Index: ompi/mca/bcol/base/bcol_base_frame.c =================================================================== --- ompi/mca/bcol/base/bcol_base_frame.c (revision 30606) +++ ompi/mca/bcol/base/bcol_base_frame.c (working copy) @@ -28,7 +28,6 @@ #include "ompi/mca/bcol/base/base.h" #include "ompi/include/ompi/constants.h" #include "ompi/mca/mpool/mpool.h" -#include "ompi/mca/coll/ml/coll_ml.h" /*frag and full message descriptors defined here*/ #include "opal/class/opal_list.h" /* * The following file was created by configure. It contains extern @@ -341,3 +340,19 @@ opal_list_item_t, NULL, NULL); + +static void lmngr_block_constructor(mca_bcol_base_lmngr_block_t *item) +{ + item->base_addr = NULL; +} + +static void lnmgr_block_destructor(mca_bcol_base_lmngr_block_t *item) +{ + /* I have nothing to do here */ +} +OBJ_CLASS_INSTANCE(mca_bcol_base_lmngr_block_t, + opal_list_item_t, + lmngr_block_constructor, + lnmgr_block_destructor); + + Index: ompi/mca/bcol/basesmuma/bcol_basesmuma.h =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma.h (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma.h (working copy) @@ -19,8 +19,6 @@ #include "ompi_config.h" #include "ompi/mca/bcol/bcol.h" #include "ompi/mca/mpool/mpool.h" -#include "ompi/mca/coll/ml/coll_ml.h" -#include "ompi/mca/coll/ml/coll_ml_allocation.h" #include "ompi/request/request.h" #include "ompi/proc/proc.h" #include "ompi/patterns/net/netpatterns.h" @@ -406,7 +404,7 @@ /* pool index */ int pool_index; - /* pointer to the ml_memory_block_desc_t structure + /* pointer to the mca_bcol_base_memory_block_desc_t structure * that is actually managing this registration. * This is meaningful when these control structures * are used in conjunction with the user payload @@ -717,7 +715,7 @@ struct mca_bcol_basesmuma_local_mlmem_desc_t { uint32_t bank_index_for_release; - struct ml_memory_block_desc_t *ml_mem_desc; + struct mca_bcol_base_memory_block_desc_t *ml_mem_desc; uint32_t num_banks; uint32_t num_buffers_per_bank; uint32_t size_buffer; @@ -934,7 +932,7 @@ size_t size_ctl_structure; size_t data_seg_alignment; bcol_basesmuma_smcm_mmap_t *sm_mmap; /* shared memory map struct */ - mca_coll_ml_release_buff_fn_t buff_release_cb; /* buffer release + mca_bcol_base_release_buff_fn_t buff_release_cb; /* buffer release call back */ } bcol_basesmuma_registration_data_t; @@ -1014,7 +1012,7 @@ /* shared memory recusive double barrier */ int bcol_basesmuma_recursive_double_barrier(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* shared memory fanin */ int bcol_basesmuma_fanin_init(mca_bcol_base_module_t *super); @@ -1028,36 +1026,36 @@ int bcol_basesmuma_bcast_init(mca_bcol_base_module_t *super); int bcol_basesmuma_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* Shared memory non-blocking broadcast */ int bcol_basesmuma_bcast_k_nomial_anyroot(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_bcast_k_nomial_knownroot(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* Shared memory non-blocking broadcast - Large message anyroot */ int bcol_basesmuma_binary_scatter_allgather_segment(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); #if 0 /*FIXME: having fun here*/ int bcol_basesmuma_hdl_zerocopy_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); #endif int bcol_basesmuma_lmsg_bcast_k_nomial_anyroot(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_lmsg_scatter_allgather_portals_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_lmsg_scatter_allgather_portals_nb_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_lmsg_scatter_allgather_portals_nb_knownroot_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* * shared memory scatter @@ -1067,23 +1065,23 @@ /* shared memory nonblocking scatter - known root */ int bcol_basesmuma_nb_scatter_k_array_knownroot( bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* shared memory non-blocking k-nomial barrier init */ int bcol_basesmuma_k_nomial_barrier_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /* shared memory non-blocking k-nomial barrier progress */ int bcol_basesmuma_k_nomial_barrier_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /*shared memory non-blocking k-nomial allgather init */ int bcol_basesmuma_k_nomial_allgather_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /* shared memory non-blocking k-nomial allgather progress */ int bcol_basesmuma_k_nomial_allgather_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /* shared memory allgather -- selection logic api */ int bcol_basesmuma_allgather_init(mca_bcol_base_module_t *super); @@ -1090,15 +1088,15 @@ /* shared memory blocking k-nomial gather */ int bcol_basesmuma_k_nomial_gather(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* shared memory non blocking k-nomial gather */ int bcol_basesmuma_k_nomial_gather_init(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* shared memory non blocking k-nomial gather progress*/ int bcol_basesmuma_k_nomial_gather_progress(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* shared memory init */ int bcol_basesmuma_gather_init(mca_bcol_base_module_t *super); @@ -1110,9 +1108,9 @@ /* Shared memory basesmuma reduce */ int bcol_basesmuma_reduce_init(mca_bcol_base_module_t *super); int bcol_basesmuma_reduce_intra_fanin(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_reduce_intra_fanin_old(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_reduce_intra_reducescatter_gather(void *sbuf, void *rbuf, int count, struct ompi_datatype_t *dtype, @@ -1125,10 +1123,10 @@ int bcol_basesmuma_allreduce_init(mca_bcol_base_module_t *super); int bcol_basesmuma_allreduce_intra_fanin_fanout(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_allreduce_intra_recursive_doubling(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); /* initialize non-blocking barrier for recycling the memory buffers. * This is not a general purpose nb_barrier, and relies on the @@ -1168,19 +1166,13 @@ int bcol_basesmuma_free_buff( sm_buffer_mgmt * buff_block, uint64_t buff_id ); -/* This function does bcol_basesmuma specific memory registration and - issues call back for ml level bank recycling -*/ -int bcol_basesmuma_bank_init(struct mca_coll_ml_module_t *ml_module, - mca_bcol_base_module_t *bcol_module, - void *reg_data); - /* bank init which is used for shared memory optimization, fall back to * the bank init above if this causes problems */ -int bcol_basesmuma_bank_init_opti(struct mca_coll_ml_module_t *ml_module, - mca_bcol_base_module_t *bcol_module, - void *reg_data); +int bcol_basesmuma_bank_init_opti(struct mca_bcol_base_memory_block_desc_t *payload_block, + uint32_t data_offset, + mca_bcol_base_module_t *bcol_module, + void *reg_data); /* used for shared memory offset exchange */ int base_bcol_basesmuma_exchange_offsets( Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_allgather.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_allgather.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_allgather.c (working copy) @@ -14,7 +14,7 @@ #include "ompi_config.h" #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" +#include "ompi/mca/bcol/base/base.h" #include "ompi/mca/bcol/bcol.h" #include "ompi/mca/bcol/basesmuma/bcol_basesmuma.h" /* @@ -51,7 +51,7 @@ * */ int bcol_basesmuma_k_nomial_allgather_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variables */ int8_t flag_offset; @@ -286,7 +286,7 @@ /* allgather progress function */ int bcol_basesmuma_k_nomial_allgather_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_allreduce.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_allreduce.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_allreduce.c (working copy) @@ -20,9 +20,10 @@ #include "opal/include/opal_stdint.h" +#include "ompi/mca/bcol/base/base.h" #include "bcol_basesmuma.h" -static int bcol_basesmuma_allreduce_intra_fanin_fanout_progress (bcol_function_args_t *input_args, coll_ml_function_t *c_input_args); +static int bcol_basesmuma_allreduce_intra_fanin_fanout_progress (bcol_function_args_t *input_args, mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_allreduce_init(mca_bcol_base_module_t *super) { @@ -185,7 +186,7 @@ } -static int bcol_basesmuma_allreduce_intra_fanin_fanout_progress (bcol_function_args_t *input_args, coll_ml_function_t *c_input_args) +static int bcol_basesmuma_allreduce_intra_fanin_fanout_progress (bcol_function_args_t *input_args, mca_bcol_base_function_t *c_input_args) { mca_bcol_basesmuma_module_t *bcol_module = (mca_bcol_basesmuma_module_t *) c_input_args->bcol_module; int buff_idx = buff_idx = input_args->src_desc->buffer_index; @@ -256,7 +257,7 @@ /** * Shared memory blocking allreduce. */ -int bcol_basesmuma_allreduce_intra_fanin_fanout(bcol_function_args_t *input_args, coll_ml_function_t *c_input_args) +int bcol_basesmuma_allreduce_intra_fanin_fanout(bcol_function_args_t *input_args, mca_bcol_base_function_t *c_input_args) { /* local variables */ mca_bcol_basesmuma_module_t *bcol_module = (mca_bcol_basesmuma_module_t *) c_input_args->bcol_module; @@ -307,7 +308,7 @@ /* this thing uses the old bcol private control structures */ int bcol_basesmuma_allreduce_intra_recursive_doubling(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { int my_rank,group_size,my_node_index; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_bcast.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_bcast.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_bcast.c (working copy) @@ -17,6 +17,9 @@ #include "ompi/datatype/ompi_datatype.h" #include "ompi/communicator/communicator.h" +#include "ompi/mca/bcol/bcol.h" +#include "ompi/mca/bcol/base/base.h" + #include "bcol_basesmuma.h" #define __TEST_BLOCKING__ 1 @@ -117,7 +120,7 @@ * @param module - basesmuma module. */ int bcol_basesmuma_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int group_size, process_shift, my_node_index; @@ -254,7 +257,7 @@ /*zero-copy large massage communication methods*/ #if 0 int bcol_basesmuma_hdl_zerocopy_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int group_size, process_shift, my_node_index; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_bcast_prime.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_bcast_prime.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_bcast_prime.c (working copy) @@ -96,7 +96,7 @@ } while (0) int bcol_basesmuma_bcast_k_nomial_knownroot(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ mca_bcol_basesmuma_module_t* bcol_module= @@ -242,7 +242,7 @@ * @param module - basesmuma module. */ int bcol_basesmuma_bcast_k_nomial_anyroot(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ mca_bcol_basesmuma_module_t* bcol_module= @@ -387,7 +387,7 @@ * algorithms. */ int bcol_basesmuma_binary_scatter_allgather_segment(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_buf_mgmt.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_buf_mgmt.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_buf_mgmt.c (working copy) @@ -21,7 +21,6 @@ #include "ompi/constants.h" #include "ompi/mca/bcol/bcol.h" #include "ompi/mca/bcol/base/base.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "ompi/patterns/comm/coll_ops.h" #include "opal/dss/dss.h" @@ -166,127 +165,6 @@ return ret; } -#if 0 -/* Basesmuma interface function used for buffer bank resource recycling and - bcol specific registration information -*/ -int bcol_basesmuma_bank_init(struct mca_coll_ml_module_t *ml_module, - mca_bcol_base_module_t *bcol_module, - void *reg_data) -{ - /* assumption here is that the block has been registered with - * sm bcol hence has been mapped by each process, need to be - * sure that memory is mapped amongst sm peers - */ - - /* local variables */ - int ret = OMPI_SUCCESS, i; - uint32_t j; - sm_buffer_mgmt *pload_mgmt; - mca_bcol_basesmuma_component_t *cs = &mca_bcol_basesmuma_component; - bcol_basesmuma_registration_data_t *sm_reg_data = - (bcol_basesmuma_registration_data_t *) reg_data; - mca_bcol_basesmuma_module_t *sm_bcol = - (mca_bcol_basesmuma_module_t *) bcol_module; - ml_memory_block_desc_t *ml_block = - ml_module->payload_block; - size_t malloc_size; - ompi_common_sm_file_t input_file; - uint64_t mem_offset; - int leading_dim,loop_limit,buf_id; - unsigned char *base_ptr; - mca_bcol_basesmuma_module_t *sm_bcol_module= - (mca_bcol_basesmuma_module_t *)bcol_module; - - fprintf(stderr,"test opti test\n"); - - /* first, we get a pointer to the payload buffer management struct */ - pload_mgmt = &(sm_bcol->colls_with_user_data); - - /* allocate memory for pointers to mine and my peers' payload buffers - */ - malloc_size = ml_block->num_banks*ml_block->num_buffers_per_bank* - pload_mgmt->size_of_group *sizeof(void *); - pload_mgmt->data_buffs = malloc(malloc_size); - if( !pload_mgmt->data_buffs) { - ret = OMPI_ERR_OUT_OF_RESOURCE; - goto exit_ERROR; - } - - /* setup the input file for the shared memory connection manager */ - input_file.file_name = sm_reg_data->file_name; - input_file.size = sm_reg_data->size; - input_file.size_ctl_structure = 0; - input_file.data_seg_alignment = BASESMUMA_CACHE_LINE_SIZE; - input_file.mpool_size = sm_reg_data->size; - - /* call the connection manager and map my shared memory peers' file - */ - ret = ompi_common_smcm_allgather_connection( - sm_bcol, - sm_bcol->super.sbgp_partner_module, - &(cs->sm_connections_list), - &(sm_bcol->payload_backing_files_info), - sm_bcol->super.sbgp_partner_module->group_comm, - input_file, - false); - if( OMPI_SUCCESS != ret ) { - goto exit_ERROR; - } - - /* now we exchange offset info - don't assume symmetric virtual memory - */ - mem_offset = (uint64_t)(ml_block->block->base_addr) - - (uint64_t)(cs->sm_payload_structs->data_addr); - - /* call into the exchange offsets function */ - ret = base_bcol_basesmuma_exchange_offsets(sm_bcol_module, - (void **)pload_mgmt->data_buffs, mem_offset, 0, - pload_mgmt->size_of_group); - if( OMPI_SUCCESS != ret ) { - goto exit_ERROR; - } - - /* convert memory offset to virtual address in current rank */ - leading_dim = pload_mgmt->size_of_group; - loop_limit = ml_block->num_banks*ml_block->num_buffers_per_bank; - for (i=0;i< sm_bcol_module->super.sbgp_partner_module->group_size;i++) { - - /* get the base pointer */ - int array_id=SM_ARRAY_INDEX(leading_dim,0,i); - if( i == sm_bcol_module->super.sbgp_partner_module->my_index) { - /* me */ - base_ptr=cs->sm_payload_structs->map_addr; - } else { - base_ptr=sm_bcol_module->payload_backing_files_info[i]-> - sm_mmap->map_addr; - } - pload_mgmt->data_buffs[array_id]=(void *) - (((uint64_t)pload_mgmt->data_buffs[array_id])+(uint64_t)base_ptr); - for( buf_id = 1 ; buf_id < loop_limit ; buf_id++ ) { - int array_id_m1=SM_ARRAY_INDEX(leading_dim,(buf_id-1),i); - array_id=SM_ARRAY_INDEX(leading_dim,buf_id,i); - pload_mgmt->data_buffs[array_id]=(void *) ((uint64_t)(pload_mgmt->data_buffs[array_id_m1])+ - (uint64_t)ml_block->size_buffer); - } - } - - /* setup the data structures needed for releasing the payload - * buffers back to the ml level - */ - for(j = 0; j < ml_block->num_banks; j++) { - sm_bcol->colls_with_user_data. - ctl_buffs_mgmt[j].nb_barrier_desc.ml_memory_block_descriptor= - ml_block; - } - - return OMPI_SUCCESS; - -exit_ERROR: - return ret; -} -#endif - /* * Allocate buffers for storing non-blocking collective descriptions, required * for making code re-entrant @@ -337,9 +215,10 @@ /* New init function used for new control scheme where we put the control * struct at the top of the payload buffer */ -int bcol_basesmuma_bank_init_opti(struct mca_coll_ml_module_t *ml_module, - mca_bcol_base_module_t *bcol_module, - void *reg_data) +int bcol_basesmuma_bank_init_opti(struct mca_bcol_base_memory_block_desc_t *payload_block, + uint32_t data_offset, + mca_bcol_base_module_t *bcol_module, + void *reg_data) { /* assumption here is that the block has been registered with * sm bcol hence has been mapped by each process, need to be @@ -354,8 +233,7 @@ (bcol_basesmuma_registration_data_t *) reg_data; mca_bcol_basesmuma_module_t *sm_bcol = (mca_bcol_basesmuma_module_t *) bcol_module; - ml_memory_block_desc_t *ml_block = - ml_module->payload_block; + mca_bcol_base_memory_block_desc_t *ml_block = payload_block; size_t malloc_size; bcol_basesmuma_smcm_file_t input_file; uint64_t mem_offset; @@ -374,7 +252,7 @@ /* go ahead and get the header size that is cached on the payload block */ - sm_bcol->total_header_size = ml_module->data_offset; + sm_bcol->total_header_size = data_offset; /* allocate memory for pointers to mine and my peers' payload buffers * difference here is that now we use our new data struct @@ -449,7 +327,7 @@ /* second, calculate where to set the data pointer */ pload_mgmt->data_buffs[array_id].payload=(void *) (uintptr_t)((uint64_t)(uintptr_t) pload_mgmt->data_buffs[array_id].ctl_struct + - (uint64_t)(uintptr_t) ml_module->data_offset); + (uint64_t)(uintptr_t) data_offset); for( buf_id = 1 ; buf_id < loop_limit ; buf_id++ ) { int array_id_m1=SM_ARRAY_INDEX(leading_dim,(buf_id-1),i); @@ -464,7 +342,7 @@ /* second, set the payload pointer */ pload_mgmt->data_buffs[array_id].payload =(void *) (uintptr_t)((uint64_t)(uintptr_t) pload_mgmt->data_buffs[array_id].ctl_struct + - (uint64_t)(uintptr_t) ml_module->data_offset); + (uint64_t)(uintptr_t) data_offset); } } @@ -513,7 +391,7 @@ ml_mem->num_banks, ml_mem->num_buffers_per_bank, ml_mem->size_buffer, - ml_module->data_offset, + data_offset, sm_bcol_module->super.sbgp_partner_module->group_size, sm_bcol_module->pow_k)) { @@ -542,7 +420,7 @@ */ mca_bcol_basesmuma_module_t *sm_bcol_module int bcol_basesmuma_free_payload_buff( - struct ml_memory_block_desc_t *block, + struct mca_bcol_base_memory_block_desc_t *block, sm_buffer_mgmt *ctl_mgmt, uint64_t buff_id) { Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_component.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_component.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_component.c (working copy) @@ -23,7 +23,6 @@ #include "ompi/mca/mpool/base/base.h" #include "ompi/mca/bcol/bcol.h" #include "ompi/mca/bcol/base/base.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "bcol_basesmuma.h" /* Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_fanin.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_fanin.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_fanin.c (working copy) @@ -18,6 +18,7 @@ #include "opal/sys/atomic.h" +#include "ompi/mca/bcol/base/base.h" #include "bcol_basesmuma.h" /********************************************************************************/ @@ -25,7 +26,7 @@ /********************************************************************************/ static int bcol_basesmuma_fanin_new(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int64_t sequence_number; @@ -107,7 +108,7 @@ } static int bcol_basesmuma_fanin_new_progress(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int64_t sequence_number; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_fanout.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_fanout.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_fanout.c (working copy) @@ -18,6 +18,7 @@ #include "opal/sys/atomic.h" +#include "ompi/mca/bcol/base/base.h" #include "bcol_basesmuma.h" /***********************************************************************************/ @@ -26,7 +27,7 @@ static int bcol_basesmuma_fanout_new( bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int64_t sequence_number; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_gather.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_gather.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_gather.c (working copy) @@ -12,6 +12,7 @@ */ #include "ompi_config.h" +#include "ompi/mca/bcol/base/base.h" #include "ompi/mca/bcol/basesmuma/bcol_basesmuma.h" #include "ompi/constants.h" #include "ompi/datatype/ompi_datatype.h" @@ -53,7 +54,7 @@ } int bcol_basesmuma_k_nomial_gather_init(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int leading_dim, buff_idx, idx; @@ -217,7 +218,7 @@ int bcol_basesmuma_k_nomial_gather_progress(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int group_size; @@ -457,7 +458,7 @@ /* original, fully blocking, fully synchronous gather - should result in worst performance when used */ #if 0 int bcol_basesmuma_k_nomial_gather(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int group_size; @@ -753,7 +754,7 @@ #if 0 /* blocking, asynchronous polling gather routine */ int bcol_basesmuma_k_nomial_gather(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int group_size; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_bcast.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_bcast.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_bcast.c (working copy) @@ -43,7 +43,7 @@ static int completed_scatter = 0; #if 0 int bcol_basesmuma_lmsg_scatter_allgather_portals_bcast_old(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ @@ -589,7 +589,7 @@ */ int bcol_basesmuma_lmsg_scatter_allgather_portals_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ @@ -1113,7 +1113,7 @@ */ int bcol_basesmuma_lmsg_scatter_allgather_portals_nb_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { int i; mca_bcol_basesmuma_portal_proc_info_t *portals_info; @@ -1526,7 +1526,7 @@ int bcol_basesmuma_lmsg_scatter_allgather_portals_nb_knownroot_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { int i; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_bcast.h =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_bcast.h (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_bcast.h (working copy) @@ -615,12 +615,12 @@ } int bcol_basesmuma_lmsg_scatter_allgather_portals_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_lmsg_scatter_allgather_portals_nb_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_lmsg_scatter_allgather_portals_nb_knownroot_bcast(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); #endif Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_knomial_bcast.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_knomial_bcast.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_lmsg_knomial_bcast.c (working copy) @@ -46,7 +46,7 @@ * @param module - basesmuma module. */ int bcol_basesmuma_lmsg_bcast_k_nomial_anyroot(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { #if 0 /* local variables */ @@ -268,7 +268,7 @@ int bcol_basesmuma_lmsg_bcast_k_nomial_anyroot(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ mca_bcol_basesmuma_module_t* bcol_module= Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_rd_barrier.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_rd_barrier.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_rd_barrier.c (working copy) @@ -24,7 +24,7 @@ #if 0 int bcol_basesmuma_recursive_double_barrier(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_rd_nb_barrier.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_rd_nb_barrier.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_rd_nb_barrier.c (working copy) @@ -21,6 +21,7 @@ #include "bcol_basesmuma.h" #include "opal/sys/atomic.h" #include "ompi/patterns/net/netpatterns.h" +#include "ompi/mca/bcol/base/base.h" /* * Initialize nonblocking barrier. This is code specific for handling @@ -369,7 +370,7 @@ } static int bcol_basesmuma_memsync(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { int rc; int memory_bank = input_args->root; @@ -414,7 +415,7 @@ } static int bcol_basesmuma_memsync_progress(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { int memory_bank = input_args->root; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_reduce.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_reduce.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_reduce.c (working copy) @@ -29,7 +29,7 @@ */ static int bcol_basesmuma_reduce_intra_fanin_progress(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_reduce_init(mca_bcol_base_module_t *super) { @@ -112,7 +112,7 @@ } static int bcol_basesmuma_reduce_intra_fanin_progress(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { mca_bcol_basesmuma_module_t* bcol_module = (mca_bcol_basesmuma_module_t *)c_input_args->bcol_module; @@ -157,7 +157,7 @@ } int bcol_basesmuma_reduce_intra_fanin(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int rc=BCOL_FN_COMPLETE; @@ -261,7 +261,7 @@ * complete? */ int bcol_basesmuma_reduce_intra_fanin_old(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args) + mca_bcol_base_function_t *c_input_args) { /* local variables */ int rc=OMPI_SUCCESS; Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_reduce.h =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_reduce.h (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_reduce.h (working copy) @@ -70,10 +70,10 @@ volatile mca_bcol_basesmuma_payload_t *data_buffs); int bcol_basesmuma_lmsg_reduce(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); int bcol_basesmuma_lmsg_reduce_extra(bcol_function_args_t *input_args, - coll_ml_function_t *c_input_args); + mca_bcol_base_function_t *c_input_args); void basesmuma_reduce_recv(int my_group_index, int peer, void *recv_buffer, Index: ompi/mca/bcol/basesmuma/bcol_basesmuma_rk_barrier.c =================================================================== --- ompi/mca/bcol/basesmuma/bcol_basesmuma_rk_barrier.c (revision 30606) +++ ompi/mca/bcol/basesmuma/bcol_basesmuma_rk_barrier.c (working copy) @@ -11,8 +11,8 @@ #include "ompi_config.h" #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "ompi/mca/bcol/bcol.h" +#include "ompi/mca/bcol/base/base.h" #include "ompi/mca/bcol/basesmuma/bcol_basesmuma.h" /* @@ -49,7 +49,7 @@ * */ int bcol_basesmuma_k_nomial_barrier_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variables */ int flag_offset = 0; @@ -236,7 +236,7 @@ /* allgather progress function */ int bcol_basesmuma_k_nomial_barrier_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { Index: ompi/mca/bcol/bcol.h =================================================================== --- ompi/mca/bcol/bcol.h (revision 30606) +++ ompi/mca/bcol/bcol.h (working copy) @@ -34,8 +34,6 @@ #endif /* Forward declaration - please do not remove it */ -struct ml_memory_block_desc_t; -struct mca_coll_ml_module_t; struct ml_buffers_t; struct mca_bcol_base_coll_fn_comm_attributes_t; @@ -318,8 +316,8 @@ /* forward declaration */ struct mca_coll_ml_descriptor_t; -struct ml_payload_buffer_desc_t; -struct mca_coll_ml_route_info_t; +struct mca_bcol_base_payload_buffer_desc_t; +struct mca_bcol_base_route_info_t; typedef struct { int order_num; /* Seq num of collective fragment */ @@ -352,7 +350,7 @@ * parameters */ /* Pasha: We don need this one for new flow - remove it */ struct mca_coll_ml_descriptor_t *full_message_descriptor; - struct mca_coll_ml_route_info_t *root_route; + struct mca_bcol_base_route_info_t *root_route; /* function status */ int function_status; /* root, for rooted operations */ @@ -361,8 +359,8 @@ void *sbuf; void *rbuf; void *userbuf; - struct ml_payload_buffer_desc_t *src_desc; - struct ml_payload_buffer_desc_t *dst_desc; + struct mca_bcol_base_payload_buffer_desc_t *src_desc; + struct mca_bcol_base_payload_buffer_desc_t *dst_desc; /* ml buffer size */ uint32_t buffer_size; /* index of buffer in ml payload cache */ @@ -389,6 +387,82 @@ }; +struct mca_bcol_base_route_info_t { + int level; + int rank; +}; +typedef struct mca_bcol_base_route_info_t mca_bcol_base_route_info_t; + +struct mca_bcol_base_lmngr_block_t { + opal_list_item_t super; + struct mca_coll_ml_lmngr_t *lmngr; + void* base_addr; +}; +typedef struct mca_bcol_base_lmngr_block_t mca_bcol_base_lmngr_block_t; +OBJ_CLASS_DECLARATION(mca_bcol_base_lmngr_block_t); + +struct mca_bcol_base_memory_block_desc_t { + + /* memory block for payload buffers */ + struct mca_bcol_base_lmngr_block_t *block; + + /* Address offset in bytes -- Indicates free memory in the block */ + uint64_t block_addr_offset; + + /* size of the memory block */ + size_t size_block; + + /* number of memory banks */ + uint32_t num_banks; + + /* number of buffers per bank */ + uint32_t num_buffers_per_bank; + + /* size of a payload buffer */ + uint32_t size_buffer; + + /* pointer to buffer descriptors initialized */ + struct mca_bcol_base_payload_buffer_desc_t *buffer_descs; + + /* index of the next free buffer in the block */ + uint64_t next_free_buffer; + + uint32_t *bank_release_counters; + + /* Counter that defines what bank should be synchronized next + * since collectives could be completed out of order, we have to make + * sure that memory synchronization collectives started in order ! */ + int memsync_counter; + + /* This arrays of flags used to signal that the bank is ready for recycling */ + bool *ready_for_memsync; + + /* This flags monitors if bank is open for usage. Usually we expect that user + * will do the check only on buffer-zero allocation */ + bool *bank_is_busy; + +}; + +/* convenience typedef */ +typedef struct mca_bcol_base_memory_block_desc_t mca_bcol_base_memory_block_desc_t; + +typedef void (*mca_bcol_base_release_buff_fn_t)(struct mca_bcol_base_memory_block_desc_t *ml_memblock, uint32_t buff_id); + +struct mca_bcol_base_payload_buffer_desc_t { + void *base_data_addr; /* buffer address */ + void *data_addr; /* buffer address + header offset */ + uint64_t generation_number; /* my generation */ + uint64_t bank_index; /* my bank */ + uint64_t buffer_index; /* my buff index */ +}; +/* convenience typedef */ +typedef struct mca_bcol_base_payload_buffer_desc_t mca_bcol_base_payload_buffer_desc_t; + + + + + + typedef struct bcol_function_args_t bcol_function_args_t; @@ -411,9 +485,9 @@ /* collective function prototype - all functions have the same interface * so that we can call them via a function pointer */ -struct coll_ml_function_t; +struct mca_bcol_base_function_t; typedef int (*mca_bcol_base_module_collective_fn_primitives_t) - (bcol_function_args_t *input_args, struct coll_ml_function_t *const_args); + (bcol_function_args_t *input_args, struct mca_bcol_base_function_t *const_args); typedef int (*mca_bcol_base_module_collective_init_fn_primitives_t) (struct mca_bcol_base_module_t *bcol_module); @@ -533,11 +607,12 @@ * */ /*typedef int (*mca_bcol_base_init_memory_fn_t) - (struct ml_memory_block_desc_t *ml_block, void *reg_data);*/ + (struct mca_bcol_base_memory_block_desc_t *ml_block, void *reg_data);*/ typedef int (*mca_bcol_base_init_memory_fn_t) - (struct mca_coll_ml_module_t *ml_module, - struct mca_bcol_base_module_t *bcol_module, + (struct mca_bcol_base_memory_block_desc_t *payload_block, + uint32_t data_offset, + struct mca_bcol_base_module_t *bcol, void *reg_data); typedef int (*mca_common_allgather_init_fn_t) @@ -656,6 +731,41 @@ typedef struct mca_bcol_base_module_t mca_bcol_base_module_t; OMPI_DECLSPEC OBJ_CLASS_DECLARATION(mca_bcol_base_module_t); +/* function description */ +struct mca_bcol_base_function_t { + int fn_idx; + /* module */ + struct mca_bcol_base_module_t *bcol_module; + + /* + * The following two parameters are used for bcol modules + * that want to do some optimizations based on the fact that + * n functions from the same bcol module are called in a row. + * For example, in the iboffload case, on the first call one + * will want to initialize the MWR, and start to instantiate + * it, but only post it at the end of the last call. + * The index of this function in a sequence of consecutive + * functions from the same bcol + */ + int index_in_consecutive_same_bcol_calls; + + /* number of times functions from this bcol are + * called in order + */ + int n_of_this_type_in_a_row; + + /* + * number of times functions from this module are called in the + * collective operation. + */ + int n_of_this_type_in_collective; + int index_of_this_type_in_collective; +}; +typedef struct mca_bcol_base_function_t mca_bcol_base_function_t; + + + + struct mca_bcol_base_descriptor_t { ompi_free_list_item_t super; /* Vasily: will be described in the future */ Index: ompi/mca/bcol/iboffload/bcol_iboffload.h =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload.h (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload.h (working copy) @@ -28,8 +28,6 @@ #include "ompi/datatype/ompi_datatype.h" #include "ompi/datatype/ompi_datatype_internal.h" -#include "ompi/mca/coll/ml/coll_ml.h" - #include "ompi/mca/bcol/bcol.h" #include "ompi/mca/bcol/base/base.h" @@ -135,7 +133,7 @@ /* IB related information first */ struct mca_bcol_iboffload_rdma_info_t ib_info; /* back pointer to original ML memory descriptor */ - struct ml_memory_block_desc_t *ml_mem_desc; + struct mca_bcol_base_memory_block_desc_t *ml_mem_desc; /* Pasha: do we really need this one ?*/ /* caching ml memory descriptor configurations localy */ mca_bcol_iboffload_rdma_block_desc_t bdesc; @@ -579,7 +577,7 @@ */ int mca_bcol_iboffload_small_msg_bcast_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_barrier_intra_recursive_doubling_start( @@ -613,7 +611,7 @@ struct mca_bcol_iboffload_collreq_t *coll_request); int mca_bcol_iboffload_nb_memory_service_barrier_intra(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_coll_support_all_types(bcol_coll coll_name); int mca_bcol_iboffload_coll_supported(int op, int dtype, bcol_elem_type elem_type); Index: ompi/mca/bcol/iboffload/bcol_iboffload_allgather.c =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_allgather.c (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_allgather.c (working copy) @@ -1170,7 +1170,7 @@ #if 0 static int mca_bcol_iboffload_neighbor_allgather_userbuffer_intra( bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *)const_args->bcol_module; @@ -1201,7 +1201,7 @@ #if 1 static int mca_bcol_iboffload_k_nomial_allgather_userbuffer_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *)const_args->bcol_module; @@ -1232,7 +1232,7 @@ #endif static int mca_bcol_iboffload_k_nomial_allgather_mlbuffer_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *)const_args->bcol_module; @@ -1266,7 +1266,7 @@ */ static int mca_bcol_iboffload_collreq_mlbuffer_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int i; mca_bcol_iboffload_collreq_t *coll_request = @@ -1294,7 +1294,7 @@ static int mca_bcol_iboffload_collreq_userbuffer_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int i; mca_bcol_iboffload_collreq_t *coll_request = Index: ompi/mca/bcol/iboffload/bcol_iboffload_allreduce.c =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_allreduce.c (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_allreduce.c (working copy) @@ -1261,7 +1261,7 @@ } static int mca_bcol_iboffload_allreduce_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variables */ int rc; @@ -1304,7 +1304,7 @@ static int mca_bcol_iboffload_allreduce_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_collreq_t *coll_request = (mca_bcol_iboffload_collreq_t *) Index: ompi/mca/bcol/iboffload/bcol_iboffload_barrier.c =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_barrier.c (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_barrier.c (working copy) @@ -324,7 +324,7 @@ } int mca_bcol_iboffload_nb_memory_service_barrier_intra(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variables */ @@ -818,7 +818,7 @@ static int mca_bcol_iboffload_new_style_barrier_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_collreq_t *coll_request = (mca_bcol_iboffload_collreq_t *) @@ -840,7 +840,7 @@ static int mca_bcol_iboffload_new_style_barrier_intra( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variables */ int rc; Index: ompi/mca/bcol/iboffload/bcol_iboffload_bcast.c =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_bcast.c (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_bcast.c (working copy) @@ -177,7 +177,7 @@ int mca_bcol_iboffload_small_msg_bcast_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_collreq_t *coll_request = (mca_bcol_iboffload_collreq_t *) @@ -369,7 +369,7 @@ } int mca_bcol_iboffload_small_msg_bcast_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *) const_args->bcol_module; @@ -501,7 +501,7 @@ } int mca_bcol_iboffload_small_msg_bcast_extra_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *)const_args->bcol_module; @@ -528,7 +528,7 @@ /* Large message scatter-allgather with zero copy */ int mca_bcol_iboffload_zero_copy_progress(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int i; mca_bcol_iboffload_collreq_t *coll_request = @@ -830,7 +830,7 @@ } int mca_bcol_iboffload_bcast_scatter_allgather_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *) const_args->bcol_module; @@ -977,7 +977,7 @@ } int mca_bcol_iboffload_bcast_scatter_allgather_extra_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *) const_args->bcol_module; Index: ompi/mca/bcol/iboffload/bcol_iboffload_bcast.h =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_bcast.h (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_bcast.h (working copy) @@ -26,17 +26,17 @@ int mca_bcol_iboffload_small_msg_bcast_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_small_msg_bcast_extra_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_small_msg_bcast_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_bcast_scatter_allgather_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_zero_copy_progress(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_bcast_scatter_allgather_extra_intra(bcol_function_args_t *fn_arguments, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int mca_bcol_iboffload_bcast_register(mca_bcol_base_module_t *super); static inline __opal_attribute_always_inline__ int Index: ompi/mca/bcol/iboffload/bcol_iboffload_fanin.c =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_fanin.c (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_fanin.c (working copy) @@ -233,7 +233,7 @@ ***********************************************************************/ static int mca_bcol_iboffload_new_style_fanin_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_collreq_t *coll_request = (mca_bcol_iboffload_collreq_t *) @@ -293,7 +293,7 @@ static int mca_bcol_iboffload_new_style_fanin_intra( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int rc = OMPI_SUCCESS; Index: ompi/mca/bcol/iboffload/bcol_iboffload_fanout.c =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_fanout.c (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_fanout.c (working copy) @@ -232,7 +232,7 @@ ***********************************************************************/ static int mca_bcol_iboffload_new_style_fanout_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_iboffload_collreq_t *coll_request = (mca_bcol_iboffload_collreq_t *) @@ -292,7 +292,7 @@ static int mca_bcol_iboffload_new_style_fanout_intra( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int rc = OMPI_SUCCESS; Index: ompi/mca/bcol/iboffload/bcol_iboffload_module.c =================================================================== --- ompi/mca/bcol/iboffload/bcol_iboffload_module.c (revision 30606) +++ ompi/mca/bcol/iboffload/bcol_iboffload_module.c (working copy) @@ -614,7 +614,7 @@ mca_bcol_iboffload_module_t *iboffload_module = (mca_bcol_iboffload_module_t *) bcol; mca_bcol_iboffload_local_rdma_block_t *rdma_block = &iboffload_module->rdma_block; - struct ml_memory_block_desc_t *desc = ml_module->payload_block; + struct mca_bcol_base_memory_block_desc_t *desc = ml_module->payload_block; struct ibv_mr *mr = (struct ibv_mr *) desc->block->lmngr->reg_desc[bcol->context_index]; int i; Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll.h =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll.h (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll.h (working copy) @@ -247,8 +247,6 @@ struct mca_bcol_ptpcoll_local_mlmem_desc_t { /* Bank index to release */ uint32_t bank_index_for_release; - /* back pointer to original ML memory descriptor */ - struct ml_memory_block_desc_t *ml_mem_desc; /* number of memory banks */ uint32_t num_banks; /* number of buffers per bank */ @@ -382,9 +380,9 @@ /* barrier routines */ int bcol_ptpcoll_barrier_recurs_dbl(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_barrier_recurs_knomial(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_barrier_init(mca_bcol_base_module_t *super); int mca_bcol_ptpcoll_memsync_init(mca_bcol_base_module_t *super); void * bcol_ptpcoll_allocate_memory(size_t length, size_t alignment, @@ -398,16 +396,16 @@ int bcol_ptpcoll_fanin( bcol_function_args_t *input_args, struct mca_bcol_base_module_t *module); int bcol_ptpcoll_fanout( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /* allgather routine */ int bcol_ptpcoll_k_nomial_allgather_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /* allgather progress */ int bcol_ptpcoll_k_nomial_allgather_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /* allgather register */ int bcol_ptpcoll_allgather_init(mca_bcol_base_module_t *super); Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allgather.c =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allgather.c (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allgather.c (working copy) @@ -11,10 +11,8 @@ #include "ompi_config.h" #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "ompi/mca/bcol/bcol.h" #include "bcol_ptpcoll_allreduce.h" -#include "ompi/mca/coll/base/coll_tags.h" /* debug */ /* * Recursive K-ing allgather */ @@ -31,7 +29,7 @@ */ int bcol_ptpcoll_k_nomial_allgather_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variables */ @@ -307,7 +305,7 @@ /* allgather progress function */ int bcol_ptpcoll_k_nomial_allgather_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allreduce.c =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allreduce.c (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allreduce.c (working copy) @@ -14,7 +14,6 @@ #include "ompi_config.h" #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "ompi/mca/bcol/bcol.h" #include "bcol_ptpcoll_allreduce.h" @@ -75,7 +74,7 @@ } static int bcol_ptpcoll_allreduce_narraying_progress (bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *) const_args->bcol_module; void *data_buffer = (void *) ( (unsigned char *) input_args->sbuf + @@ -228,7 +227,7 @@ } int bcol_ptpcoll_allreduce_narraying_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args){ + struct mca_bcol_base_function_t *const_args){ mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; uint64_t sequence_number = input_args->sequence_num; @@ -696,7 +695,7 @@ int bcol_ptpcoll_allreduce_recursivek_scatter_reduce_allgather_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args){ + struct mca_bcol_base_function_t *const_args){ mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; struct ompi_op_t *op = input_args->op; @@ -895,7 +894,7 @@ } int bcol_ptpcoll_allreduce_recursivek_scatter_reduce_allgather_extra_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args){ + struct mca_bcol_base_function_t *const_args){ mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; struct ompi_op_t *op = input_args->op; Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allreduce.h =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allreduce.h (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_allreduce.h (working copy) @@ -34,7 +34,7 @@ int bcol_ptpcoll_allreduce_narraying_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_allreduce_recursivek_scatter_reduce(mca_bcol_ptpcoll_module_t *ptpcoll_module, const int buffer_index, void *sbuf, @@ -52,7 +52,7 @@ const int padded_start_byte); int bcol_ptpcoll_allreduce_recursivek_scatter_reduce_allgather_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int compute_knomial_allgather_offsets(int group_index, int count, struct @@ -74,7 +74,7 @@ const int count, struct ompi_datatype_t *dtype); int bcol_ptpcoll_allreduce_recursivek_scatter_reduce_allgather_extra_init(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_allreduce_init(mca_bcol_base_module_t *super); Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_barrier.c =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_barrier.c (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_barrier.c (working copy) @@ -14,7 +14,6 @@ #include "ompi_config.h" #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "bcol_ptpcoll.h" #include "bcol_ptpcoll_utils.h" @@ -30,7 +29,7 @@ /*****************************************************************************************/ static int bcol_ptpcoll_barrier_recurs_knomial_new( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ uint64_t sequence_number; @@ -220,7 +219,7 @@ static int bcol_ptpcoll_barrier_recurs_knomial_new_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ mca_bcol_ptpcoll_module_t *ptpcoll_module = @@ -364,7 +363,7 @@ static int bcol_ptpcoll_barrier_recurs_knomial_extra_new( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ uint64_t sequence_number; @@ -450,7 +449,7 @@ static int bcol_ptpcoll_barrier_recurs_dbl_new( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ uint64_t sequence_number; @@ -625,7 +624,7 @@ static int bcol_ptpcoll_barrier_recurs_dbl_new_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ mca_bcol_ptpcoll_module_t *ptp_module = @@ -758,7 +757,7 @@ static int bcol_ptpcoll_barrier_recurs_dbl_extra_new( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ uint64_t sequence_number; @@ -837,7 +836,7 @@ /* We have the same progress func for both cases (R-D and K-Nominal) */ static int bcol_ptpcoll_barrier_extra_node_progress( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ ompi_request_t **requests; Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_bcast.c =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_bcast.c (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_bcast.c (working copy) @@ -11,7 +11,6 @@ #include "ompi_config.h" #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "ompi/mca/bcol/bcol.h" #include "bcol_ptpcoll_bcast.h" #include "bcol_ptpcoll_utils.h" @@ -83,7 +82,7 @@ int bcol_ptpcoll_bcast_k_nomial_anyroot_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int completed = 0; int rc; @@ -112,7 +111,7 @@ /* K-nomial tree ( with any root ) algorithm */ int bcol_ptpcoll_bcast_k_nomial_anyroot(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; mca_bcol_ptpcoll_component_t *cm = &mca_bcol_ptpcoll_component; @@ -293,7 +292,7 @@ } static int bcol_ptpcoll_bcast_k_nomial_extra_known_and_anyroot(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -392,7 +391,7 @@ } static int bcol_ptpcoll_bcast_k_nomial_extra_known_and_anyroot_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int rc; int completed = 0; /* not completed */ @@ -467,7 +466,7 @@ int bcol_ptpcoll_bcast_k_nomial_known_root_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -596,7 +595,7 @@ } int bcol_ptpcoll_bcast_k_nomial_known_root(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -767,7 +766,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot_extra(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -866,7 +865,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot_extra_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int rc; int completed = 0; /* not completed */ @@ -920,7 +919,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1014,7 +1013,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1147,7 +1146,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1254,7 +1253,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1435,7 +1434,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root_extra(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1522,7 +1521,7 @@ } int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root_extra_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int rc; int completed = 0; /* not completed */ @@ -1544,7 +1543,7 @@ } static int bcol_ptpcoll_bcast_narray_knomial_scatter_gatther_known_root_progress( - bcol_function_args_t *input_args, struct coll_ml_function_t *const_args) + bcol_function_args_t *input_args, struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1676,7 +1675,7 @@ static int bcol_ptpcoll_bcast_narray_knomial_scatter_gatther_known_root(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1899,7 +1898,7 @@ /* Pasha : need to move this code to some common function */ static int bcol_ptpcoll_bcast_narray_knomial_scatter_gatther_known_root_extra(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -1984,7 +1983,7 @@ } static int bcol_ptpcoll_bcast_known_root_extra_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { int rc; int completed = 0; /* not completed */ @@ -2008,7 +2007,7 @@ static int bcol_ptpcoll_bcast_narray_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -2093,7 +2092,7 @@ } static int bcol_ptpcoll_bcast_narray(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_bcast.h =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_bcast.h (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_bcast.h (working copy) @@ -20,34 +20,34 @@ int bcol_ptpcoll_bcast_init(mca_bcol_base_module_t *super); int bcol_ptpcoll_bcast_k_nomial_anyroot (bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_k_nomial_anyroot_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_k_nomial_known_root(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_k_nomial_known_root_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot_extra(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_anyroot_extra_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root_extra(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); int bcol_ptpcoll_bcast_binomial_scatter_gatther_known_root_extra_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); /* macros */ Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_fanout.c =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_fanout.c (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_fanout.c (working copy) @@ -10,7 +10,6 @@ #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "ompi/mca/bcol/ptpcoll/bcol_ptpcoll.h" /* @@ -18,7 +17,7 @@ */ int bcol_ptpcoll_fanout( bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { /* local variable */ int ret = OMPI_SUCCESS; Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_module.c =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_module.c (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_module.c (working copy) @@ -30,8 +30,6 @@ #include "ompi/mca/bcol/base/base.h" #include "ompi/mca/pml/pml.h" /* need this for the max tag size */ -#include "ompi/mca/coll/ml/coll_ml.h" -#include "ompi/mca/coll/ml/coll_ml_allocation.h" #include "bcol_ptpcoll.h" #include "bcol_ptpcoll_utils.h" #include "bcol_ptpcoll_bcast.h" @@ -244,13 +242,14 @@ /* * Cache information about ML memory */ -static int mca_bcol_ptpcoll_cache_ml_memory_info(struct mca_coll_ml_module_t *ml_module, +static int mca_bcol_ptpcoll_cache_ml_memory_info(struct mca_bcol_base_memory_block_desc_t *payload_block, + uint32_t data_offset, struct mca_bcol_base_module_t *bcol, void *reg_data) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *) bcol; mca_bcol_ptpcoll_local_mlmem_desc_t *ml_mem = &ptpcoll_module->ml_mem; - struct ml_memory_block_desc_t *desc = ml_module->payload_block; + struct mca_bcol_base_memory_block_desc_t *desc = payload_block; int group_size = ptpcoll_module->super.sbgp_partner_module->group_size; PTPCOLL_VERBOSE(10, ("mca_bcol_ptpcoll_init_buffer_memory was called")); @@ -263,9 +262,6 @@ PTPCOLL_VERBOSE(10, ("ML buffer configuration num banks %d num_per_bank %d size %d base addr %p", desc->num_banks, desc->num_buffers_per_bank, desc->size_buffer, desc->block->base_addr)); - /* pointer to ml level descriptor */ - ml_mem->ml_mem_desc = desc; - /* Set first bank index for release */ ml_mem->bank_index_for_release = 0; @@ -274,7 +270,7 @@ ml_mem->num_banks, ml_mem->num_buffers_per_bank, ml_mem->size_buffer, - ml_module->data_offset, + data_offset, group_size, ptpcoll_module->pow_k)) { PTPCOLL_VERBOSE(10, ("Failed to allocate rdma memory descriptor\n")); @@ -281,8 +277,8 @@ return OMPI_ERROR; } - PTPCOLL_VERBOSE(10, ("ml_module = %p, ptpcoll_module = %p, ml_mem_desc = %p.\n", - ml_module, ptpcoll_module, ml_mem->ml_mem_desc)); + PTPCOLL_VERBOSE(10, ("ptpcoll_module = %p, ml_mem_desc = %p.\n", + ptpcoll_module)); return OMPI_SUCCESS; } Index: ompi/mca/bcol/ptpcoll/bcol_ptpcoll_reduce.c =================================================================== --- ompi/mca/bcol/ptpcoll/bcol_ptpcoll_reduce.c (revision 30606) +++ ompi/mca/bcol/ptpcoll/bcol_ptpcoll_reduce.c (working copy) @@ -14,16 +14,15 @@ #include "ompi_config.h" #include "ompi/include/ompi/constants.h" -#include "ompi/mca/coll/ml/coll_ml.h" #include "ompi/mca/bcol/bcol.h" #include "bcol_ptpcoll_reduce.h" #include "bcol_ptpcoll_utils.h" static int bcol_ptpcoll_reduce_narray_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); static int bcol_ptpcoll_reduce_narray(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args); + struct mca_bcol_base_function_t *const_args); #define NARRAY_RECV_NB(narray_node, process_shift, group_size, \ @@ -85,7 +84,7 @@ return OMPI_SUCCESS; } static int bcol_ptpcoll_reduce_narray_progress(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; @@ -226,7 +225,7 @@ } static int bcol_ptpcoll_reduce_narray(bcol_function_args_t *input_args, - struct coll_ml_function_t *const_args) + struct mca_bcol_base_function_t *const_args) { mca_bcol_ptpcoll_module_t *ptpcoll_module = (mca_bcol_ptpcoll_module_t *)const_args->bcol_module; Index: ompi/mca/coll/ml/coll_ml.h =================================================================== --- ompi/mca/coll/ml/coll_ml.h (revision 30606) +++ ompi/mca/coll/ml/coll_ml.h (working copy) @@ -101,40 +101,7 @@ }; struct mca_bcol_base_module_t; -/* function description */ -struct coll_ml_function_t { - int fn_idx; - /* module */ - struct mca_bcol_base_module_t *bcol_module; - /* - * The following two parameters are used for bcol modules - * that want to do some optimizations based on the fact that - * n functions from the same bcol module are called in a row. - * For example, in the iboffload case, on the first call one - * will want to initialize the MWR, and start to instantiate - * it, but only post it at the end of the last call. - * The index of this function in a sequence of consecutive - * functions from the same bcol - */ - int index_in_consecutive_same_bcol_calls; - - /* number of times functions from this bcol are - * called in order - */ - int n_of_this_type_in_a_row; - - /* - * number of times functions from this module are called in the - * collective operation. - */ - int n_of_this_type_in_collective; - int index_of_this_type_in_collective; -}; -typedef struct coll_ml_function_t coll_ml_function_t; - - - /* collective function arguments - gives * one function signature for calling all collective setup * routines, with the initial call to a collective function having @@ -205,7 +172,7 @@ mpi_coll_algorithm_params_t alg_params; /* list of functions */ - coll_ml_function_t *functions; + mca_bcol_base_function_t *functions; /* function names - for debugging */ char **function_names; @@ -599,12 +566,6 @@ */ OMPI_MODULE_DECLSPEC extern mca_coll_ml_component_t mca_coll_ml_component; -struct mca_coll_ml_route_info_t { - int level; - int rank; -}; -typedef struct mca_coll_ml_route_info_t mca_coll_ml_route_info_t; - struct mca_coll_ml_leader_offset_info_t { size_t offset; int level_one_index; @@ -623,7 +584,7 @@ int n_levels; /* bcols bits that describe supported features/modes */ uint64_t all_bcols_mode; - mca_coll_ml_route_info_t *route_vector; + mca_bcol_base_route_info_t *route_vector; coll_ml_collective_description_t *hierarchical_algorithms[BCOL_NUM_OF_FUNCTIONS]; sub_group_params_t *array_of_all_subgroups; /* (sbgp, bcol) pairs */ @@ -691,7 +652,7 @@ ompi_free_list_t fragment_descriptors; /** pointer to the payload memory block **/ - struct ml_memory_block_desc_t *payload_block; + struct mca_bcol_base_memory_block_desc_t *payload_block; /** the maximum size of collective function description */ int max_dag_size; Index: ompi/mca/coll/ml/coll_ml_allgather.c =================================================================== --- ompi/mca/coll/ml/coll_ml_allgather.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_allgather.c (working copy) @@ -133,7 +133,7 @@ size_t frag_len, dt_size; void *buf; - ml_payload_buffer_desc_t *src_buffer_desc; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc; mca_coll_ml_collective_operation_progress_t *new_op; mca_coll_ml_module_t *ml_module = OP_ML_MODULE(coll_op); @@ -296,7 +296,7 @@ int ret, n_fragments = 1, comm_size; mca_coll_ml_topology_t *topo_info; - ml_payload_buffer_desc_t *src_buffer_desc; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc; mca_coll_ml_component_t *cm = &mca_coll_ml_component; Index: ompi/mca/coll/ml/coll_ml_allocation.c =================================================================== --- ompi/mca/coll/ml/coll_ml_allocation.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_allocation.c (working copy) @@ -21,11 +21,11 @@ long memory_buffer_index; -ml_memory_block_desc_t *mca_coll_ml_allocate_block(struct mca_coll_ml_component_t *ml_component, - ml_memory_block_desc_t *ml_memblock) +mca_bcol_base_memory_block_desc_t *mca_coll_ml_allocate_block(struct mca_coll_ml_component_t *ml_component, + mca_bcol_base_memory_block_desc_t *ml_memblock) { - ml_memory_block_desc_t *ret = NULL; - ml_memory_block_desc_t *memory_block = NULL; + mca_bcol_base_memory_block_desc_t *ret = NULL; + mca_bcol_base_memory_block_desc_t *memory_block = NULL; mca_coll_ml_lmngr_t *memory_manager = NULL; if (ml_memblock) { @@ -32,7 +32,7 @@ ML_ERROR(("Memory already allocated - expecting NULL pointer")); return ret; } - memory_block = (ml_memory_block_desc_t*) calloc(1, sizeof(ml_memory_block_desc_t)); + memory_block = (mca_bcol_base_memory_block_desc_t*) calloc(1, sizeof(mca_bcol_base_memory_block_desc_t)); if (NULL == memory_block){ ML_ERROR(("Couldn't allocate memory for ml_memblock")); @@ -60,7 +60,7 @@ return ret; } -void mca_coll_ml_free_block (ml_memory_block_desc_t *ml_memblock) +void mca_coll_ml_free_block (mca_bcol_base_memory_block_desc_t *ml_memblock) { if (!ml_memblock) return; @@ -76,7 +76,7 @@ free(ml_memblock); } -int mca_coll_ml_initialize_block(ml_memory_block_desc_t *ml_memblock, +int mca_coll_ml_initialize_block(mca_bcol_base_memory_block_desc_t *ml_memblock, uint32_t num_buffers, uint32_t num_banks, uint32_t buffer_size, @@ -86,7 +86,7 @@ int ret = OMPI_SUCCESS; uint32_t bank_loop, buff_loop; uint64_t addr_offset = 0; - ml_payload_buffer_desc_t *pbuff_descs = NULL,*pbuff_desc = NULL; + mca_bcol_base_payload_buffer_desc_t *pbuff_descs = NULL,*pbuff_desc = NULL; if (NULL == ml_memblock){ ML_ERROR(("Memory block not initialized")); @@ -100,7 +100,7 @@ goto exit_ERROR; } - pbuff_descs = (ml_payload_buffer_desc_t*) malloc(sizeof(ml_payload_buffer_desc_t) + pbuff_descs = (mca_bcol_base_payload_buffer_desc_t*) malloc(sizeof(mca_bcol_base_payload_buffer_desc_t) * num_banks * num_buffers); for(bank_loop = 0; bank_loop < num_banks; bank_loop++) @@ -167,12 +167,12 @@ return ret; } -ml_payload_buffer_desc_t *mca_coll_ml_alloc_buffer (mca_coll_ml_module_t *module) +mca_bcol_base_payload_buffer_desc_t *mca_coll_ml_alloc_buffer (mca_coll_ml_module_t *module) { uint64_t bindex; uint32_t bank, buffer, num_buffers; - ml_memory_block_desc_t *ml_memblock = module->payload_block; - ml_payload_buffer_desc_t *pbuff_descs = NULL, + mca_bcol_base_memory_block_desc_t *ml_memblock = module->payload_block; + mca_bcol_base_payload_buffer_desc_t *pbuff_descs = NULL, *ml_membuffer = NULL; /* Return a buffer */ Index: ompi/mca/coll/ml/coll_ml_allocation.h =================================================================== --- ompi/mca/coll/ml/coll_ml_allocation.h (revision 30606) +++ ompi/mca/coll/ml/coll_ml_allocation.h (working copy) @@ -20,67 +20,6 @@ #include "ompi/mca/mpool/base/base.h" #include "coll_ml_lmngr.h" -typedef void (*mca_coll_ml_release_buff_fn_t)(struct ml_memory_block_desc_t *ml_memblock, uint32_t buff_id); - -struct ml_payload_buffer_desc_t { - void *base_data_addr; /* buffer address */ - void *data_addr; /* buffer address + header offset */ - uint64_t generation_number; /* my generation */ - uint64_t bank_index; /* my bank */ - uint64_t buffer_index; /* my buff index */ -}; -/* convenience typedef */ -typedef struct ml_payload_buffer_desc_t ml_payload_buffer_desc_t; - - -struct mca_coll_ml_lmngr_block_t; -struct ml_memory_block_desc_t { - - /* memory block for payload buffers */ - struct mca_coll_ml_lmngr_block_t *block; - - /* Address offset in bytes -- Indicates free memory in the block */ - uint64_t block_addr_offset; - - /* size of the memory block */ - size_t size_block; - - /* number of memory banks */ - uint32_t num_banks; - - /* number of buffers per bank */ - uint32_t num_buffers_per_bank; - - /* size of a payload buffer */ - uint32_t size_buffer; - - /* pointer to buffer descriptors initialized */ - ml_payload_buffer_desc_t *buffer_descs; - - /* index of the next free buffer in the block */ - uint64_t next_free_buffer; - - uint32_t *bank_release_counters; - - /* Counter that defines what bank should be synchronized next - * since collectives could be completed out of order, we have to make - * sure that memory synchronization collectives started in order ! */ - int memsync_counter; - - /* This arrays of flags used to signal that the bank is ready for recycling */ - bool *ready_for_memsync; - - /* This flags monitors if bank is open for usage. Usually we expect that user - * will do the check only on buffer-zero allocation */ - bool *bank_is_busy; - -}; -/* convenience typedef */ -typedef struct ml_memory_block_desc_t ml_memory_block_desc_t; - - - - /* Returns a block of memory from mpool @@ -99,15 +38,15 @@ struct mca_coll_ml_component_t; struct mca_coll_ml_module_t; -ml_memory_block_desc_t *mca_coll_ml_allocate_block( +mca_bcol_base_memory_block_desc_t *mca_coll_ml_allocate_block( struct mca_coll_ml_component_t *ml_component, - struct ml_memory_block_desc_t *ml_memblock + struct mca_bcol_base_memory_block_desc_t *ml_memblock ); /* Allocate the memory from mpool */ /* Register the memory block with bcols */ void mca_coll_ml_free_block( - ml_memory_block_desc_t *ml_memblock + mca_bcol_base_memory_block_desc_t *ml_memblock ); @@ -125,7 +64,7 @@ On Failure: OMPI_ERROR */ int mca_coll_ml_initialize_block( - ml_memory_block_desc_t *ml_memblock, + mca_bcol_base_memory_block_desc_t *ml_memblock, uint32_t num_buffers, uint32_t num_banks, uint32_t buffer_size, @@ -146,12 +85,12 @@ On Sucess: OMPI_SUCCESS On Failure: OMPI_ERROR */ -ml_payload_buffer_desc_t *mca_coll_ml_alloc_buffer( +mca_bcol_base_payload_buffer_desc_t *mca_coll_ml_alloc_buffer( struct mca_coll_ml_module_t *module); int mca_coll_ml_free_buffer( - ml_memory_block_desc_t *ml_memblock, - struct ml_payload_buffer_desc_t *ml_membuffer + mca_bcol_base_memory_block_desc_t *ml_memblock, + struct mca_bcol_base_payload_buffer_desc_t *ml_membuffer ); /* @@ -166,7 +105,7 @@ */ int mca_coll_ml_register_block_bcol( - ml_memory_block_desc_t *ml_memblock + mca_bcol_base_memory_block_desc_t *ml_memblock ); #endif /* MCA_ML_ALLOC_H */ Index: ompi/mca/coll/ml/coll_ml_allreduce.c =================================================================== --- ompi/mca/coll/ml/coll_ml_allreduce.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_allreduce.c (working copy) @@ -91,7 +91,7 @@ ptrdiff_t lb, extent; - ml_payload_buffer_desc_t *src_buffer_desc; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc; mca_coll_ml_collective_operation_progress_t *new_op; mca_coll_ml_module_t *ml_module = OP_ML_MODULE(coll_op); @@ -253,7 +253,7 @@ ptrdiff_t lb, extent; size_t pack_len, dt_size; - ml_payload_buffer_desc_t *src_buffer_desc; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc; mca_coll_ml_collective_operation_progress_t *coll_op; mca_coll_ml_component_t *cm = &mca_coll_ml_component; Index: ompi/mca/coll/ml/coll_ml_barrier.c =================================================================== --- ompi/mca/coll/ml/coll_ml_barrier.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_barrier.c (working copy) @@ -37,7 +37,7 @@ { ompi_free_list_item_t *item; mca_coll_ml_collective_operation_progress_t *coll_op; - ml_payload_buffer_desc_t *src_buffer_desc = NULL; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc = NULL; /* allocate an ml buffer for signaling purposes */ src_buffer_desc = mca_coll_ml_alloc_buffer(ml_module); Index: ompi/mca/coll/ml/coll_ml_bcast.c =================================================================== --- ompi/mca/coll/ml/coll_ml_bcast.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_bcast.c (working copy) @@ -177,7 +177,7 @@ int ret, frag_len; size_t max_data = 0; - ml_payload_buffer_desc_t *src_buffer_desc = NULL; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc = NULL; mca_coll_ml_collective_operation_progress_t *new_op = NULL; mca_coll_ml_task_setup_fn_t task_setup = NULL; mca_coll_ml_module_t *ml_module = OP_ML_MODULE(coll_op); @@ -315,7 +315,7 @@ size_t dt_size; void *buf; - ml_payload_buffer_desc_t *src_buffer_desc = NULL; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc = NULL; mca_coll_ml_collective_operation_progress_t *new_op = NULL; mca_coll_ml_task_setup_fn_t task_setup = NULL; @@ -448,7 +448,7 @@ mca_coll_ml_collective_operation_progress_t * coll_op = NULL; mca_coll_ml_module_t *ml_module = (mca_coll_ml_module_t *) module; - ml_payload_buffer_desc_t *src_buffer_desc = NULL; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc = NULL; mca_coll_ml_task_setup_fn_t task_setup; OPAL_PTRDIFF_TYPE lb, extent; @@ -710,7 +710,7 @@ mca_coll_ml_collective_operation_progress_t * coll_op = NULL; mca_coll_ml_compound_functions_t *fixed_schedule; mca_coll_ml_module_t *ml_module = (mca_coll_ml_module_t *) module; - ml_payload_buffer_desc_t *src_buffer_desc = NULL; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc = NULL; mca_bcol_base_coll_fn_desc_t *func; OPAL_PTRDIFF_TYPE lb, extent; @@ -813,7 +813,7 @@ func = fixed_schedule[fn_idx].bcol_function; ret = func->coll_fn(&coll_op->variable_fn_params, - (struct coll_ml_function_t *) &fixed_schedule[fn_idx].constant_group_data); + (struct mca_bcol_base_function_t *) &fixed_schedule[fn_idx].constant_group_data); /* set the coll_fn_started flag to true */ if (BCOL_FN_COMPLETE == ret) { /* done with this routine, bump the active counter */ Index: ompi/mca/coll/ml/coll_ml_colls.h =================================================================== --- ompi/mca/coll/ml/coll_ml_colls.h (revision 30606) +++ ompi/mca/coll/ml/coll_ml_colls.h (working copy) @@ -25,7 +25,7 @@ /* RLG - temp fix !!!! - really need to remove this, but right now do not want to change the signature of the collective primitives to - use coll_ml_utility_data_t rather than coll_ml_function_t */ + use coll_ml_utility_data_t rather than mca_bcol_base_function_t */ int dummy; /* module */ @@ -239,7 +239,7 @@ } mca_coll_ml_pending_type_t; /* Forward declaration */ -struct ml_payload_buffer_desc_t; +struct mca_bcol_base_payload_buffer_desc_t; /* Data structure used to track ML level collective operation * progress. */ @@ -348,7 +348,7 @@ struct full_message_t *message_descriptor; /* ML buffer descriptor attached to this buffer */ - struct ml_payload_buffer_desc_t *buffer_desc; + struct mca_bcol_base_payload_buffer_desc_t *buffer_desc; /* handle for collective progress, e.g. alltoall */ bcol_fragment_descriptor_t bcol_fragment_desc; Index: ompi/mca/coll/ml/coll_ml_component.c =================================================================== --- ompi/mca/coll/ml/coll_ml_component.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_component.c (working copy) @@ -159,9 +159,9 @@ const_args = &seq_coll_op->coll_schedule->component_functions[fn_idx].constant_group_data; /* RLG - note need to move to useing coll_ml_utility_data_t as - * collective argument, rather than coll_ml_function_t + * collective argument, rather than mca_bcol_base_function_t */ - rc = progress_fn(&(seq_coll_op->variable_fn_params), (coll_ml_function_t *)const_args); + rc = progress_fn(&(seq_coll_op->variable_fn_params), (mca_bcol_base_function_t *)const_args); if (BCOL_FN_COMPLETE == rc) { /* done with this routine */ seq_coll_op->sequential_routine.current_active_bcol_fn++; @@ -205,7 +205,7 @@ const_args = &task_status->ml_coll_operation->coll_schedule-> component_functions[INDEX(task_status)].constant_group_data; rc = progress_fn(&(task_status->ml_coll_operation->variable_fn_params), - (coll_ml_function_t *)const_args); + (mca_bcol_base_function_t *)const_args); if (BCOL_FN_COMPLETE == rc) { ML_VERBOSE(3, ("GOT BCOL_COMPLETED!!!!")); rc = mca_coll_ml_task_completion_processing(&task_status, ACTIVE_L); @@ -233,7 +233,7 @@ const_args = &task_status->ml_coll_operation->coll_schedule-> component_functions[INDEX(task_status)].constant_group_data; rc = coll_fn(&(task_status->ml_coll_operation->variable_fn_params), - (coll_ml_function_t *)const_args); + (mca_bcol_base_function_t *)const_args); if (BCOL_FN_COMPLETE == rc) { ML_VERBOSE(3, ("GOT BCOL_COMPLETED!")); rc = mca_coll_ml_task_completion_processing(&task_status, PENDING_L); Index: ompi/mca/coll/ml/coll_ml_hier_algorithms_ibarrier.c =================================================================== --- ompi/mca/coll/ml/coll_ml_hier_algorithms_ibarrier.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_hier_algorithms_ibarrier.c (working copy) @@ -66,8 +66,8 @@ hierarchical_algorithms[BCOL_IBARRIER]->n_buffers=0; /* allocate space for the functions */ - hierarchical_algorithms[BCOL_IBARRIER][0].functions=(coll_ml_function_t *) - malloc(sizeof(coll_ml_function_t)* + hierarchical_algorithms[BCOL_IBARRIER][0].functions=(mca_bcol_base_function_t *) + malloc(sizeof(mca_bcol_base_function_t)* hierarchical_algorithms[BCOL_IBARRIER][0].n_functions); if( NULL == hierarchical_algorithms[BCOL_IBARRIER][0].functions) { ret=OMPI_ERROR; Index: ompi/mca/coll/ml/coll_ml_hier_algorithms_setup.c =================================================================== --- ompi/mca/coll/ml/coll_ml_hier_algorithms_setup.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_hier_algorithms_setup.c (working copy) @@ -75,8 +75,8 @@ collective_alg->n_functions )); /* allocate space for the functions */ - collective_alg->functions = (coll_ml_function_t *) - malloc(sizeof(coll_ml_function_t) * collective_alg->n_functions); + collective_alg->functions = (mca_bcol_base_function_t *) + malloc(sizeof(mca_bcol_base_function_t) * collective_alg->n_functions); if( NULL == collective_alg->functions) { ML_ERROR(("Can't allocate memory.\n")); ret = OMPI_ERR_OUT_OF_RESOURCE; Index: ompi/mca/coll/ml/coll_ml_inlines.h =================================================================== --- ompi/mca/coll/ml/coll_ml_inlines.h (revision 30606) +++ ompi/mca/coll/ml/coll_ml_inlines.h (working copy) @@ -45,7 +45,7 @@ int mca_coll_ml_buffer_recycling(mca_coll_ml_collective_operation_progress_t *ml_request) { mca_coll_ml_module_t *ml_module = (mca_coll_ml_module_t *)ml_request->coll_module; - ml_memory_block_desc_t *ml_memblock = ml_module->payload_block; + mca_bcol_base_memory_block_desc_t *ml_memblock = ml_module->payload_block; uint64_t bank_index = ml_request->fragment_data.buffer_desc->bank_index; int rc; @@ -417,7 +417,7 @@ rc = func->bcol_function->coll_fn(&op_prog->variable_fn_params, /* Pasha: Need to update the prototype of the func, right now it is ugly hack for compilation */ - (struct coll_ml_function_t *)&func->constant_group_data); + (struct mca_bcol_base_function_t *)&func->constant_group_data); switch(rc) { case BCOL_FN_NOT_STARTED: /* put it on pending list */ @@ -604,7 +604,7 @@ bcol_func = (sched->component_functions[ifunc].bcol_function); ret = bcol_func->coll_fn(&coll_op->variable_fn_params, - (struct coll_ml_function_t *) &sched->component_functions[ifunc].constant_group_data); + (struct mca_bcol_base_function_t *) &sched->component_functions[ifunc].constant_group_data); if (BCOL_FN_COMPLETE == ret) { if (ifunc == n_fn - 1) { Index: ompi/mca/coll/ml/coll_ml_lmngr.c =================================================================== --- ompi/mca/coll/ml/coll_ml_lmngr.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_lmngr.c (working copy) @@ -89,22 +89,6 @@ construct_lmngr, destruct_lmngr); -static void lmngr_block_constructor(mca_coll_ml_lmngr_block_t *item) -{ - item->base_addr = NULL; -} - -static void lnmgr_block_destructor(mca_coll_ml_lmngr_block_t *item) -{ - /* I have nothing to do here */ -} - -OBJ_CLASS_INSTANCE(mca_coll_ml_lmngr_block_t, - opal_list_item_t, - lmngr_block_constructor, - lnmgr_block_destructor); - - int mca_coll_ml_lmngr_tune(mca_coll_ml_lmngr_t *lmngr, size_t block_size, size_t list_size, size_t alignment) { @@ -246,7 +230,7 @@ /* slice the memory to blocks */ addr = (unsigned char *) lmngr->base_addr; for(num_blocks = 0; num_blocks < (int)lmngr->list_size; num_blocks++) { - mca_coll_ml_lmngr_block_t *item = OBJ_NEW(mca_coll_ml_lmngr_block_t); + mca_bcol_base_lmngr_block_t *item = OBJ_NEW(mca_bcol_base_lmngr_block_t); item->base_addr = (void *)addr; item->lmngr = lmngr; /* ML_VERBOSE(10, ("Appending block # %d %p", num_blocks, (void *)addr)); */ @@ -260,7 +244,7 @@ return OMPI_SUCCESS; } -mca_coll_ml_lmngr_block_t* mca_coll_ml_lmngr_alloc ( +mca_bcol_base_lmngr_block_t* mca_coll_ml_lmngr_alloc ( mca_coll_ml_lmngr_t *lmngr) { int rc; @@ -282,10 +266,10 @@ return NULL; } - return (mca_coll_ml_lmngr_block_t *)opal_list_remove_first(list); + return (mca_bcol_base_lmngr_block_t *)opal_list_remove_first(list); } -void mca_coll_ml_lmngr_free(mca_coll_ml_lmngr_block_t *block) +void mca_coll_ml_lmngr_free(mca_bcol_base_lmngr_block_t *block) { opal_list_append(&block->lmngr->blocks_list, (opal_list_item_t *)block); } Index: ompi/mca/coll/ml/coll_ml_lmngr.h =================================================================== --- ompi/mca/coll/ml/coll_ml_lmngr.h (revision 30606) +++ ompi/mca/coll/ml/coll_ml_lmngr.h (working copy) @@ -55,14 +55,6 @@ typedef struct mca_coll_ml_lmngr_t mca_coll_ml_lmngr_t; OBJ_CLASS_DECLARATION(mca_coll_ml_lmngr_t); -struct mca_coll_ml_lmngr_block_t { - opal_list_item_t super; - mca_coll_ml_lmngr_t *lmngr; - void* base_addr; -}; -typedef struct mca_coll_ml_lmngr_block_t mca_coll_ml_lmngr_block_t; -OBJ_CLASS_DECLARATION(mca_coll_ml_lmngr_block_t); - /* read user defined parametres for list manager */ int mca_coll_ml_lmngr_reg(void); /* If programmer want to user other than default mca @@ -76,10 +68,10 @@ int mca_coll_ml_lmngr_append_nc(mca_coll_ml_lmngr_t *lmngr, bcol_base_network_context_t *nc); /* Allocate a block from memory list manager */ -mca_coll_ml_lmngr_block_t* mca_coll_ml_lmngr_alloc ( +mca_bcol_base_lmngr_block_t* mca_coll_ml_lmngr_alloc ( mca_coll_ml_lmngr_t *lmngr); /* Return block to list memory manager */ -void mca_coll_ml_lmngr_free (mca_coll_ml_lmngr_block_t *block); +void mca_coll_ml_lmngr_free (mca_bcol_base_lmngr_block_t *block); #endif Index: ompi/mca/coll/ml/coll_ml_memsync.c =================================================================== --- ompi/mca/coll/ml/coll_ml_memsync.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_memsync.c (working copy) @@ -27,7 +27,7 @@ static int mca_coll_ml_memsync_recycle_memory(mca_coll_ml_collective_operation_progress_t *coll_op) { mca_coll_ml_module_t *ml_module = (mca_coll_ml_module_t *)coll_op->coll_module; - ml_memory_block_desc_t *ml_memblock = ml_module->payload_block; + mca_bcol_base_memory_block_desc_t *ml_memblock = ml_module->payload_block; mca_coll_ml_collective_operation_progress_t *pending_op = NULL; int bank = coll_op->full_message.bank_index_to_recycle; int rc; Index: ompi/mca/coll/ml/coll_ml_module.c =================================================================== --- ompi/mca/coll/ml/coll_ml_module.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_module.c (working copy) @@ -219,7 +219,7 @@ /* gvm Leak FIX Remove fragment free list */ OBJ_DESTRUCT(&(module->fragment_descriptors)); OBJ_DESTRUCT(&(module->message_descriptors)); - /* push ml_memory_block_desc_t back on list manager */ + /* push mca_bcol_base_memory_block_desc_t back on list manager */ mca_coll_ml_free_block(module->payload_block); /* release the cinvertor if it was allocated */ if (NULL != module->reference_convertor) { @@ -510,7 +510,8 @@ for (j = 0; j < topo->component_pairs[i].num_bcol_modules; j++) { bcol_module = topo->component_pairs[i].bcol_modules[j]; if (NULL != bcol_module->bcol_memory_init) { - ret = bcol_module->bcol_memory_init(ml_module, + ret = bcol_module->bcol_memory_init(ml_module->payload_block, + ml_module->data_offset, bcol_module, (NULL != bcol_module->network_context) ? bcol_module->network_context->context_data: NULL); @@ -2378,8 +2379,8 @@ goto exit_ERROR; } - topo->route_vector = (mca_coll_ml_route_info_t *) - calloc(comm_size, sizeof(mca_coll_ml_route_info_t)); + topo->route_vector = (mca_bcol_base_route_info_t *) + calloc(comm_size, sizeof(mca_bcol_base_route_info_t)); if (NULL == topo->route_vector) { ML_VERBOSE(10, ("Cannot allocate memory.\n")); rc = OMPI_ERR_OUT_OF_RESOURCE; Index: ompi/mca/coll/ml/coll_ml_reduce.c =================================================================== --- ompi/mca/coll/ml/coll_ml_reduce.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_reduce.c (working copy) @@ -120,7 +120,7 @@ ptrdiff_t lb, extent; - ml_payload_buffer_desc_t *src_buffer_desc; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc; mca_coll_ml_collective_operation_progress_t *new_op; mca_coll_ml_module_t *ml_module = OP_ML_MODULE(coll_op); @@ -278,7 +278,7 @@ int large_data_reduce) { ptrdiff_t lb, extent; size_t pack_len, dt_size; - ml_payload_buffer_desc_t *src_buffer_desc = NULL; + mca_bcol_base_payload_buffer_desc_t *src_buffer_desc = NULL; mca_coll_ml_collective_operation_progress_t * coll_op = NULL; bool contiguous = ompi_datatype_is_contiguous_memory_layout(dtype, count); mca_coll_ml_component_t *cm = &mca_coll_ml_component; Index: ompi/mca/coll/ml/coll_ml_select.c =================================================================== --- ompi/mca/coll/ml/coll_ml_select.c (revision 30606) +++ ompi/mca/coll/ml/coll_ml_select.c (working copy) @@ -310,7 +310,7 @@ int mca_select_bcol_function(mca_bcol_base_module_t *bcol_module, int bcoll_type, bcol_function_args_t *bcol_fn_arguments, - coll_ml_function_t *ml_fn_arguments ) + mca_bcol_base_function_t *ml_fn_arguments ) { struct mca_bcol_base_coll_fn_desc_t *fn_filtered = NULL; Index: ompi/mca/coll/ml/coll_ml_select.h =================================================================== --- ompi/mca/coll/ml/coll_ml_select.h (revision 30606) +++ ompi/mca/coll/ml/coll_ml_select.h (working copy) @@ -19,7 +19,7 @@ int mca_select_bcol_function(mca_bcol_base_module_t *bcol_module, int bcoll_type, bcol_function_args_t *bcol_fn_arguments, - coll_ml_function_t *ml_fn_arguments ); + mca_bcol_base_function_t *ml_fn_arguments ); /* * Goes through the function table and filters the collectives functions * based on comm-time attributes.