C++ OBJ_DESTRUCT函数代码示例

static void mca_oob_ud_module_fini (mca_oob_ud_peer_t **peer)
{
    opal_process_name_t key;
    void *node1, *node2;
    int rc;

    rc = opal_proc_table_get_first_key (&mca_oob_ud_module.peers, &key,
                                        (void **) peer, &node1, &node2);
    if (OPAL_SUCCESS == rc) {
        do {
            if (NULL != *peer) {
                mca_oob_ud_peer_release (*peer);
            }
            rc = opal_proc_table_get_next_key (&mca_oob_ud_module.peers, &key,
                                               (void **) peer, node1, &node1, node2, &node2);
        } while (OPAL_SUCCESS == rc);
    }

    opal_proc_table_remove_all(&mca_oob_ud_module.peers);

    OBJ_DESTRUCT(&mca_oob_ud_module.peers);

    return;
}

//.........这里部分代码省略.........
                    ORTE_ERROR_LOG(ret);
                    goto relay;
                }

                if (ORTE_DAEMON_ADD_LOCAL_PROCS == command) {
                    OBJ_RELEASE(relay);
                    relay = OBJ_NEW(opal_buffer_t);
                    /* repack the command */
                    if (OPAL_SUCCESS != (ret = opal_dss.pack(relay, &command, 1, ORTE_DAEMON_CMD))) {
                        ORTE_ERROR_LOG(ret);
                        goto relay;
                    }
                    if (0 == flag) {
                        /* copy the remainder of the payload */
                        opal_dss.copy_payload(relay, buffer);
                        /* no - just return */
                        goto relay;
                    }
                }

                /* unpack the byte object */
                cnt=1;
                if (ORTE_SUCCESS != (ret = opal_dss.unpack(buffer, &bo, &cnt, OPAL_BYTE_OBJECT))) {
                    ORTE_ERROR_LOG(ret);
                    goto relay;
                }
                if (0 < bo->size) {
                    /* load it into a buffer */
                    OBJ_CONSTRUCT(&wireup, opal_buffer_t);
                    opal_dss.load(&wireup, bo->bytes, bo->size);
                    /* pass it for processing */
                    if (ORTE_SUCCESS != (ret = orte_routed.init_routes(ORTE_PROC_MY_NAME->jobid, &wireup))) {
                        ORTE_ERROR_LOG(ret);
                        OBJ_DESTRUCT(&wireup);
                        goto relay;
                    }
                    /* done with the wireup buffer - dump it */
                    OBJ_DESTRUCT(&wireup);
                }
                free(bo);
                if (ORTE_DAEMON_ADD_LOCAL_PROCS == command) {
                    /* copy the remainder of the payload */
                    opal_dss.copy_payload(relay, buffer);
                }
            }
        } else {
            ORTE_ERROR_LOG(ret);
            goto CLEANUP;
        }
    }

 relay:

    /* get the list of next recipients from the routed module */
    orte_routed.get_routing_list(&coll);

    /* if list is empty, no relay is required */
    if (opal_list_is_empty(&coll)) {
        OPAL_OUTPUT_VERBOSE((5, orte_grpcomm_base_framework.framework_output,
                             "%s grpcomm:direct:send_relay - recipient list is empty!",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
        OBJ_RELEASE(rly);
        goto CLEANUP;
    }

    /* send the message to each recipient on list, deconstructing it as we go */

//.........这里部分代码省略.........
        return ret;
    }

    /* wait for barrier to complete */
    OMPI_LAZY_WAIT_FOR_COMPLETION(coll->active);
    OBJ_RELEASE(coll);

    /* check for timing request - get stop time and report elapsed
     time if so */
    if (ompi_enable_timing && 0 == OMPI_PROC_MY_NAME->vpid) {
        gettimeofday(&ompistop, NULL);
        opal_output(0, "ompi_mpi_finalize[%ld]: time to execute barrier %ld usec",
                    (long)OMPI_PROC_MY_NAME->vpid,
                    (long int)((ompistop.tv_sec - ompistart.tv_sec)*1000000 +
                               (ompistop.tv_usec - ompistart.tv_usec)));
    }

    /*
     * Shutdown the Checkpoint/Restart Mech.
     */
    if (OMPI_SUCCESS != (ret = ompi_cr_finalize())) {
        OMPI_ERROR_LOG(ret);
    }

    /* Shut down any bindings-specific issues: C++, F77, F90 */

    /* Remove all memory associated by MPI_REGISTER_DATAREP (per
       MPI-2:9.5.3, there is no way for an MPI application to
       *un*register datareps, but we don't want the OMPI layer causing
       memory leaks). */
    while (NULL != (item = opal_list_remove_first(&ompi_registered_datareps))) {
        OBJ_RELEASE(item);
    }
    OBJ_DESTRUCT(&ompi_registered_datareps);

    /* Remove all F90 types from the hash tables. As the OBJ_DESTRUCT will
     * call a special destructor able to release predefined types, we can
     * simply call the OBJ_DESTRUCT on the hash table and all memory will
     * be correctly released.
     */
    OBJ_DESTRUCT( &ompi_mpi_f90_integer_hashtable );
    OBJ_DESTRUCT( &ompi_mpi_f90_real_hashtable );
    OBJ_DESTRUCT( &ompi_mpi_f90_complex_hashtable );

    /* Free communication objects */

    /* free file resources */
    if (OMPI_SUCCESS != (ret = ompi_file_finalize())) {
        return ret;
    }

    /* free window resources */
    if (OMPI_SUCCESS != (ret = ompi_win_finalize())) {
        return ret;
    }
    if (OMPI_SUCCESS != (ret = ompi_osc_base_finalize())) {
        return ret;
    }

    /* free pml resource */ 
    if(OMPI_SUCCESS != (ret = mca_pml_base_finalize())) { 
      return ret;
    }
    /* free communicator resources */
    if (OMPI_SUCCESS != (ret = ompi_comm_finalize())) {
        return ret;

static void des(ompi_orte_tracker_t *p)
{
    OBJ_DESTRUCT(&p->lock);
    OBJ_DESTRUCT(&p->cond);
}

static int s1_init(void)
{
    PMI_BOOL initialized;
    int spawned;
    int rc, ret = OPAL_ERROR;
    int i, rank, lrank, nrank;
    char *pmix_id, tmp[64];
    opal_value_t kv;
    char *str;
    uint32_t ui32;
    opal_process_name_t ldr;
    char **localranks=NULL;

    if (PMI_SUCCESS != (rc = PMI_Initialized(&initialized))) {
        OPAL_PMI_ERROR(rc, "PMI_Initialized");
        return OPAL_ERROR;
    }

    if (PMI_TRUE != initialized && PMI_SUCCESS != (rc = PMI_Init(&spawned))) {
        OPAL_PMI_ERROR(rc, "PMI_Init");
        return OPAL_ERROR;
    }

    // setup hash table
    opal_pmix_base_hash_init();

    // Initialize space demands
    rc = PMI_KVS_Get_value_length_max(&pmix_vallen_max);
    if (PMI_SUCCESS != rc) {
        OPAL_PMI_ERROR(rc, "PMI_KVS_Get_value_length_max");
        goto err_exit;
    }
    pmix_vallen_threshold = pmix_vallen_max * 3;
    pmix_vallen_threshold >>= 2;

    rc = PMI_KVS_Get_name_length_max(&pmix_kvslen_max);
    if (PMI_SUCCESS != rc) {
        OPAL_PMI_ERROR(rc, "PMI_KVS_Get_name_length_max");
        goto err_exit;
    }

    rc = PMI_KVS_Get_key_length_max(&pmix_keylen_max);
    if (PMI_SUCCESS != rc) {
        OPAL_PMI_ERROR(rc, "PMI_KVS_Get_key_length_max");
        goto err_exit;
    }

    // Initialize job environment information
    pmix_id = (char*)malloc(pmix_vallen_max);
    if (pmix_id == NULL) {
        ret = OPAL_ERR_OUT_OF_RESOURCE;
        goto err_exit;
    }
    /* Get domain id */
    if (PMI_SUCCESS != (rc = PMI_Get_kvs_domain_id(pmix_id, pmix_vallen_max))) {
        free(pmix_id);
        goto err_exit;
    }

    /* get our rank */
    ret = PMI_Get_rank(&rank);
    if( PMI_SUCCESS != ret ) {
        OPAL_PMI_ERROR(ret, "PMI_Get_rank");
        free(pmix_id);
        goto err_exit;
    }

    /* Slurm PMI provides the job id as an integer followed
     * by a '.', followed by essentially a stepid. The first integer
     * defines an overall job number. The second integer is the number of
     * individual jobs we have run within that allocation. */
    s1_pname.jobid = strtoul(pmix_id, &str, 10);
    s1_pname.jobid = (s1_pname.jobid << 16) & 0xffff0000;
    if (NULL != str) {
        ui32 = strtoul(str, NULL, 10);
        s1_pname.jobid |= (ui32 & 0x0000ffff);
    }
    ldr.jobid = s1_pname.jobid;
    s1_pname.vpid = rank;
    /* store our name in the opal_proc_t so that
     * debug messages will make sense - an upper
     * layer will eventually overwrite it, but that
     * won't do any harm */
    opal_proc_set_name(&s1_pname);
    opal_output_verbose(2, opal_pmix_base_framework.framework_output,
                        "%s pmix:s1: assigned tmp name",
                        OPAL_NAME_PRINT(s1_pname));

    OBJ_CONSTRUCT(&kv, opal_value_t);
    kv.key = strdup(OPAL_PMIX_JOBID);
    kv.type = OPAL_UINT32;
    kv.data.uint32 = s1_pname.jobid;
    if (OPAL_SUCCESS != (ret = opal_pmix_base_store(&OPAL_PROC_MY_NAME, &kv))) {
        OPAL_ERROR_LOG(ret);
        OBJ_DESTRUCT(&kv);
        goto err_exit;
    }
    OBJ_DESTRUCT(&kv);

    /* save it */
//.........这里部分代码省略.........

int
main(int argc, char *argv[]){
    int count;
    int msgsize;
    uint8_t *msg;
    int i, j, rc;
    orte_process_name_t peer;
    double maxpower;
    opal_buffer_t *buf;
    orte_rml_recv_cb_t blob;

    /*
     * Init
     */
    orte_init(&argc, &argv, ORTE_PROC_NON_MPI);

    if (argc > 1) {
        count = atoi(argv[1]);
        if (count < 0) {
            count = INT_MAX-1;
        }
    } else {
        count = MAX_COUNT;
    }

    peer.jobid = ORTE_PROC_MY_NAME->jobid;
    peer.vpid = ORTE_PROC_MY_NAME->vpid + 1;
    if (peer.vpid == orte_process_info.num_procs) {
        peer.vpid = 0;
    }

    for (j=1; j < count+1; j++) {
        /* rank0 starts ring */
        if (ORTE_PROC_MY_NAME->vpid == 0) {
            /* setup the initiating buffer - put random sized message in it */
            buf = OBJ_NEW(opal_buffer_t);

            maxpower = (double)(j%7);
            msgsize = (int)pow(10.0, maxpower);
            opal_output(0, "Ring %d message size %d bytes", j, msgsize);
            msg = (uint8_t*)malloc(msgsize);
            opal_dss.pack(buf, msg, msgsize, OPAL_BYTE);
            free(msg);
            orte_rml.send_buffer_nb(&peer, buf, MY_TAG, orte_rml_send_callback, NULL);

            /* wait for it to come around */
            OBJ_CONSTRUCT(&blob, orte_rml_recv_cb_t);
            blob.active = true;
            orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, MY_TAG,
                                    ORTE_RML_NON_PERSISTENT,
                                    orte_rml_recv_callback, &blob);
            ORTE_WAIT_FOR_COMPLETION(blob.active);
            OBJ_DESTRUCT(&blob);

            opal_output(0, "%s Ring %d completed", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), j);
        } else {
            /* wait for msg */
            OBJ_CONSTRUCT(&blob, orte_rml_recv_cb_t);
            blob.active = true;
            orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD, MY_TAG,
                                    ORTE_RML_NON_PERSISTENT,
                                    orte_rml_recv_callback, &blob);
            ORTE_WAIT_FOR_COMPLETION(blob.active);

            opal_output(0, "%s received message %d from %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), j, ORTE_NAME_PRINT(&blob.name));

            /* send it along */
            buf = OBJ_NEW(opal_buffer_t);
            opal_dss.copy_payload(buf, &blob.data);
            OBJ_DESTRUCT(&blob);
            msg_active = true;
            orte_rml.send_buffer_nb(&peer, buf, MY_TAG, send_callback, NULL);
            ORTE_WAIT_FOR_COMPLETION(msg_active);
        }
    }

    orte_finalize();

    return 0;
}

//.........这里部分代码省略.........

    /* Allocate the protection domain for the device */
    device->ib_pd = ibv_alloc_pd(device->ib_dev_context);
    if (NULL == device->ib_pd) {
        rc = OSHMEM_ERR_RESOURCE_BUSY;
        goto out;
    }

    /* Allocate memory */
    if (!rc) {
        void *addr = NULL;
        size_t size = getpagesize();
        struct ibv_mr *ib_mr = NULL;
        uint64_t access_flag = IBV_ACCESS_LOCAL_WRITE |
                          IBV_ACCESS_REMOTE_WRITE |
                          IBV_ACCESS_REMOTE_READ; 
        uint64_t exp_access_flag = 0;

        OBJ_CONSTRUCT(&device->ib_mr_array, opal_value_array_t);
        opal_value_array_init(&device->ib_mr_array, sizeof(struct ibv_mr *));

#if defined(MPAGE_ENABLE) && (MPAGE_ENABLE > 0)
        exp_access_flag = IBV_EXP_ACCESS_ALLOCATE_MR  |
                          IBV_EXP_ACCESS_SHARED_MR_USER_READ |
                          IBV_EXP_ACCESS_SHARED_MR_USER_WRITE; 
#endif /* MPAGE_ENABLE */

        struct ibv_exp_reg_mr_in in = {device->ib_pd, addr, size, access_flag|exp_access_flag, 0};
        ib_mr = ibv_exp_reg_mr(&in);
        if (NULL == ib_mr) {
            rc = OSHMEM_ERR_OUT_OF_RESOURCE;
        } else {
            device->ib_mr_shared = ib_mr;
            opal_value_array_append_item(&device->ib_mr_array, &ib_mr);
        }

#if defined(MPAGE_ENABLE) && (MPAGE_ENABLE > 0)
        if (!rc) {
            struct ibv_exp_reg_shared_mr_in in_smr;

            access_flag = IBV_ACCESS_LOCAL_WRITE |
                          IBV_ACCESS_REMOTE_WRITE |
                          IBV_ACCESS_REMOTE_READ|
                          IBV_EXP_ACCESS_NO_RDMA;

            addr = (void *)mca_sshmem_base_start_address;
            mca_sshmem_verbs_fill_shared_mr(&in_smr, device->ib_pd, device->ib_mr_shared->handle,  addr, access_flag);
            ib_mr = ibv_exp_reg_shared_mr(&in_smr);
            if (NULL == ib_mr) {
                mca_sshmem_verbs_component.has_shared_mr = 0;
            } else {
                opal_value_array_append_item(&device->ib_mr_array, &ib_mr);
                mca_sshmem_verbs_component.has_shared_mr = 1;
            }
        }
#endif /* MPAGE_ENABLE */
    }

    /* all is well - rainbows and butterflies */
    if (!rc) {
        *priority = mca_sshmem_verbs_component.priority;
        *module = (mca_base_module_t *)&mca_sshmem_verbs_module.super;
    }

out:
    if (device) {
        if (opal_value_array_get_size(&device->ib_mr_array)) {
            struct ibv_mr** array;
            struct ibv_mr* ib_mr = NULL;
            array = OPAL_VALUE_ARRAY_GET_BASE(&device->ib_mr_array, struct ibv_mr *);
            while (opal_value_array_get_size(&device->ib_mr_array) > 0) {
                ib_mr = array[0];
                ibv_dereg_mr(ib_mr);
                opal_value_array_remove_item(&device->ib_mr_array, 0);
            }

            if (device->ib_mr_shared) {
                device->ib_mr_shared = NULL;
            }
            OBJ_DESTRUCT(&device->ib_mr_array);
        }

        if (device->ib_pd) {
            ibv_dealloc_pd(device->ib_pd);
            device->ib_pd = NULL;
        }

        if(device->ib_dev_context) {
            ibv_close_device(device->ib_dev_context);
            device->ib_dev_context = NULL;
        }

        if(device->ib_devs) {
            ibv_free_device_list(device->ib_devs);
            device->ib_devs = NULL;
        }
    }

    return rc;
}

//.........这里部分代码省略.........
            ret = ORTE_ERROR;
            goto cleanup;
        }

        /* Do we have enough processors on the available nodes? 
         * Question: How do we get the required number of processors?
         */
        if ( (Status == NodeStatus_Ready) && (idle_processors > 0) ) {

            /* Get node name. */
            hr = pNode->get_Name(&node_name);
            if (FAILED(hr)) {
                OPAL_OUTPUT_VERBOSE((1, orte_ras_base.ras_output,
                                    "ras:ccp:pNode->get_Name failed."));
                ret = ORTE_ERROR;
                goto cleanup;
            }

            /* Get node processor architecture. */
            hr = pNode->get_ProcessorArchitecture(&node_arch);
            if (FAILED(hr)) {
                OPAL_OUTPUT_VERBOSE((1, orte_ras_base.ras_output,
                                    "ras:ccp:pNode->get_ProcessorArchitecture failed."));
                ret = ORTE_ERROR;
                goto cleanup;
            }

            /* Prevent duplicated nodes in the list*/
            for (item = opal_list_get_first(&new_nodes);
                 opal_list_get_end(&new_nodes) != item;
                 item = opal_list_get_next(item)) {

                node = (orte_node_t*) item;
                if (0 == strcmp(node->name, (char *)node_name)) {
                    ++node->slots;
                    OPAL_OUTPUT_VERBOSE((1, orte_ras_base.ras_output,
                                        "ras:ccp:allocate:discover: found -- bumped slots to %d",
                                        node->slots));
                    break;
                }
            }
            /* Did we find it? */

            if (opal_list_get_end(&new_nodes) == item) {

                /* Nope -- didn't find it, so add a new item to the list */

                OPAL_OUTPUT_VERBOSE((1, orte_ras_base.ras_output,
                                    "ras:ccp:allocate:discover: not found -- added to list"));

                node = OBJ_NEW(orte_node_t);

                /* The function _dupenv_s is much safer than getenv on Windows. */
                _dupenv_s(&node->username, &len, "username");

                node->name = _com_util::ConvertBSTRToString(node_name);
                node->launch_id = nodeid;
                node->slots_inuse = 0;
                node->slots_max = 0;
                node->slots = 1;
                opal_list_append(nodelist, &node->super);
            } 
            /* up the nodeid */
            nodeid++;
        }

        pNode->Release();
        VariantClear(&var);
    }

    pNodes->Release();

    if (nodeid > 0) ret = ORTE_SUCCESS;

    /* All done */
cleanup:

    if (ORTE_SUCCESS == ret) {
        OPAL_OUTPUT_VERBOSE((1, orte_ras_base.ras_output,
                            "ras:ccp:allocate:discover: success"));
    } else {
        OPAL_OUTPUT_VERBOSE((1, orte_ras_base.ras_output,
                            "ras:ccp:allocate:discover: failed (rc=%d)", ret));
    }

    OBJ_DESTRUCT(&new_nodes);
    SysFreeString(node_name);
    SysFreeString(node_arch);

    /* check for timing request - get stop time and report elapsed time if so */
    if (orte_timing) {
        gettimeofday(&stop, NULL);
        opal_output(0, "ras_ccp: time to allocate is %ld usec",
                    (long int)((stop.tv_sec - start.tv_sec)*1000000 +
                               (stop.tv_usec - start.tv_usec)));
        gettimeofday(&start, NULL);
    }
    
    return ret;
}

int vprotocol_pessimist_event_logger_connect(int el_rank, ompi_communicator_t **el_comm)
{
    int rc;
    opal_buffer_t buffer;
    char *port;
    orte_process_name_t el_proc;
    char *hnp_uri, *rml_uri;
    orte_rml_tag_t el_tag;
    char name[MPI_MAX_PORT_NAME];
    int rank;
    vprotocol_pessimist_clock_t connect_info[2];
    
    snprintf(name, MPI_MAX_PORT_NAME, VPROTOCOL_EVENT_LOGGER_NAME_FMT, el_rank);
    port = ompi_pubsub.lookup(name, MPI_INFO_NULL);
    if(NULL == port)
    {
        return OMPI_ERR_NOT_FOUND;
    }
    V_OUTPUT_VERBOSE(45, "Found port < %s >", port);
    
    /* separate the string into the HNP and RML URI and tag */
    if (OMPI_SUCCESS != (rc = ompi_dpm.parse_port(port, &hnp_uri, &rml_uri, &el_tag))) {
        ORTE_ERROR_LOG(rc);
        return rc;
    }
    /* extract the originating proc's name */
    if (ORTE_SUCCESS != (rc = orte_rml_base_parse_uris(rml_uri, &el_proc, NULL))) {
        ORTE_ERROR_LOG(rc);
        free(rml_uri); free(hnp_uri);
        return rc;
    }
    /* make sure we can route rml messages to the destination */
    if (OMPI_SUCCESS != (rc = ompi_dpm.route_to_port(hnp_uri, &el_proc))) {
        ORTE_ERROR_LOG(rc);
        free(rml_uri); free(hnp_uri);
        return rc;
    }
    free(rml_uri); free(hnp_uri);
    
    /* Send an rml message to tell the remote end to wake up and jump into 
     * connect/accept */
    OBJ_CONSTRUCT(&buffer, opal_buffer_t);
    rc = orte_rml.send_buffer(&el_proc, &buffer, el_tag+1, 0);
    if(ORTE_SUCCESS > rc) {
        ORTE_ERROR_LOG(rc);
        OBJ_DESTRUCT(&buffer);        
        return rc;
    }
    OBJ_DESTRUCT(&buffer);

    rc = ompi_dpm.connect_accept(MPI_COMM_SELF, 0, port, true, el_comm);
    if(OMPI_SUCCESS != rc) {
        ORTE_ERROR_LOG(rc);
    }
    
    /* Send Rank, receive max buffer size and max_clock back */
    MPI_Comm_rank(MPI_COMM_WORLD, &rank);
    rc = mca_pml_v.host_pml.pml_send(&rank, 1, MPI_INTEGER, 0, 
                                     VPROTOCOL_PESSIMIST_EVENTLOG_NEW_CLIENT_CMD,
                                     MCA_PML_BASE_SEND_STANDARD, 
                                     mca_vprotocol_pessimist.el_comm);
    if(OPAL_UNLIKELY(MPI_SUCCESS != rc))
        OMPI_ERRHANDLER_INVOKE(mca_vprotocol_pessimist.el_comm, rc,
                               __FILE__ ": failed sending event logger handshake");
    rc = mca_pml_v.host_pml.pml_recv(&connect_info, 2, MPI_UNSIGNED_LONG_LONG, 
                                     0, VPROTOCOL_PESSIMIST_EVENTLOG_NEW_CLIENT_CMD,
                                     mca_vprotocol_pessimist.el_comm, MPI_STATUS_IGNORE);
    if(OPAL_UNLIKELY(MPI_SUCCESS != rc))                                  \
        OMPI_ERRHANDLER_INVOKE(mca_vprotocol_pessimist.el_comm, rc,       \
                               __FILE__ ": failed receiving event logger handshake");   
    
    return rc;
}

static void ofacm_base_proc_destructor (opal_common_ofacm_base_proc_t *proc)
{
    OBJ_DESTRUCT(&proc->all_contexts);
}

int
main(int argc, char *argv[])
{
    orcm_alloc_t alloc, *aptr;
    orte_rml_recv_cb_t xfer;
    opal_buffer_t *buf;
    int rc, n;
    orcm_scd_cmd_flag_t command=ORCM_SESSION_REQ_COMMAND;
    orcm_alloc_id_t id;
    struct timeval tv;

    /* initialize, parse command line, and setup frameworks */
    orcm_osub_init(argc, argv);

    /* create an allocation request */
    OBJ_CONSTRUCT(&alloc, orcm_alloc_t);

    alloc.priority = 1;                                // session priority
    alloc.account = orcm_osub_globals.account;         // account to be charged
    alloc.name = orcm_osub_globals.name;               // user-assigned project name
    alloc.gid = orcm_osub_globals.gid;                 // group id to be run under
    alloc.max_nodes = orcm_osub_globals.max_nodes;     // max number of nodes
    alloc.max_pes = orcm_osub_globals.max_pes;         // max number of processing elements
    alloc.min_nodes = orcm_osub_globals.min_nodes;     // min number of nodes required
    alloc.min_pes = orcm_osub_globals.min_pes;         // min number of pe's required
    alloc.exclusive = orcm_osub_globals.exclusive;     // true if nodes to be exclusively allocated (i.e., not shared across sessions)
    alloc.interactive = orcm_osub_globals.interactive; // true if in interactive mode
    alloc.nodes = '\0';                                // regex of nodes to be used
    alloc.parent_name = ORTE_NAME_PRINT(ORTE_PROC_MY_NAME); // my_daemon_name
    alloc.parent_uri = '\0';                           // my_daemon uri address
    /* alloc.constraints = orcm_osub_globals.resources */ ; // list of resource constraints to be applied when selecting hosts
    alloc.hnpname = '\0'; //my hnp name
    alloc.hnpuri = '\0'; //my hnp uri

    alloc.caller_uid = getuid();   // caller uid, not from args
    alloc.caller_gid = getgid();   // caller gid, not from args

    if (NULL == orcm_osub_globals.starttime || 0 == strlen(orcm_osub_globals.starttime)) {
        gettimeofday(&tv,NULL);
        /* desired start time for allocation deafults to now */
        alloc.begin = tv.tv_sec;
    } else {
        /* TODO: eventually parse the string to figure out what user means, for now its now */
        gettimeofday(&tv,NULL);
        alloc.begin = tv.tv_sec;
    }

    if (NULL == orcm_osub_globals.walltime || 0 == strlen(orcm_osub_globals.walltime)) {
        /* desired walltime default to 10 min */
        alloc.walltime = 600;
    } else {
        /* get this in seconds for now, but will be parsed for more complexity later */
        alloc.walltime = (time_t)strtol(orcm_osub_globals.walltime, NULL, 10);                               // max execution time
    }

    /* setup to receive the result */
    OBJ_CONSTRUCT(&xfer, orte_rml_recv_cb_t);
    xfer.active = true;
    orte_rml.recv_buffer_nb(ORTE_NAME_WILDCARD,
                            ORCM_RML_TAG_SCD,
                            ORTE_RML_NON_PERSISTENT,
                            orte_rml_recv_callback, &xfer);

    /* send it to the scheduler */
    buf = OBJ_NEW(opal_buffer_t);
    /* pack the alloc command flag */
    if (OPAL_SUCCESS != (rc = opal_dss.pack(buf, &command,1, ORCM_SCD_CMD_T))) {
        ORTE_ERROR_LOG(rc);
        return rc;
    }

    aptr = &alloc;
    if (OPAL_SUCCESS != (rc = opal_dss.pack(buf, &aptr, 1, ORCM_ALLOC))) {
        ORTE_ERROR_LOG(rc);
        return rc;
    }
    if (ORTE_SUCCESS != (rc = orte_rml.send_buffer_nb(ORTE_PROC_MY_SCHEDULER, buf,
                                                      ORCM_RML_TAG_SCD,
                                                      orte_rml_send_callback, NULL))) {
        ORTE_ERROR_LOG(rc);
        OBJ_RELEASE(buf);
        OBJ_DESTRUCT(&xfer);
        return rc;
    }

    /* get our allocated jobid */
    n=1;
    ORTE_WAIT_FOR_COMPLETION(xfer.active);
    if (OPAL_SUCCESS != (rc = opal_dss.unpack(&xfer.data, &id, &n, ORCM_ALLOC_ID_T))) {
        ORTE_ERROR_LOG(rc);
        OBJ_DESTRUCT(&xfer);
        return rc;
    }
    opal_output(0, "RECEIVED ALLOC ID %d", (int)id);

    if (ORTE_SUCCESS != orcm_finalize()) {
        fprintf(stderr, "Failed orcm_finalize\n");
        exit(1);
    }

//.........这里部分代码省略.........

//.........这里部分代码省略.........
       if (NULL == component->fcollm_file_query) {
          opal_output_verbose(10, mca_fcoll_base_output,
                             "select: no query, ignoring the component");
       } else {
           /*
            * call the query function and see what it returns
            */ 
           module = component->fcollm_file_query (file, &priority);

           if (NULL == module ||
               NULL == module->fcoll_module_init) {
               /*
                * query did not return any action which can be used
                */ 
               opal_output_verbose(10, mca_fcoll_base_output,
                                  "select: query returned failure");
           } else {
               opal_output_verbose(10, mca_fcoll_base_output,
                                  "select: query returned priority %d",
                                  priority);
               /* 
                * is this the best component we have found till now?
                */
               if (priority > best_priority) {
                   best_priority = priority;
                   best_component = component;
               }

               om = OBJ_NEW(queried_module_t);
               /*
                * check if we have run out of space
                */
               if (NULL == om) {
                   OBJ_DESTRUCT(&queried);
                   return OMPI_ERR_OUT_OF_RESOURCE;
               }
               om->om_component = component;
               om->om_module = module; 
               opal_list_append(&queried, (opal_list_item_t *)om); 
           } /* end else of if (NULL == module) */
       } /* end else of if (NULL == component->fcollm_init) */
    } /* end for ... end of traversal */

    /* We have to remove empty out the selectable list if the selectable 
     * list was constructed as a duplicate and not as a pointer to the
     * mca_base_components_available list. So, check and destroy */

    if (was_selectable_constructed) {

        /* remove all the items first */
        for (item = opal_list_get_first(&mca_fcoll_base_components_available);
             item != opal_list_get_end(&mca_fcoll_base_components_available);
             item = next_item) {
             next_item = opal_list_get_next(item);
             OBJ_RELEASE (item);
        }
                
        /* release the list itself */
        OBJ_RELEASE (selectable);
        was_selectable_constructed = false;
    }

    /*
     * Now we have alist of components which successfully returned
     * their module struct.  One of these components has the best
     * priority. The rest have to be comm_unqueried to counter the

int orte_ess_base_app_setup(bool db_restrict_local)
{
    int ret;
    char *error = NULL;
    opal_value_t kv;

    /*
     * stdout/stderr buffering
     * If the user requested to override the default setting then do
     * as they wish.
     */
    if( orte_ess_base_std_buffering > -1 ) {
        if( 0 == orte_ess_base_std_buffering ) {
            setvbuf(stdout, NULL, _IONBF, 0);
            setvbuf(stderr, NULL, _IONBF, 0);
        }
        else if( 1 == orte_ess_base_std_buffering ) {
            setvbuf(stdout, NULL, _IOLBF, 0);
            setvbuf(stderr, NULL, _IOLBF, 0);
        }
        else if( 2 == orte_ess_base_std_buffering ) {
            setvbuf(stdout, NULL, _IOFBF, 0);
            setvbuf(stderr, NULL, _IOFBF, 0);
        }
    }

    /* if I am an MPI app, we will let the MPI layer define and
     * control the opal_proc_t structure. Otherwise, we need to
     * do so here */
    if (ORTE_PROC_NON_MPI) {
        orte_process_info.super.proc_name = *(opal_process_name_t*)ORTE_PROC_MY_NAME;
        orte_process_info.super.proc_hostname = strdup(orte_process_info.nodename);
        orte_process_info.super.proc_flags = OPAL_PROC_ALL_LOCAL;
        orte_process_info.super.proc_arch = opal_local_arch;
        opal_proc_local_set(&orte_process_info.super);
    }

    /* get an async event base - we use the opal_async one so
     * we don't startup extra threads if not needed */
    orte_event_base = opal_progress_thread_init(NULL);
    progress_thread_running = true;
    /* open and setup the state machine */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_state_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orte_state_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = orte_state_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "orte_state_base_select";
        goto error;
    }

    /* open the errmgr */
    if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_errmgr_base_framework, 0))) {
        ORTE_ERROR_LOG(ret);
        error = "orte_errmgr_base_open";
        goto error;
    }

    /* setup my session directory */
    if (orte_create_session_dirs) {
        OPAL_OUTPUT_VERBOSE((2, orte_ess_base_framework.framework_output,
                             "%s setting up session dir with\n\ttmpdir: %s\n\thost %s",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                             (NULL == orte_process_info.tmpdir_base) ? "UNDEF" : orte_process_info.tmpdir_base,
                             orte_process_info.nodename));
        if (ORTE_SUCCESS != (ret = orte_session_dir(true,
                                                    orte_process_info.tmpdir_base,
                                                    orte_process_info.nodename, NULL,
                                                    ORTE_PROC_MY_NAME))) {
            ORTE_ERROR_LOG(ret);
            error = "orte_session_dir";
            goto error;
        }
        /* Once the session directory location has been established, set
           the opal_output env file location to be in the
           proc-specific session directory. */
        opal_output_set_output_file_info(orte_process_info.proc_session_dir,
                                         "output-", NULL, NULL);
        /* store the session directory location */
        OBJ_CONSTRUCT(&kv, opal_value_t);
        kv.key = strdup(OPAL_PMIX_NSDIR);
        kv.type = OPAL_STRING;
        kv.data.string = strdup(orte_process_info.job_session_dir);
        if (OPAL_SUCCESS != (ret = opal_pmix.store_local(ORTE_PROC_MY_NAME, &kv))) {
            ORTE_ERROR_LOG(ret);
            OBJ_DESTRUCT(&kv);
            error = "opal pmix put job sessiondir";
            goto error;
        }
        OBJ_DESTRUCT(&kv);
        OBJ_CONSTRUCT(&kv, opal_value_t);
        kv.key = strdup(OPAL_PMIX_PROCDIR);
        kv.type = OPAL_STRING;
        kv.data.string = strdup(orte_process_info.proc_session_dir);
        if (OPAL_SUCCESS != (ret = opal_pmix.store_local(ORTE_PROC_MY_NAME, &kv))) {
            ORTE_ERROR_LOG(ret);
            OBJ_DESTRUCT(&kv);
            error = "opal pmix put proc sessiondir";
//.........这里部分代码省略.........

//.........这里部分代码省略.........
                }
#endif
                local_cellid = 0;
                need_cellid = false;
            }

            rds_item->cellid      = local_cellid;
            ras_item->node_cellid = local_cellid;

            new_attr = OBJ_NEW(orte_rds_cell_attr_t);
            if (NULL == new_attr) {
                ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
                return ORTE_ERR_OUT_OF_RESOURCE;
            }
            new_attr->keyval.key          = strdup(ORTE_RDS_NAME);
            new_attr->keyval.value = OBJ_NEW(orte_data_value_t);
            if (NULL == new_attr->keyval.value) {
                ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
                return ORTE_ERR_OUT_OF_RESOURCE;
            }
            new_attr->keyval.value->type   = ORTE_STRING;
            new_attr->keyval.value->data   = strdup(ras_item->node_name);
            opal_list_append(&(rds_item->attributes), &new_attr->super);

            new_attr = OBJ_NEW(orte_rds_cell_attr_t);
            if (NULL == new_attr) {
                ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
                return ORTE_ERR_OUT_OF_RESOURCE;
            }
            new_attr->keyval.key          = strdup(ORTE_CELLID_KEY);
            new_attr->keyval.value = OBJ_NEW(orte_data_value_t);
            if (NULL == new_attr->keyval.value) {
                ORTE_ERROR_LOG(ORTE_ERR_OUT_OF_RESOURCE);
                return ORTE_ERR_OUT_OF_RESOURCE;
            }
            new_attr->keyval.value->type   = ORTE_CELLID;
            if (ORTE_SUCCESS != (rc = orte_dss.copy(&(new_attr->keyval.value->data), &(rds_item->cellid), ORTE_CELLID))) {
                ORTE_ERROR_LOG(rc);
                return rc;
            }
            opal_list_append(&(rds_item->attributes), &new_attr->super);

            opal_list_append(&rds_updates, &rds_item->super);
        }

        /* Insert the new node into the RDS */
        rc = orte_rds.store_resource(&rds_updates);
        if (ORTE_SUCCESS != rc) {
            goto cleanup;
        }

        /* Then the RAS, since we can assume that any
         * resources listed in the hostfile have been
         * already allocated for our use.
         */
        rc = orte_ras_base_node_insert(&updates);
        if (ORTE_SUCCESS != rc) {
            goto cleanup;
        }
        
        /* and now, indicate that ORTE should override any oversubscribed conditions
         * based on local hardware limits since the user (a) might not have
         * provided us any info on the #slots for a node, and (b) the user
         * might have been wrong! If we don't check the number of local physical
         * processors, then we could be too aggressive on our sched_yield setting
         * and cause performance problems.
         */
        rc = orte_ras_base_set_oversubscribe_override(job);
        if (ORTE_SUCCESS != rc) {
            goto cleanup;
        }
    }

cleanup:
    if (NULL != mca_rds_hostfile_component.path) {
        free(mca_rds_hostfile_component.path);
        mca_rds_hostfile_component.path = NULL;
    }

    while(NULL != (item = opal_list_remove_first(&existing))) {
        OBJ_RELEASE(item);
    }

    while(NULL != (item = opal_list_remove_first(&updates))) {
        OBJ_RELEASE(item);
    }

    while (NULL != (rds_item = (orte_rds_cell_desc_t*)opal_list_remove_first(&rds_updates))) {
        while (NULL != (new_attr = (orte_rds_cell_attr_t*)opal_list_remove_first(&(rds_item->attributes)))) {
            OBJ_RELEASE(new_attr);
        }
        OBJ_RELEASE(rds_item);
    }

    OBJ_DESTRUCT(&existing);
    OBJ_DESTRUCT(&updates);
    OBJ_DESTRUCT(&rds_updates);

    return rc;
}

static void ompi_osc_rdma_replyreq_destruct(ompi_osc_rdma_replyreq_t *replyreq)
{
    OBJ_DESTRUCT(&(replyreq->rep_target_convertor));
}