +void RUNTIME_ipiv_init( const RUNTIME_sequence_t *sequence, + CHAM_ipiv_t *ipiv ) { assert( 0 ); + (void)sequence; (void)ipiv; } diff --git a/runtime/quark/control/runtime_descriptor.c b/runtime/quark/control/runtime_descriptor.c index 1e9be06faf1e072269dfeea60eda2ef463214d3e..740522a08e5e6169b72cfc6659bf6d83e348ea13 100644 --- a/runtime/quark/control/runtime_descriptor.c +++ b/runtime/quark/control/runtime_descriptor.c @@ -70,8 +70,9 @@ RUNTIME_desc_flush( const CHAM_desc_t *desc, void -RUNTIME_flush( ) +RUNTIME_flush( CHAM_context_t *chamctxt ) { + (void)chamctxt; return; } diff --git a/runtime/quark/control/runtime_descriptor_ipiv.c b/runtime/quark/control/runtime_descriptor_ipiv.c index 88e8f886e8578f99e066868e6dfb2880fc4035d0..3053435b1f00d155606dd871eb042fabd061321a 100644 --- a/runtime/quark/control/runtime_descriptor_ipiv.c +++ b/runtime/quark/control/runtime_descriptor_ipiv.c @@ -107,9 +107,11 @@ void RUNTIME_ipiv_reducek( const RUNTIME_option_t *options, (void)h; } -void RUNTIME_ipiv_init( CHAM_ipiv_t *ipiv ) +void RUNTIME_ipiv_init( const RUNTIME_sequence_t *sequence, + CHAM_ipiv_t *ipiv ) { assert( 0 ); + (void)sequence; (void)ipiv; } diff --git a/runtime/starpu/codelets/codelet_convert.c b/runtime/starpu/codelets/codelet_convert.c index 0b07c069a1d509dc2fbbc09bd97f611154aa8ee0..101268134ba61e71b81375f122bcf3c1f6b95063 100644 --- a/runtime/starpu/codelets/codelet_convert.c +++ b/runtime/starpu/codelets/codelet_convert.c @@ -119,6 +119,7 @@ insert_task_convert( const RUNTIME_option_t *options, } rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_dlag2h.c b/runtime/starpu/codelets/codelet_dlag2h.c index b910559db38c20a147e23173c884ea5364d923e1..993026a3d68194ea52477a0e1bdce2587bf857c7 100644 --- a/runtime/starpu/codelets/codelet_dlag2h.c +++ b/runtime/starpu/codelets/codelet_dlag2h.c @@ -81,6 +81,7 @@ void INSERT_TASK_dlag2h( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), @@ -146,6 +147,7 @@ void INSERT_TASK_hlag2d( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_dlag2z.c b/runtime/starpu/codelets/codelet_dlag2z.c index 34e0b16c0e29207f514110cb9333876e40524319..0edc438e45a7da5d992d64e4b18db23c213476c8 100644 --- a/runtime/starpu/codelets/codelet_dlag2z.c +++ b/runtime/starpu/codelets/codelet_dlag2z.c @@ -61,6 +61,7 @@ void INSERT_TASK_dlag2z( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(uplo), STARPU_VALUE, &m, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_dzasum.c b/runtime/starpu/codelets/codelet_dzasum.c index 1fc2cf3090bf046b8ffdda4dd021619ce2d08090..859ce1d3663ba6e94aa1953c9d73542aaa7e98bc 100644 --- a/runtime/starpu/codelets/codelet_dzasum.c +++ b/runtime/starpu/codelets/codelet_dzasum.c @@ -62,6 +62,7 @@ void INSERT_TASK_dzasum( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &storev, sizeof(cham_store_t), STARPU_VALUE, &uplo, sizeof(cham_uplo_t), diff --git a/runtime/starpu/codelets/codelet_gemm.c b/runtime/starpu/codelets/codelet_gemm.c index 69cd67a294d2fcdf116e28326072702527a8d699..bdec166e3046f07aa964941d48bb834cb5559f2d 100644 --- a/runtime/starpu/codelets/codelet_gemm.c +++ b/runtime/starpu/codelets/codelet_gemm.c @@ -184,6 +184,7 @@ INSERT_TASK_gemm( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, codelet, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, argssize, diff --git a/runtime/starpu/codelets/codelet_gemmex.c b/runtime/starpu/codelets/codelet_gemmex.c index efc33dab33d5fef8c8186aebf0c53e401f9ef2ba..75473ab548f082bfc405cd26f9e62aa3ce48c63a 100644 --- a/runtime/starpu/codelets/codelet_gemmex.c +++ b/runtime/starpu/codelets/codelet_gemmex.c @@ -179,6 +179,7 @@ void INSERT_TASK_gemmex_Astat( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_gemmex, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_gemmex_args_s), @@ -257,6 +258,7 @@ void INSERT_TASK_gemmex( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_gemmex, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_gemmex_args_s), diff --git a/runtime/starpu/codelets/codelet_hgemm.c b/runtime/starpu/codelets/codelet_hgemm.c index a595bf990cd336e50aee8c541954ef5591c5ed8b..7a5dc716d57252c2b3d5820aa7dcce05246b0ff3 100644 --- a/runtime/starpu/codelets/codelet_hgemm.c +++ b/runtime/starpu/codelets/codelet_hgemm.c @@ -158,6 +158,7 @@ void INSERT_TASK_hgemm_Astat( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_hgemm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_hgemm_args_s), @@ -229,6 +230,7 @@ void INSERT_TASK_hgemm( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_hgemm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_hgemm_args_s), diff --git a/runtime/starpu/codelets/codelet_ipiv_to_perm.c b/runtime/starpu/codelets/codelet_ipiv_to_perm.c index 31183c11505a0f19fa3505691684c37810c0f10e..f5883767fdc7567d47d8cc10cc5e64f764f45eda 100644 --- a/runtime/starpu/codelets/codelet_ipiv_to_perm.c +++ b/runtime/starpu/codelets/codelet_ipiv_to_perm.c @@ -56,6 +56,7 @@ void INSERT_TASK_ipiv_to_perm( const RUNTIME_option_t *options, struct starpu_codelet *codelet = &cl_ipiv_to_perm; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m0, sizeof(int), STARPU_VALUE, &m, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_map.c b/runtime/starpu/codelets/codelet_map.c index a55c56ef8fc79e84db3a04a9fa332e889443192b..eadd45bfd0db8cfa0fc89444e6cdd0e796bde8fc 100644 --- a/runtime/starpu/codelets/codelet_map.c +++ b/runtime/starpu/codelets/codelet_map.c @@ -59,6 +59,7 @@ void INSERT_TASK_map( const RUNTIME_option_t *options, A->get_blktile( A, Am, An ) ); rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &A, sizeof(CHAM_desc_t*), STARPU_VALUE, &uplo, sizeof(cham_uplo_t), diff --git a/runtime/starpu/codelets/codelet_zaxpy.c b/runtime/starpu/codelets/codelet_zaxpy.c index be1fbbbf422325c84068513392c11bb7352c424c..08de84a228dde3cfceb1f0ae9decc940c36cc8cc 100644 --- a/runtime/starpu/codelets/codelet_zaxpy.c +++ b/runtime/starpu/codelets/codelet_zaxpy.c @@ -62,6 +62,7 @@ void INSERT_TASK_zaxpy( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &M, sizeof(int), STARPU_VALUE, &alpha, sizeof(CHAMELEON_Complex64_t), diff --git a/runtime/starpu/codelets/codelet_zbuild.c b/runtime/starpu/codelets/codelet_zbuild.c index 3ce93dea06a18299c2ac4fa0d7c471b8b975d297..71938c7d40bc73d141ae5bba5181aa8f72c23120 100644 --- a/runtime/starpu/codelets/codelet_zbuild.c +++ b/runtime/starpu/codelets/codelet_zbuild.c @@ -74,6 +74,7 @@ CODELETS_CPU(zbuild, cl_zbuild_cpu_func) col_min = An*A->nb ; col_max = An == A->nt-1 ? A->n-1 : col_min+A->nb-1 ; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &row_min, sizeof(int), STARPU_VALUE, &row_max, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zcesca.c b/runtime/starpu/codelets/codelet_zcesca.c index 74cb9c7c6834723ae1a536b933b450c8b6439344..fb6a24b9913bcb7b1496657ee718a23fe5b091c4 100644 --- a/runtime/starpu/codelets/codelet_zcesca.c +++ b/runtime/starpu/codelets/codelet_zcesca.c @@ -94,6 +94,7 @@ void INSERT_TASK_zcesca( const RUNTIME_option_t *options, } rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_CL_ARGS, clargs, sizeof(struct cl_zcesca_args_s), STARPU_R, RTBLKADDR(Gi, ChamComplexDouble, Gim, Gin), diff --git a/runtime/starpu/codelets/codelet_zgeadd.c b/runtime/starpu/codelets/codelet_zgeadd.c index 60daa3b177a7ec86bcbba840c0e64dd33e72b39b..fda00982faac7cafee0f5c490ac059edcbafaeaa 100644 --- a/runtime/starpu/codelets/codelet_zgeadd.c +++ b/runtime/starpu/codelets/codelet_zgeadd.c @@ -99,6 +99,7 @@ void INSERT_TASK_zgeadd( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &trans, sizeof(int), STARPU_VALUE, &m, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgelqt.c b/runtime/starpu/codelets/codelet_zgelqt.c index 846482361badc28556fb6da43492b38754354ece..36b1cde9368085a94879750456f73b5fe7b31b4a 100644 --- a/runtime/starpu/codelets/codelet_zgelqt.c +++ b/runtime/starpu/codelets/codelet_zgelqt.c @@ -76,6 +76,7 @@ void INSERT_TASK_zgelqt(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgemm.c b/runtime/starpu/codelets/codelet_zgemm.c index 4b47627258ba618751e82a33945f3fbeed27c8bc..fe62de383657a7b2f049f57dfbadf1cdc94b639e 100644 --- a/runtime/starpu/codelets/codelet_zgemm.c +++ b/runtime/starpu/codelets/codelet_zgemm.c @@ -186,6 +186,7 @@ void INSERT_TASK_zgemm_Astat( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zgemm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zgemm_args_s), @@ -257,6 +258,7 @@ void INSERT_TASK_zgemm( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zgemm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zgemm_args_s), diff --git a/runtime/starpu/codelets/codelet_zgemv.c b/runtime/starpu/codelets/codelet_zgemv.c index a2529d99995bf272d4e960916c61c82fea26b762..307c58c7d0dd1a0e946c58d53b83446e67b07709 100644 --- a/runtime/starpu/codelets/codelet_zgemv.c +++ b/runtime/starpu/codelets/codelet_zgemv.c @@ -104,6 +104,7 @@ void INSERT_TASK_zgemv( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &trans, sizeof(cham_trans_t), STARPU_VALUE, &m, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgeqrt.c b/runtime/starpu/codelets/codelet_zgeqrt.c index 0ee282e4a02c5bb1dc47e82c474c88a445144114..dea86bb8fcfeb58ee03f7634fde45c9f1b3bf8d1 100644 --- a/runtime/starpu/codelets/codelet_zgeqrt.c +++ b/runtime/starpu/codelets/codelet_zgeqrt.c @@ -77,6 +77,7 @@ void INSERT_TASK_zgeqrt(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgered.c b/runtime/starpu/codelets/codelet_zgered.c index 394ef74c24ac9023a1b594b348ba21479df48d3a..e4cdf784022c4c40e4b7c86b5f404b53c58c5997 100644 --- a/runtime/starpu/codelets/codelet_zgered.c +++ b/runtime/starpu/codelets/codelet_zgered.c @@ -78,6 +78,7 @@ void INSERT_TASK_zgered( const RUNTIME_option_t *options, starpu_cham_tile_register( &handleAout, -1, tileA, ChamComplexHalf ); rt_starpu_insert_task( + options->sequence->comm, &cl_dlag2h, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), @@ -119,6 +120,7 @@ void INSERT_TASK_zgered( const RUNTIME_option_t *options, starpu_cham_tile_register( &handleAout, -1, tileA, ChamComplexFloat ); rt_starpu_insert_task( + options->sequence->comm, &cl_zlag2c, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgerst.c b/runtime/starpu/codelets/codelet_zgerst.c index 68490f011eef8227bfa656bafdf60ab4c0e70e79..a6e5821ff0ac1e2d3dd54985f71216175775e59e 100644 --- a/runtime/starpu/codelets/codelet_zgerst.c +++ b/runtime/starpu/codelets/codelet_zgerst.c @@ -70,6 +70,7 @@ void INSERT_TASK_zgerst( const RUNTIME_option_t *options, A->myrank, Am, An ); #endif rt_starpu_insert_task( + options->sequence->comm, &cl_hlag2d, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), @@ -92,6 +93,7 @@ void INSERT_TASK_zgerst( const RUNTIME_option_t *options, A->myrank, Am, An ); #endif rt_starpu_insert_task( + options->sequence->comm, &cl_clag2z, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgersum.c b/runtime/starpu/codelets/codelet_zgersum.c index 7c218ab0075ef5c56ceda76360c9797fe765546d..a1053aa03d5bb1d67b1aedf1e89402611cfdedc3 100644 --- a/runtime/starpu/codelets/codelet_zgersum.c +++ b/runtime/starpu/codelets/codelet_zgersum.c @@ -128,7 +128,7 @@ RUNTIME_zgersum_submit_tree( const RUNTIME_option_t *options, const CHAM_desc_t *A, int Am, int An ) { #if defined(HAVE_STARPU_MPI_REDUX) && defined(CHAMELEON_USE_MPI) - starpu_mpi_redux_data_prio_tree( MPI_COMM_WORLD, + starpu_mpi_redux_data_prio_tree( options->sequence->comm, RTBLKADDR(A, ChamComplexDouble, Am, An), options->priority + 1, 2 /* Binary tree */ ); diff --git a/runtime/starpu/codelets/codelet_zgessm.c b/runtime/starpu/codelets/codelet_zgessm.c index 2afae321494854f744c6ad1fcf9161db8701b402..f592f89dc7d5812401bf1a171475ecb27948eeed 100644 --- a/runtime/starpu/codelets/codelet_zgessm.c +++ b/runtime/starpu/codelets/codelet_zgessm.c @@ -71,6 +71,7 @@ void INSERT_TASK_zgessm( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgessq.c b/runtime/starpu/codelets/codelet_zgessq.c index e7de6b666101fbb6d240e85ead4bcb849f9dbe24..43abb20e76a5aedb9b72a33416321a89e00f5b13 100644 --- a/runtime/starpu/codelets/codelet_zgessq.c +++ b/runtime/starpu/codelets/codelet_zgessq.c @@ -61,6 +61,7 @@ void INSERT_TASK_zgessq( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &storev, sizeof(cham_store_t), STARPU_VALUE, &m, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgesum.c b/runtime/starpu/codelets/codelet_zgesum.c index d1947b4a1ecb6666008e63c9b82f558c1bb60e8e..a9755cd7305644affee4c28a07200888e647f1d4 100644 --- a/runtime/starpu/codelets/codelet_zgesum.c +++ b/runtime/starpu/codelets/codelet_zgesum.c @@ -67,6 +67,7 @@ void INSERT_TASK_zgesum( const RUNTIME_option_t *options, } rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_CL_ARGS, clargs, sizeof(struct cl_zgesum_args_s), STARPU_R, RTBLKADDR(A, ChamComplexDouble, Am, An), diff --git a/runtime/starpu/codelets/codelet_zgetrf.c b/runtime/starpu/codelets/codelet_zgetrf.c index 568265ee27721d0c34d80012e860514772a414b4..5abf85cdc98ddf6920ac486b028a1fdbf1983464 100644 --- a/runtime/starpu/codelets/codelet_zgetrf.c +++ b/runtime/starpu/codelets/codelet_zgetrf.c @@ -71,6 +71,7 @@ void INSERT_TASK_zgetrf( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgetrf_blocked.c b/runtime/starpu/codelets/codelet_zgetrf_blocked.c index 6e4662060aa66a18945561352d9ae7fd3e03824e..71767ce14f4339bffa02ace2ac74f9d33980674d 100644 --- a/runtime/starpu/codelets/codelet_zgetrf_blocked.c +++ b/runtime/starpu/codelets/codelet_zgetrf_blocked.c @@ -112,6 +112,7 @@ void INSERT_TASK_zgetrf_blocked_diag( const RUNTIME_option_t *options, } rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &h, sizeof(int), STARPU_VALUE, &m0, sizeof(int), @@ -185,6 +186,7 @@ void INSERT_TASK_zgetrf_blocked_offdiag( const RUNTIME_option_t *options, void (*callback)(void*) = options->profiling ? cl_zgetrf_blocked_offdiag_callback : NULL; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &h, sizeof(int), STARPU_VALUE, &m0, sizeof(int), @@ -253,6 +255,7 @@ void INSERT_TASK_zgetrf_blocked_trsm( const RUNTIME_option_t *options, void (*callback)(void*) = options->profiling ? cl_zgetrf_blocked_trsm_callback : NULL; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgetrf_incpiv.c b/runtime/starpu/codelets/codelet_zgetrf_incpiv.c index 03398f6447c34fb15f85dba005d2a92cb9d1aa9a..99690a73a763be98e28abcca1fd906bd39e6093c 100644 --- a/runtime/starpu/codelets/codelet_zgetrf_incpiv.c +++ b/runtime/starpu/codelets/codelet_zgetrf_incpiv.c @@ -79,6 +79,7 @@ void INSERT_TASK_zgetrf_incpiv(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgetrf_nopiv.c b/runtime/starpu/codelets/codelet_zgetrf_nopiv.c index 4488882e796ff7d9cb87a07f57e9b40c3f49036c..44dcfcac203a0428016faabb30658d1a24ede0a3 100644 --- a/runtime/starpu/codelets/codelet_zgetrf_nopiv.c +++ b/runtime/starpu/codelets/codelet_zgetrf_nopiv.c @@ -71,6 +71,7 @@ void INSERT_TASK_zgetrf_nopiv(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgetrf_nopiv_percol.c b/runtime/starpu/codelets/codelet_zgetrf_nopiv_percol.c index 554735a47e654cadb16c77b209f1e6ed2f28d398..0fc72a70bb13d48e3651f0cfa1f6f42d31e5e63a 100644 --- a/runtime/starpu/codelets/codelet_zgetrf_nopiv_percol.c +++ b/runtime/starpu/codelets/codelet_zgetrf_nopiv_percol.c @@ -90,6 +90,7 @@ void INSERT_TASK_zgetrf_nopiv_percol_diag( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), @@ -161,6 +162,7 @@ void INSERT_TASK_zgetrf_nopiv_percol_trsm( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgetrf_percol.c b/runtime/starpu/codelets/codelet_zgetrf_percol.c index 3248ee0ae4df619ed6988de838ccc9c80c473e33..ef865a7f16f72c9a7d83a7a1a0d06b23fb9bf5c0 100644 --- a/runtime/starpu/codelets/codelet_zgetrf_percol.c +++ b/runtime/starpu/codelets/codelet_zgetrf_percol.c @@ -90,6 +90,7 @@ void INSERT_TASK_zgetrf_percol_diag( const RUNTIME_option_t *options, int access_ppiv = ( h == 0 ) ? STARPU_NONE : STARPU_R; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &h, sizeof(int), STARPU_VALUE, &m0, sizeof(int), @@ -149,6 +150,7 @@ void INSERT_TASK_zgetrf_percol_offdiag( const RUNTIME_option_t *options, void (*callback)(void*) = options->profiling ? cl_zgetrf_percol_offdiag_callback : NULL; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &h, sizeof(int), STARPU_VALUE, &m0, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zgram.c b/runtime/starpu/codelets/codelet_zgram.c index 883093966a5355c00a2fc9f0de9d35bd4b0c823c..7b8067496777d38693cd891024b2cf740a2ec3cc 100644 --- a/runtime/starpu/codelets/codelet_zgram.c +++ b/runtime/starpu/codelets/codelet_zgram.c @@ -64,6 +64,7 @@ void INSERT_TASK_zgram( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &m, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zhe2ge.c b/runtime/starpu/codelets/codelet_zhe2ge.c index 2fa77bfb0464567fdbdf1f4decac0627c39c8076..1ad5865c7e9ec9d505766de14894a050fc9ca9ac 100644 --- a/runtime/starpu/codelets/codelet_zhe2ge.c +++ b/runtime/starpu/codelets/codelet_zhe2ge.c @@ -61,6 +61,7 @@ void INSERT_TASK_zhe2ge( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &m, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zhemm.c b/runtime/starpu/codelets/codelet_zhemm.c index c38e525963aed0aef6789129b8ae2cb8aa8a96d7..4f299e9f4b88fc86d1dba64cbf6c16a9c1c12083 100644 --- a/runtime/starpu/codelets/codelet_zhemm.c +++ b/runtime/starpu/codelets/codelet_zhemm.c @@ -188,6 +188,7 @@ void INSERT_TASK_zhemm_Astat( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zhemm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zhemm_args_s), @@ -258,6 +259,7 @@ void INSERT_TASK_zhemm( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zhemm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zhemm_args_s), diff --git a/runtime/starpu/codelets/codelet_zher2k.c b/runtime/starpu/codelets/codelet_zher2k.c index 223522bbb6559cb1b753579bf48ac608089bdd92..541d1bf36c6fb2b942bb40cb988638ec948712a2 100644 --- a/runtime/starpu/codelets/codelet_zher2k.c +++ b/runtime/starpu/codelets/codelet_zher2k.c @@ -148,6 +148,7 @@ INSERT_TASK_zher2k( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zherfb.c b/runtime/starpu/codelets/codelet_zherfb.c index 1429529684ee9630f21717f1bc1d594c06697469..47a1449ccb836a69800d795ea07468c6dc0b7eb0 100644 --- a/runtime/starpu/codelets/codelet_zherfb.c +++ b/runtime/starpu/codelets/codelet_zherfb.c @@ -99,6 +99,7 @@ void INSERT_TASK_zherfb(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zherk.c b/runtime/starpu/codelets/codelet_zherk.c index 6e866ae6940e3e71af1c59b4b87c59476978c71b..9a03d77af260d5703945233f392bb3c69d74f767 100644 --- a/runtime/starpu/codelets/codelet_zherk.c +++ b/runtime/starpu/codelets/codelet_zherk.c @@ -154,6 +154,7 @@ void INSERT_TASK_zherk( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zherk, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zherk_args_s), diff --git a/runtime/starpu/codelets/codelet_zlacpy.c b/runtime/starpu/codelets/codelet_zlacpy.c index 4d4f1018582d90182eb246f0a5fac8615222b0ac..2fedd9d5230ca716c7f5af56bfc51099c5108751 100644 --- a/runtime/starpu/codelets/codelet_zlacpy.c +++ b/runtime/starpu/codelets/codelet_zlacpy.c @@ -120,9 +120,9 @@ insert_task_zlacpy_on_remote_node( const RUNTIME_option_t *options, { void (*callback)(void*) = options->profiling ? cl_zlacpy_callback : NULL; #if defined(CHAMELEON_RUNTIME_SYNC) - starpu_mpi_data_cpy_priority( handleB, handleA, MPI_COMM_WORLD, 0, callback, NULL, options->priority ); + starpu_mpi_data_cpy_priority( handleB, handleA, options->sequence->comm, 0, callback, NULL, options->priority ); #else - starpu_mpi_data_cpy_priority( handleB, handleA, MPI_COMM_WORLD, 1, callback, NULL, options->priority ); + starpu_mpi_data_cpy_priority( handleB, handleA, options->sequence->comm, 1, callback, NULL, options->priority ); #endif } #endif @@ -183,6 +183,7 @@ void INSERT_TASK_zlacpyx( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zlacpyx, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zlacpy_args_s), @@ -258,6 +259,7 @@ void INSERT_TASK_zlacpy( const RUNTIME_option_t *options, callback = options->profiling ? cl_zlacpy_callback : NULL; rt_starpu_insert_task( + options->sequence->comm, &cl_zlacpy, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zlacpy_args_s), diff --git a/runtime/starpu/codelets/codelet_zlag2c.c b/runtime/starpu/codelets/codelet_zlag2c.c index 961cc9c0376c0fa48ce6bbf0fe2b4e183e3714ff..e0b2151f918782ab74987b2e2257082cb677b871 100644 --- a/runtime/starpu/codelets/codelet_zlag2c.c +++ b/runtime/starpu/codelets/codelet_zlag2c.c @@ -97,6 +97,7 @@ void INSERT_TASK_zlag2c(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), @@ -175,6 +176,7 @@ void INSERT_TASK_clag2z( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zlange.c b/runtime/starpu/codelets/codelet_zlange.c index 1747215a60e72b1cdac0c7e77bfff04a31725ce7..3e944050bf8faf98b23f152ccffd3187e393be72 100644 --- a/runtime/starpu/codelets/codelet_zlange.c +++ b/runtime/starpu/codelets/codelet_zlange.c @@ -66,6 +66,7 @@ void INSERT_TASK_zlange( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &norm, sizeof(cham_normtype_t), STARPU_VALUE, &M, sizeof(int), @@ -120,6 +121,7 @@ void INSERT_TASK_zlange_max(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_R, RTBLKADDR(A, ChamRealDouble, Am, An), STARPU_RW, RTBLKADDR(B, ChamRealDouble, Bm, Bn), diff --git a/runtime/starpu/codelets/codelet_zlanhe.c b/runtime/starpu/codelets/codelet_zlanhe.c index c45ecc83987918785f212fd8922453e679ecb63f..6c80e092490939e5438c45d707987ea4bc55dc10 100644 --- a/runtime/starpu/codelets/codelet_zlanhe.c +++ b/runtime/starpu/codelets/codelet_zlanhe.c @@ -65,6 +65,7 @@ void INSERT_TASK_zlanhe(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &norm, sizeof(int), STARPU_VALUE, &uplo, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zlansy.c b/runtime/starpu/codelets/codelet_zlansy.c index 21b5a1140c3a9d94d032308cf81a55b751844dbd..1c88303e7e3c5d89e5551d1a95cd4c69018f679e 100644 --- a/runtime/starpu/codelets/codelet_zlansy.c +++ b/runtime/starpu/codelets/codelet_zlansy.c @@ -66,6 +66,7 @@ void INSERT_TASK_zlansy( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &norm, sizeof(int), STARPU_VALUE, &uplo, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zlantr.c b/runtime/starpu/codelets/codelet_zlantr.c index e2067a02c392391c39dc28e9bf81ebf922defc4e..86a5e6ce2e94ff716b0b4921910add202d8ab9a2 100644 --- a/runtime/starpu/codelets/codelet_zlantr.c +++ b/runtime/starpu/codelets/codelet_zlantr.c @@ -65,6 +65,7 @@ void INSERT_TASK_zlantr( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &norm, sizeof(int), STARPU_VALUE, &uplo, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zlascal.c b/runtime/starpu/codelets/codelet_zlascal.c index 1246af7734068fdb4b836e03ba87adee2751ba69..c0acf8c78fcdfb241e9ff9365ba49854519020fd 100644 --- a/runtime/starpu/codelets/codelet_zlascal.c +++ b/runtime/starpu/codelets/codelet_zlascal.c @@ -88,6 +88,7 @@ void INSERT_TASK_zlascal( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zlascal, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zlascal_args_s), diff --git a/runtime/starpu/codelets/codelet_zlaset.c b/runtime/starpu/codelets/codelet_zlaset.c index b78055d3385b4609cb43c1a6056e99cd58815e07..91bb2d459669dc32175b8f272d8ed9c7a0e9a0f5 100644 --- a/runtime/starpu/codelets/codelet_zlaset.c +++ b/runtime/starpu/codelets/codelet_zlaset.c @@ -82,6 +82,7 @@ void INSERT_TASK_zlaset( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zlaset, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zlaset_args_s), diff --git a/runtime/starpu/codelets/codelet_zlaset2.c b/runtime/starpu/codelets/codelet_zlaset2.c index 33a136f359bb8979d09f6222bddc866e2ef2d437..f7d3cc3ff5fd4991d79eb4bd1db28dd1efab69d9 100644 --- a/runtime/starpu/codelets/codelet_zlaset2.c +++ b/runtime/starpu/codelets/codelet_zlaset2.c @@ -61,6 +61,7 @@ void INSERT_TASK_zlaset2(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &M, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zlaswp.c b/runtime/starpu/codelets/codelet_zlaswp.c index 2d8fc31d422fa39da13db5f2b2240cd7096d64e3..8c7d716be2dce7d929ab7334c846f1e900c9a684 100644 --- a/runtime/starpu/codelets/codelet_zlaswp.c +++ b/runtime/starpu/codelets/codelet_zlaswp.c @@ -51,6 +51,7 @@ void INSERT_TASK_zlaswp_get( const RUNTIME_option_t *options, //void (*callback)(void*) = options->profiling ? cl_zlaswp_get_callback : NULL; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m0, sizeof(int), STARPU_VALUE, &k, sizeof(int), @@ -95,6 +96,7 @@ void INSERT_TASK_zlaswp_set( const RUNTIME_option_t *options, //void (*callback)(void*) = options->profiling ? cl_zlaswp_set_callback : NULL; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m0, sizeof(int), STARPU_VALUE, &k, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zlatro.c b/runtime/starpu/codelets/codelet_zlatro.c index 42fa0637df433f277656c09c259b55d33df347e3..0e9471ec34c2ab960ada674420a3d003d059137c 100644 --- a/runtime/starpu/codelets/codelet_zlatro.c +++ b/runtime/starpu/codelets/codelet_zlatro.c @@ -72,6 +72,7 @@ void INSERT_TASK_zlatro( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zlauum.c b/runtime/starpu/codelets/codelet_zlauum.c index bdb000e520bebbc92b8aaa45adbf883bb6d5ed52..dbdb47e92ec297fb745e9f4c23113be247e46536 100644 --- a/runtime/starpu/codelets/codelet_zlauum.c +++ b/runtime/starpu/codelets/codelet_zlauum.c @@ -80,6 +80,7 @@ void INSERT_TASK_zlauum( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zlauum, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zlauum_args_s), diff --git a/runtime/starpu/codelets/codelet_zplghe.c b/runtime/starpu/codelets/codelet_zplghe.c index bc3052e2730e8d394bb2bc4ac7a93f44a0642e1a..3c90a52b29bd2e55003be33522c57131607c25fa 100644 --- a/runtime/starpu/codelets/codelet_zplghe.c +++ b/runtime/starpu/codelets/codelet_zplghe.c @@ -86,6 +86,7 @@ void INSERT_TASK_zplghe( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zplghe, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zplghe_args_s), diff --git a/runtime/starpu/codelets/codelet_zplgsy.c b/runtime/starpu/codelets/codelet_zplgsy.c index fd29f8d0a0263bf97c7e76aa97b746e5febf6099..18dd13e7d5773a07f96e3083184bf99f10c2bc35 100644 --- a/runtime/starpu/codelets/codelet_zplgsy.c +++ b/runtime/starpu/codelets/codelet_zplgsy.c @@ -86,6 +86,7 @@ void INSERT_TASK_zplgsy( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zplgsy, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zplgsy_args_s), diff --git a/runtime/starpu/codelets/codelet_zplrnt.c b/runtime/starpu/codelets/codelet_zplrnt.c index abb1ebd399c75d3b8db3cc77dd3239e1c81f5047..b242aecca5b6cb58efb971e748aca02979ad07c9 100644 --- a/runtime/starpu/codelets/codelet_zplrnt.c +++ b/runtime/starpu/codelets/codelet_zplrnt.c @@ -85,6 +85,7 @@ void INSERT_TASK_zplrnt( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zplrnt, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zplrnt_args_s), diff --git a/runtime/starpu/codelets/codelet_zplssq.c b/runtime/starpu/codelets/codelet_zplssq.c index 6a1c7a54d0410bb1a1e5aa64c03c25235c60ebdd..cbd42e5ce45b1511e3f70b128c795518b7f11b9d 100644 --- a/runtime/starpu/codelets/codelet_zplssq.c +++ b/runtime/starpu/codelets/codelet_zplssq.c @@ -65,6 +65,7 @@ void INSERT_TASK_zplssq( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &storev, sizeof(int), STARPU_VALUE, &M, sizeof(int), @@ -113,6 +114,7 @@ void INSERT_TASK_zplssq2( const RUNTIME_option_t *options, int N, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &N, sizeof(int), STARPU_RW, RTBLKADDR(RESULT, ChamRealDouble, RESULTm, RESULTn), diff --git a/runtime/starpu/codelets/codelet_zpotrf.c b/runtime/starpu/codelets/codelet_zpotrf.c index 1195a1e93e5d13db02e75c3bcd962a95394cd688..101d1acb4bdbca68acf77b3540c874f4789ffd4e 100644 --- a/runtime/starpu/codelets/codelet_zpotrf.c +++ b/runtime/starpu/codelets/codelet_zpotrf.c @@ -97,6 +97,7 @@ void INSERT_TASK_zpotrf( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zpotrf, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zpotrf_args_s), diff --git a/runtime/starpu/codelets/codelet_zssssm.c b/runtime/starpu/codelets/codelet_zssssm.c index f9cece0bf2156ec58922f77fdd10d077c7150875..9c07f1268e43eb04823ed8e873267eccae0344d1 100644 --- a/runtime/starpu/codelets/codelet_zssssm.c +++ b/runtime/starpu/codelets/codelet_zssssm.c @@ -78,6 +78,7 @@ void INSERT_TASK_zssssm( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m1, sizeof(int), STARPU_VALUE, &n1, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zsymm.c b/runtime/starpu/codelets/codelet_zsymm.c index 6748325a666032c7d7cb210da4f49cb1b3d247ed..1046cbcb3c142057025994111398ea7273327657 100644 --- a/runtime/starpu/codelets/codelet_zsymm.c +++ b/runtime/starpu/codelets/codelet_zsymm.c @@ -189,6 +189,7 @@ void INSERT_TASK_zsymm_Astat( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zsymm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zsymm_args_s), @@ -259,6 +260,7 @@ void INSERT_TASK_zsymm( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zsymm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zsymm_args_s), diff --git a/runtime/starpu/codelets/codelet_zsyr2k.c b/runtime/starpu/codelets/codelet_zsyr2k.c index 51f72d20ea50b88bc00120ce412338ff0d00bad0..074371252ba489c667342974f7eb13b70b3ea18d 100644 --- a/runtime/starpu/codelets/codelet_zsyr2k.c +++ b/runtime/starpu/codelets/codelet_zsyr2k.c @@ -147,6 +147,7 @@ void INSERT_TASK_zsyr2k(const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zsyrk.c b/runtime/starpu/codelets/codelet_zsyrk.c index 78874feba5e687dda17b9c9170c82a66fc3e5bf5..68eb97b3c3c2c231cb4c99baa0fdb10fea0e975e 100644 --- a/runtime/starpu/codelets/codelet_zsyrk.c +++ b/runtime/starpu/codelets/codelet_zsyrk.c @@ -155,6 +155,7 @@ void INSERT_TASK_zsyrk( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_zsyrk, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_zsyrk_args_s), diff --git a/runtime/starpu/codelets/codelet_zsyssq.c b/runtime/starpu/codelets/codelet_zsyssq.c index cf710e468f8e3f6135357b38549a177903981d97..63d22d8d489c860819b78d1f3ffb6a72f1554a3a 100644 --- a/runtime/starpu/codelets/codelet_zsyssq.c +++ b/runtime/starpu/codelets/codelet_zsyssq.c @@ -59,6 +59,7 @@ void INSERT_TASK_zsyssq( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &storev, sizeof(cham_store_t), STARPU_VALUE, &uplo, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zsytrf_nopiv.c b/runtime/starpu/codelets/codelet_zsytrf_nopiv.c index 5882c5855b303b90c56636aa0a3f9a64afcc11e5..8e840bf83180bac83c15621c814e1094178d9532 100644 --- a/runtime/starpu/codelets/codelet_zsytrf_nopiv.c +++ b/runtime/starpu/codelets/codelet_zsytrf_nopiv.c @@ -61,6 +61,7 @@ void INSERT_TASK_zsytrf_nopiv( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztplqt.c b/runtime/starpu/codelets/codelet_ztplqt.c index 3471d4878436fd2d49f4ca1835195eb73f03de37..e9c8f78ca79c58c3eadbeab60cf6f638ba73802a 100644 --- a/runtime/starpu/codelets/codelet_ztplqt.c +++ b/runtime/starpu/codelets/codelet_ztplqt.c @@ -67,6 +67,7 @@ void INSERT_TASK_ztplqt( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &M, sizeof(int), STARPU_VALUE, &N, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztpmlqt.c b/runtime/starpu/codelets/codelet_ztpmlqt.c index fb2db26e4d978a4018dcf0d65fe3f47689c10004..57733a115dc3747ac3a7ebd8b5ca60823b40d0d3 100644 --- a/runtime/starpu/codelets/codelet_ztpmlqt.c +++ b/runtime/starpu/codelets/codelet_ztpmlqt.c @@ -109,6 +109,7 @@ void INSERT_TASK_ztpmlqt( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &side, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztpmqrt.c b/runtime/starpu/codelets/codelet_ztpmqrt.c index bab1d8d0a7dfc41c74edd98b1143cc62ebd8330b..6930734f8118ba0bad751885d33fe65c69f2e731 100644 --- a/runtime/starpu/codelets/codelet_ztpmqrt.c +++ b/runtime/starpu/codelets/codelet_ztpmqrt.c @@ -109,6 +109,7 @@ void INSERT_TASK_ztpmqrt( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &side, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztpqrt.c b/runtime/starpu/codelets/codelet_ztpqrt.c index 0ade7095a124201ccddb2b39ef804d355ecb0ee7..c65c93bcfee039b6d194486164a05470e92786ab 100644 --- a/runtime/starpu/codelets/codelet_ztpqrt.c +++ b/runtime/starpu/codelets/codelet_ztpqrt.c @@ -68,6 +68,7 @@ void INSERT_TASK_ztpqrt( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &M, sizeof(int), STARPU_VALUE, &N, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztradd.c b/runtime/starpu/codelets/codelet_ztradd.c index 57b146d6a70ad2a93a455762cae0e5c01d16856a..788873cfdb17e31913f5136c9f5a7c3438e3b05c 100644 --- a/runtime/starpu/codelets/codelet_ztradd.c +++ b/runtime/starpu/codelets/codelet_ztradd.c @@ -95,6 +95,7 @@ void INSERT_TASK_ztradd( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_ztradd, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_ztradd_args_s), diff --git a/runtime/starpu/codelets/codelet_ztrasm.c b/runtime/starpu/codelets/codelet_ztrasm.c index c5412d41679a63c779aee6aa63a6a02c5ac2caf7..271a6f6795057e8f6cde67edcdbcb16cb2919e7b 100644 --- a/runtime/starpu/codelets/codelet_ztrasm.c +++ b/runtime/starpu/codelets/codelet_ztrasm.c @@ -62,6 +62,7 @@ void INSERT_TASK_ztrasm( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &storev, sizeof(int), STARPU_VALUE, &uplo, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztrmm.c b/runtime/starpu/codelets/codelet_ztrmm.c index a5f2a18773fbb964db65b9e28573c7bd85c945e8..42160f4c6486968f899eeff47229bbc14f981781 100644 --- a/runtime/starpu/codelets/codelet_ztrmm.c +++ b/runtime/starpu/codelets/codelet_ztrmm.c @@ -145,6 +145,7 @@ void INSERT_TASK_ztrmm( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_ztrmm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_ztrmm_args_s), diff --git a/runtime/starpu/codelets/codelet_ztrsm.c b/runtime/starpu/codelets/codelet_ztrsm.c index 00d38429e460dd32999e85d3e9b882e5d57acdd8..6d2baba8b6dc277065e3ac8e083d6951872e5ca6 100644 --- a/runtime/starpu/codelets/codelet_ztrsm.c +++ b/runtime/starpu/codelets/codelet_ztrsm.c @@ -149,6 +149,7 @@ void INSERT_TASK_ztrsm( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_ztrsm, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_ztrsm_args_s), diff --git a/runtime/starpu/codelets/codelet_ztrssq.c b/runtime/starpu/codelets/codelet_ztrssq.c index 9d800319438bd57f759305c4f2553ccbbbbffcef..a0ecb6757817efccae152aa03479e27da0cc1e87 100644 --- a/runtime/starpu/codelets/codelet_ztrssq.c +++ b/runtime/starpu/codelets/codelet_ztrssq.c @@ -61,6 +61,7 @@ void INSERT_TASK_ztrssq( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &uplo, sizeof(int), STARPU_VALUE, &diag, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztrtri.c b/runtime/starpu/codelets/codelet_ztrtri.c index 4918849af0fe53409ef0530160767dc3794bdc03..c6b1919101677ab5b7bd349c302fb9633097efa2 100644 --- a/runtime/starpu/codelets/codelet_ztrtri.c +++ b/runtime/starpu/codelets/codelet_ztrtri.c @@ -93,6 +93,7 @@ void INSERT_TASK_ztrtri( const RUNTIME_option_t *options, /* Insert the task */ rt_starpu_insert_task( + options->sequence->comm, &cl_ztrtri, /* Task codelet arguments */ STARPU_CL_ARGS, clargs, sizeof(struct cl_ztrtri_args_s), diff --git a/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c b/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c index 05510b05943b0858de0910e80cbe6dd8df96102e..883ffc0133b9a22de2e01b6e0ecbe1324de2e2f1 100644 --- a/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c +++ b/runtime/starpu/codelets/codelet_ztsmlq_hetra1.c @@ -86,6 +86,7 @@ void INSERT_TASK_ztsmlq_hetra1( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &side, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztsmqr_hetra1.c b/runtime/starpu/codelets/codelet_ztsmqr_hetra1.c index be370a8878c703fe08491a19d8c3cff62b79f315..8ee4217414870e173f431e263b72f59113857f25 100644 --- a/runtime/starpu/codelets/codelet_ztsmqr_hetra1.c +++ b/runtime/starpu/codelets/codelet_ztsmqr_hetra1.c @@ -86,6 +86,7 @@ void INSERT_TASK_ztsmqr_hetra1( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &side, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_ztstrf.c b/runtime/starpu/codelets/codelet_ztstrf.c index 4839c0d5abd627dea9c64df0a6d0ec972edd0ac1..b7c5f6895bfaa1effddd0f06f122c5836a1864ce 100644 --- a/runtime/starpu/codelets/codelet_ztstrf.c +++ b/runtime/starpu/codelets/codelet_ztstrf.c @@ -91,6 +91,7 @@ void INSERT_TASK_ztstrf( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &m, sizeof(int), STARPU_VALUE, &n, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zunmlq.c b/runtime/starpu/codelets/codelet_zunmlq.c index db6b3726ee0375fa0055c64d7f61067989c438f9..bc39c8dd2cb2c352b2739a67ea0bb7cd27b96b6f 100644 --- a/runtime/starpu/codelets/codelet_zunmlq.c +++ b/runtime/starpu/codelets/codelet_zunmlq.c @@ -111,6 +111,7 @@ void INSERT_TASK_zunmlq( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &side, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/codelets/codelet_zunmqr.c b/runtime/starpu/codelets/codelet_zunmqr.c index 657c6ca87230410ac7ed7fe6e72435ed31fec886..0b7aa9233465ebaa8bc77625b2a3d3a469434118 100644 --- a/runtime/starpu/codelets/codelet_zunmqr.c +++ b/runtime/starpu/codelets/codelet_zunmqr.c @@ -110,6 +110,7 @@ void INSERT_TASK_zunmqr( const RUNTIME_option_t *options, CHAMELEON_END_ACCESS_DECLARATION; rt_starpu_insert_task( + options->sequence->comm, codelet, STARPU_VALUE, &side, sizeof(int), STARPU_VALUE, &trans, sizeof(int), diff --git a/runtime/starpu/control/runtime_async.c b/runtime/starpu/control/runtime_async.c index 5e4f87e6b0b9a64e4fab1700bd23997b0cb44078..2f06fb4c04099c4c1030b59250b208d8c5dbb667 100644 --- a/runtime/starpu/control/runtime_async.c +++ b/runtime/starpu/control/runtime_async.c @@ -28,7 +28,7 @@ int RUNTIME_sequence_create( CHAM_context_t *chamctxt, RUNTIME_sequence_t *sequence ) { (void)chamctxt; - (void)sequence; + sequence->comm = chamctxt->comm; return CHAMELEON_SUCCESS; } @@ -58,10 +58,10 @@ int RUNTIME_sequence_wait( CHAM_context_t *chamctxt, #if defined(CHAMELEON_USE_MPI) # if defined(HAVE_STARPU_MPI_WAIT_FOR_ALL) - starpu_mpi_wait_for_all(MPI_COMM_WORLD); + starpu_mpi_wait_for_all(sequence->comm); # else starpu_task_wait_for_all(); - starpu_mpi_barrier(MPI_COMM_WORLD); + starpu_mpi_barrier(sequence->comm); # endif #else starpu_task_wait_for_all(); diff --git a/runtime/starpu/control/runtime_control.c b/runtime/starpu/control/runtime_control.c index 62c89a00f70c160a8c0ceef38ad78a086189923e..07b67944f78b64a3a1bad47302f4dffd20c02e94 100644 --- a/runtime/starpu/control/runtime_control.c +++ b/runtime/starpu/control/runtime_control.c @@ -100,7 +100,7 @@ void chameleon_starpu_parallel_worker_fini( starpu_sched_opt_t *sched_opt ) /** * */ -static int chameleon_starpu_init( struct starpu_conf *conf ) +static int chameleon_starpu_init( MPI_Comm comm, struct starpu_conf *conf ) { int hres = CHAMELEON_SUCCESS; int rc; @@ -118,7 +118,7 @@ static int chameleon_starpu_init( struct starpu_conf *conf ) # endif # if defined(HAVE_STARPU_MPI_INIT_CONF) - rc = starpu_mpi_init_conf(NULL, NULL, !flag, MPI_COMM_WORLD, conf); + rc = starpu_mpi_init_conf(NULL, NULL, !flag, comm, conf); # else rc = starpu_init(conf); if (rc < 0) { @@ -186,7 +186,7 @@ int RUNTIME_init( CHAM_context_t *chamctxt, if ((ncpus == -1)||(nthreads_per_worker == -1)) { - hres = chameleon_starpu_init( conf ); + hres = chameleon_starpu_init( chamctxt->comm, conf ); chamctxt->nworkers = ncpus; chamctxt->nthreads_per_worker = nthreads_per_worker; @@ -202,7 +202,7 @@ int RUNTIME_init( CHAM_context_t *chamctxt, conf->use_explicit_workers_bindid = 1; - hres = chameleon_starpu_init( conf ); + hres = chameleon_starpu_init( chamctxt->comm, conf ); chamctxt->nworkers = ncpus; chamctxt->nthreads_per_worker = nthreads_per_worker; @@ -300,11 +300,11 @@ void RUNTIME_barrier( CHAM_context_t *chamctxt ) #if defined(CHAMELEON_USE_MPI) # if defined(HAVE_STARPU_MPI_WAIT_FOR_ALL) - starpu_mpi_wait_for_all(MPI_COMM_WORLD); - starpu_mpi_barrier(MPI_COMM_WORLD); + starpu_mpi_wait_for_all(chamctxt->comm); + starpu_mpi_barrier(chamctxt->comm); # else starpu_task_wait_for_all(); - starpu_mpi_barrier(MPI_COMM_WORLD); + starpu_mpi_barrier(chamctxt->comm); # endif #else starpu_task_wait_for_all(); @@ -380,13 +380,12 @@ int RUNTIME_comm_rank( CHAM_context_t *chamctxt ) #if defined(CHAMELEON_USE_MPI) # if defined(HAVE_STARPU_MPI_COMM_RANK) - starpu_mpi_comm_rank( MPI_COMM_WORLD, &rank ); + starpu_mpi_comm_rank( chamctxt->comm, &rank ); # else - MPI_Comm_rank( MPI_COMM_WORLD, &rank ); + MPI_Comm_rank( chamctxt->comm, &rank ); # endif #endif - (void)chamctxt; return rank; } @@ -398,15 +397,14 @@ int RUNTIME_comm_size( CHAM_context_t *chamctxt ) int size; #if defined(CHAMELEON_USE_MPI) # if defined(HAVE_STARPU_MPI_COMM_RANK) - starpu_mpi_comm_size( MPI_COMM_WORLD, &size ); + starpu_mpi_comm_size( chamctxt->comm, &size ); # else - MPI_Comm_size( MPI_COMM_WORLD, &size ); + MPI_Comm_size( chamctxt->comm, &size ); # endif #else size = 1; #endif - (void)chamctxt; return size; } diff --git a/runtime/starpu/control/runtime_descriptor.c b/runtime/starpu/control/runtime_descriptor.c index 2ed4ba05a14db73dc624e812b14ad36e6769d70d..64ba0bb05b388ba84b38b5f8480bde82de096bef 100644 --- a/runtime/starpu/control/runtime_descriptor.c +++ b/runtime/starpu/control/runtime_descriptor.c @@ -267,10 +267,10 @@ int RUNTIME_desc_release( const CHAM_desc_t *desc ) /** * Flush cached data */ -void RUNTIME_flush() +void RUNTIME_flush( CHAM_context_t *chamctxt ) { #if defined(CHAMELEON_USE_MPI) - starpu_mpi_cache_flush_all_data(MPI_COMM_WORLD); + starpu_mpi_cache_flush_all_data(chamctxt->comm); #endif } @@ -317,7 +317,7 @@ void RUNTIME_data_flush( const RUNTIME_sequence_t *sequence, } #if defined(CHAMELEON_USE_MPI) - starpu_mpi_cache_flush( MPI_COMM_WORLD, *handlebis ); + starpu_mpi_cache_flush( sequence->comm, *handlebis ); #endif if ( local ) { @@ -345,7 +345,7 @@ void RUNTIME_data_migrate( const RUNTIME_sequence_t *sequence, old_rank = starpu_mpi_data_get_rank( lhandle ); if ( old_rank != new_rank ) { - starpu_mpi_data_migrate( MPI_COMM_WORLD, lhandle, new_rank ); + starpu_mpi_data_migrate( sequence->comm, lhandle, new_rank ); } (void)sequence; diff --git a/runtime/starpu/control/runtime_descriptor_ipiv.c b/runtime/starpu/control/runtime_descriptor_ipiv.c index f7a6519223d2b56ff4774351d1c5acc29d3b9b51..088b7d44a7b3e239bc2349abcc283d7df91b5433 100644 --- a/runtime/starpu/control/runtime_descriptor_ipiv.c +++ b/runtime/starpu/control/runtime_descriptor_ipiv.c @@ -223,7 +223,7 @@ void RUNTIME_ipiv_flushk( const RUNTIME_sequence_t *sequence, if ( *handle != NULL ) { #if defined(CHAMELEON_USE_MPI) - starpu_mpi_cache_flush( MPI_COMM_WORLD, *handle ); + starpu_mpi_cache_flush( sequence->comm, *handle ); if ( starpu_mpi_data_get_rank( *handle ) == A->myrank ) #endif { @@ -236,7 +236,7 @@ void RUNTIME_ipiv_flushk( const RUNTIME_sequence_t *sequence, if ( *handle != NULL ) { #if defined(CHAMELEON_USE_MPI) - starpu_mpi_cache_flush( MPI_COMM_WORLD, *handle ); + starpu_mpi_cache_flush( sequence->comm, *handle ); if ( starpu_mpi_data_get_rank( *handle ) == A->myrank ) #endif { @@ -272,7 +272,7 @@ void RUNTIME_perm_flushk( const RUNTIME_sequence_t *sequence, if ( *handle != NULL ) { #if defined(CHAMELEON_USE_MPI) - starpu_mpi_cache_flush( MPI_COMM_WORLD, *handle ); + starpu_mpi_cache_flush( sequence->comm, *handle ); if ( starpu_mpi_data_get_rank( *handle ) == A->myrank ) #endif { @@ -285,7 +285,7 @@ void RUNTIME_perm_flushk( const RUNTIME_sequence_t *sequence, if ( *handle != NULL ) { #if defined(CHAMELEON_USE_MPI) - starpu_mpi_cache_flush( MPI_COMM_WORLD, *handle ); + starpu_mpi_cache_flush( sequence->comm, *handle ); if ( starpu_mpi_data_get_rank( *handle ) == A->myrank ) #endif { @@ -307,7 +307,7 @@ void RUNTIME_ipiv_reducek( const RUNTIME_option_t *options, if ( h < ipiv->n ) { #if defined(HAVE_STARPU_MPI_REDUX) && defined(CHAMELEON_USE_MPI) #if !defined(HAVE_STARPU_MPI_REDUX_WRAPUP) - starpu_mpi_redux_data_prio_tree( MPI_COMM_WORLD, nextpiv, + starpu_mpi_redux_data_prio_tree( options->sequence->comm, nextpiv, options->priority, 2 /* Binary tree */ ); #endif #endif @@ -343,7 +343,8 @@ struct starpu_codelet cl_ipiv_init = { .nbuffers = 1, }; -void RUNTIME_ipiv_init( CHAM_ipiv_t *ipiv ) +void RUNTIME_ipiv_init( const RUNTIME_sequence_t *sequence, + CHAM_ipiv_t *ipiv ) { int64_t mt = ipiv->mt; int64_t mb = ipiv->mb; @@ -355,6 +356,7 @@ void RUNTIME_ipiv_init( CHAM_ipiv_t *ipiv ) int n = (m == (mt-1)) ? ipiv->m - m0 : mb; rt_starpu_insert_task( + sequence->comm, &cl_ipiv_init, STARPU_VALUE, &m0, sizeof(int), STARPU_VALUE, &n, sizeof(int), @@ -368,6 +370,7 @@ void RUNTIME_ipiv_gather( CHAM_ipiv_t *desc, int *ipiv, int node ) int64_t mt = desc->mt; int64_t mb = desc->mb; int64_t tag = chameleon_starpu_tag_book( (int64_t)(desc->mt) ); + CHAM_context_t *chamctxt = chameleon_context_self(); int rank = CHAMELEON_Comm_rank(); int owner = rank; int m; @@ -387,7 +390,7 @@ void RUNTIME_ipiv_gather( CHAM_ipiv_t *desc, int *ipiv, int node ) if (already_received == 0) { MPI_Status status; - starpu_mpi_recv( ipiv_src, owner, tag, MPI_COMM_WORLD, &status ); + starpu_mpi_recv( ipiv_src, owner, tag, chamctxt->comm, &status ); } } else if ( rank == owner ) @@ -396,7 +399,7 @@ void RUNTIME_ipiv_gather( CHAM_ipiv_t *desc, int *ipiv, int node ) int already_sent = starpu_mpi_cached_send_set( ipiv_src, node ); if (already_sent == 0) { - starpu_mpi_send( ipiv_src, node, tag, MPI_COMM_WORLD ); + starpu_mpi_send( ipiv_src, node, tag, chamctxt->comm ); } } } diff --git a/runtime/starpu/control/runtime_tags.c b/runtime/starpu/control/runtime_tags.c index 31cc94090a074d7ff6ae05b5d3f187767996ff01..e966c461b99f807d27486234e5c385fb2ceec8a6 100644 --- a/runtime/starpu/control/runtime_tags.c +++ b/runtime/starpu/control/runtime_tags.c @@ -58,7 +58,8 @@ chameleon_starpu_tag_init( void ) int ok = 0; void *tag_ub_p = NULL; - starpu_mpi_comm_get_attr( MPI_COMM_WORLD, STARPU_MPI_TAG_UB, &tag_ub_p, &ok ); + CHAM_context_t *chamctxt = chameleon_context_self(); + starpu_mpi_comm_get_attr( chamctxt->comm, STARPU_MPI_TAG_UB, &tag_ub_p, &ok ); starpu_tag_ub = (uint64_t)((intptr_t)tag_ub_p); if ( !ok ) { diff --git a/runtime/starpu/include/chameleon_starpu.h.in b/runtime/starpu/include/chameleon_starpu.h.in index bd3123ab5e81f53f983c80d2889c3a88010ec7b0..b5840fc1b56a7c43859af1d09860eab4bfb98f66 100644 --- a/runtime/starpu/include/chameleon_starpu.h.in +++ b/runtime/starpu/include/chameleon_starpu.h.in @@ -129,20 +129,20 @@ void *RUNTIME_data_getaddr_withconversion( const RUNTIME_option_t *options, #if defined(CHAMELEON_USE_MPI) #if defined(CHAMELEON_RUNTIME_SYNC) -#define rt_starpu_insert_task( _codelet_, ... ) \ - starpu_mpi_insert_task( MPI_COMM_WORLD, (_codelet_), STARPU_TASK_SYNCHRONOUS, 1, ##__VA_ARGS__ ) +#define rt_starpu_insert_task( _comm_, _codelet_, ... ) \ + starpu_mpi_insert_task( (_comm_), (_codelet_), STARPU_TASK_SYNCHRONOUS, 1, ##__VA_ARGS__ ) #else -#define rt_starpu_insert_task( _codelet_, ... ) \ - starpu_mpi_insert_task( MPI_COMM_WORLD, (_codelet_), ##__VA_ARGS__ ) +#define rt_starpu_insert_task( _comm_, _codelet_, ... ) \ + starpu_mpi_insert_task( (_comm_), (_codelet_), ##__VA_ARGS__ ) #endif #else #if defined(CHAMELEON_RUNTIME_SYNC) -#define rt_starpu_insert_task( _codelet_, ... ) \ +#define rt_starpu_insert_task( _comm_, _codelet_, ... ) \ starpu_insert_task( (_codelet_), STARPU_TASK_SYNCHRONOUS, 1, ##__VA_ARGS__ ) #else -#define rt_starpu_insert_task( _codelet_, ... ) \ +#define rt_starpu_insert_task( _comm_, _codelet_, ... ) \ starpu_insert_task( (_codelet_), ##__VA_ARGS__ ) #endif