C++ ORTE_NAME_PRINT函数代码示例

//.........这里部分代码省略.........
            ORTE_ERROR_LOG(ret);
            return ret;
        }
    } else {
        if (ORTE_SUCCESS != (ret = orte_init(&argc, &argv, ORTE_PROC_DAEMON))) {
            ORTE_ERROR_LOG(ret);
            return ret;
        }
    }
    /* finalize the OPAL utils. As they are opened again from orte_init->opal_init
     * we continue to have a reference count on them. So we have to finalize them twice...
     */
    opal_finalize_util();

    if ((int)ORTE_VPID_INVALID != orted_globals.fail) {
        orted_globals.abort=false;
        /* some vpid was ordered to fail. The value can be positive
         * or negative, depending upon the desired method for failure,
         * so need to check both here
         */
        if (0 > orted_globals.fail) {
            orted_globals.fail = -1*orted_globals.fail;
            orted_globals.abort = true;
        }
        /* are we the specified vpid? */
        if ((int)ORTE_PROC_MY_NAME->vpid == orted_globals.fail) {
            /* if the user specified we delay, then setup a timer
             * and have it kill us
             */
            if (0 < orted_globals.fail_delay) {
                ORTE_TIMER_EVENT(orted_globals.fail_delay, 0, shutdown_callback, ORTE_SYS_PRI);
                
            } else {
                opal_output(0, "%s is executing clean %s", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                            orted_globals.abort ? "abort" : "abnormal termination");

                /* do -not- call finalize as this will send a message to the HNP
                 * indicating clean termination! Instead, just forcibly cleanup
                 * the local session_dir tree and exit
                 */
                orte_session_dir_cleanup(ORTE_JOBID_WILDCARD);
                
                /* if we were ordered to abort, do so */
                if (orted_globals.abort) {
                    abort();
                }
                
                /* otherwise, return with non-zero status */
                ret = ORTE_ERROR_DEFAULT_EXIT_CODE;
                goto DONE;
            }
        }
    }

    /* insert our contact info into our process_info struct so we
     * have it for later use and set the local daemon field to our name
     */
    orte_process_info.my_daemon_uri = orte_rml.get_contact_info();
    ORTE_PROC_MY_DAEMON->jobid = ORTE_PROC_MY_NAME->jobid;
    ORTE_PROC_MY_DAEMON->vpid = ORTE_PROC_MY_NAME->vpid;
    
    /* if I am also the hnp, then update that contact info field too */
    if (ORTE_PROC_IS_HNP) {
        orte_process_info.my_hnp_uri = orte_rml.get_contact_info();
        ORTE_PROC_MY_HNP->jobid = ORTE_PROC_MY_NAME->jobid;
        ORTE_PROC_MY_HNP->vpid = ORTE_PROC_MY_NAME->vpid;

static void _send_notification(int status,
                               orte_proc_state_t state,
                               orte_process_name_t *proc,
                               orte_process_name_t *target)
{
    opal_buffer_t *buf;
    orte_grpcomm_signature_t sig;
    int rc;
    opal_value_t kv, *kvptr;
    orte_process_name_t daemon;

    buf = OBJ_NEW(opal_buffer_t);

    opal_output_verbose(5, orte_state_base_framework.framework_output,
                        "%s state:hnp:sending notification %s proc %s target %s",
                        ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                        ORTE_ERROR_NAME(status),
                        ORTE_NAME_PRINT(proc),
                        ORTE_NAME_PRINT(target));

    /* pack the status */
    if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &status, 1, OPAL_INT))) {
        ORTE_ERROR_LOG(rc);
        OBJ_RELEASE(buf);
        return;
    }

    /* the source is the proc */
    if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, proc, 1, ORTE_NAME))) {
        ORTE_ERROR_LOG(rc);
        OBJ_RELEASE(buf);
        return;
    }

    if (OPAL_ERR_PROC_ABORTED == status) {
        /* we will pass three opal_value_t's */
        rc = 3;
        if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &rc, 1, OPAL_INT))) {
            ORTE_ERROR_LOG(rc);
            OBJ_RELEASE(buf);
            return;
        }
        /* pass along the affected proc(s) */
        OBJ_CONSTRUCT(&kv, opal_value_t);
        kv.key = strdup(OPAL_PMIX_EVENT_AFFECTED_PROC);
        kv.type = OPAL_NAME;
        kv.data.name.jobid = proc->jobid;
        kv.data.name.vpid = proc->vpid;
        kvptr = &kv;
        if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &kvptr, 1, OPAL_VALUE))) {
            ORTE_ERROR_LOG(rc);
            OBJ_DESTRUCT(&kv);
            OBJ_RELEASE(buf);
            return;
        }
        OBJ_DESTRUCT(&kv);
    } else {
        /* we are going to pass two opal_value_t's */
        rc = 2;
        if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &rc, 1, OPAL_INT))) {
            ORTE_ERROR_LOG(rc);
            OBJ_RELEASE(buf);
            return;
        }
    }

    /* pass along the affected proc(s) */
    OBJ_CONSTRUCT(&kv, opal_value_t);
    kv.key = strdup(OPAL_PMIX_EVENT_AFFECTED_PROC);
    kv.type = OPAL_NAME;
    kv.data.name.jobid = proc->jobid;
    kv.data.name.vpid = proc->vpid;
    kvptr = &kv;
    if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &kvptr, 1, OPAL_VALUE))) {
        ORTE_ERROR_LOG(rc);
        OBJ_DESTRUCT(&kv);
        OBJ_RELEASE(buf);
        return;
    }
    OBJ_DESTRUCT(&kv);

    /* pass along the proc(s) to be notified */
    OBJ_CONSTRUCT(&kv, opal_value_t);
    kv.key = strdup(OPAL_PMIX_EVENT_CUSTOM_RANGE);
    kv.type = OPAL_NAME;
    kv.data.name.jobid = target->jobid;
    kv.data.name.vpid = target->vpid;
    kvptr = &kv;
    if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &kvptr, 1, OPAL_VALUE))) {
        ORTE_ERROR_LOG(rc);
        OBJ_DESTRUCT(&kv);
        OBJ_RELEASE(buf);
        return;
    }
    OBJ_DESTRUCT(&kv);

    /* if the targets are a wildcard, then xcast it to everyone */
    if (ORTE_VPID_WILDCARD == target->vpid) {
        OBJ_CONSTRUCT(&sig, orte_grpcomm_signature_t);
        sig.signature = (orte_process_name_t*)malloc(sizeof(orte_process_name_t));
//.........这里部分代码省略.........

static int mca_oob_ud_process_messages (struct ibv_cq *event_cq, mca_oob_ud_port_t *port)
{
    mca_oob_ud_msg_item_t *msg_item, *next_item;
    opal_list_t *processing_msgs = &mca_oob_ud_component.ud_event_processing_msgs;
    mca_oob_ud_peer_t *peer;
    mca_oob_ud_msg_hdr_t *msg_hdr;
    int msg_num, i, count;
    struct ibv_wc wc[40];
    bool peer_nacked;

    count = ibv_poll_cq (event_cq, 40, wc);
    if (count < 0)
        return count;

    /* acknowlege the events */
    ibv_ack_cq_events (event_cq, count);

    for (i = 0 ; i < count ; ++i) {
        msg_num = (int)(wc[i].wr_id & (~MCA_OOB_UD_RECV_WR));
        msg_hdr = (mca_oob_ud_msg_hdr_t *) (port->msg_buf.ptr + msg_num * port->mtu);

        VALGRIND_MAKE_MEM_DEFINED(msg_hdr, wc[i].byte_len);

        if (!(wc[i].wr_id & MCA_OOB_UD_RECV_WR) || IBV_WC_SUCCESS != wc[i].status) {
            mca_oob_ud_port_post_one_recv (port, msg_num);
            continue;
        }

        peer = mca_oob_ud_get_peer (port, &msg_hdr->ra.name, wc[i].src_qp, msg_hdr->ra.qkey,
                                    wc[i].slid, msg_hdr->ra.port_num);

        if (peer) {
            if (MCA_OOB_UD_MSG_ACK != msg_hdr->msg_type && MCA_OOB_UD_MSG_NACK != msg_hdr->msg_type &&
                MCA_OOB_UD_MSG_END != msg_hdr->msg_type) {
                mca_oob_ud_msg_item_t *msg_item = OBJ_NEW(mca_oob_ud_msg_item_t);

                msg_item->msg_num = msg_num;
                msg_item->hdr     = msg_hdr;
                msg_item->port    = port;
                msg_item->peer    = peer;

                opal_list_append (processing_msgs, (opal_list_item_t *) msg_item);
            } else {
                if (MCA_OOB_UD_MSG_ACK == msg_hdr->msg_type) {
                    (void) mca_oob_ud_event_handle_ack (port, peer, msg_hdr);
                } else if (MCA_OOB_UD_MSG_NACK == msg_hdr->msg_type) {
                    (void) mca_oob_ud_event_handle_nack (port, peer, msg_hdr);
                } else {
                    mca_oob_ud_event_handle_end (peer, msg_hdr);
                }

                mca_oob_ud_port_post_one_recv (port, msg_num);
            }
        } else {
            OPAL_OUTPUT_VERBOSE((10, mca_oob_base_output, "%s oob:ud:process_message got a null peer for message id %"
                                 PRIu64, ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), msg_hdr->msg_id));
            mca_oob_ud_port_post_one_recv (port, msg_num);
        }
    }

    /* Sort messages by peer then id */
    opal_list_sort (processing_msgs, mca_oob_ud_msg_item_cmp);

    /* Send ACKs/NACKs and throw away out-of-order messages */
    msg_item = (mca_oob_ud_msg_item_t *) mca_oob_ud_list_get_first (processing_msgs);

    for (peer = NULL, peer_nacked = false ; NULL != msg_item ; msg_item = next_item) {
        if (peer != msg_item->peer) {
            peer_nacked = false;
        }

        peer = msg_item->peer;

        next_item = (mca_oob_ud_msg_item_t *) mca_oob_ud_list_get_next (processing_msgs,
                                                                        (opal_list_item_t *)msg_item);

        if (false == peer_nacked) {
            if (msg_item->hdr->msg_id > peer->peer_expected_id) {
                (void) mca_oob_ud_event_send_nack (msg_item->port, peer, msg_item->hdr);
                peer_nacked = true;
            } else if (NULL == next_item || (next_item->peer != msg_item->peer)) {
                (void) mca_oob_ud_event_send_ack (msg_item->port, msg_item->peer, msg_item->hdr);
            }
        }

        if (msg_item->hdr->msg_id != peer->peer_expected_id) {
            opal_list_remove_item (processing_msgs, (opal_list_item_t *) msg_item);
            OBJ_RELEASE(msg_item);
        } else {
            peer->peer_expected_id++;
        }
    }

    /* Process remaining messages */
    while (NULL !=
           (msg_item = (mca_oob_ud_msg_item_t *) opal_list_remove_first (processing_msgs))) {
        switch (msg_item->hdr->msg_type) {
        case MCA_OOB_UD_MSG_REQUEST:
            mca_oob_ud_event_handle_req (port, msg_item->peer, msg_item->hdr);
            break;
//.........这里部分代码省略.........

void orte_iof_hnp_recv(int status, orte_process_name_t* sender,
                       opal_buffer_t* buffer, orte_rml_tag_t tag,
                       void* cbdata)
{
    orte_process_name_t origin;
    unsigned char data[ORTE_IOF_BASE_MSG_MAX];
    orte_iof_tag_t stream;
    int32_t count, numbytes;
    orte_iof_sink_t *sink;
    opal_list_item_t *item, *next;
    int rc;

    
    /* unpack the stream first as this may be flow control info */
    count = 1;
    if (ORTE_SUCCESS != (rc = opal_dss.unpack(buffer, &stream, &count, ORTE_IOF_TAG))) {
        ORTE_ERROR_LOG(rc);
        goto CLEAN_RETURN;
    }

    if (ORTE_IOF_XON & stream) {
        /* re-start the stdin read event */
        if (NULL != mca_iof_hnp_component.stdinev &&
            !orte_job_term_ordered &&
            !mca_iof_hnp_component.stdinev->active) {
            mca_iof_hnp_component.stdinev->active = true;
            opal_event_add(mca_iof_hnp_component.stdinev->ev, 0);
        }
        goto CLEAN_RETURN;
    } else if (ORTE_IOF_XOFF & stream) {
        /* stop the stdin read event */
        if (NULL != mca_iof_hnp_component.stdinev &&
            !mca_iof_hnp_component.stdinev->active) {
            opal_event_del(mca_iof_hnp_component.stdinev->ev);
            mca_iof_hnp_component.stdinev->active = false;
        }
        goto CLEAN_RETURN;
    }
    
    /* get name of the process whose io we are discussing */
    count = 1;
    if (ORTE_SUCCESS != (rc = opal_dss.unpack(buffer, &origin, &count, ORTE_NAME))) {
        ORTE_ERROR_LOG(rc);
        goto CLEAN_RETURN;
    }
    
    /* check to see if a tool has requested something */
    if (ORTE_IOF_PULL & stream) {
        OPAL_OUTPUT_VERBOSE((1, orte_iof_base.iof_output,
                             "%s received pull cmd from remote tool %s for proc %s",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                             ORTE_NAME_PRINT(sender),
                             ORTE_NAME_PRINT(&origin)));
        /* a tool is requesting that we send it a copy of the specified stream(s)
         * from the specified process(es), so create a sink for it
         */
        if (ORTE_IOF_STDOUT & stream) {
            ORTE_IOF_SINK_DEFINE(&sink, &origin, -1, ORTE_IOF_STDOUT,
                                 NULL, &mca_iof_hnp_component.sinks);
            sink->daemon.jobid = sender->jobid;
            sink->daemon.vpid = sender->vpid;
        }
        if (ORTE_IOF_STDERR & stream) {
            ORTE_IOF_SINK_DEFINE(&sink, &origin, -1, ORTE_IOF_STDERR,
                                 NULL, &mca_iof_hnp_component.sinks);
            sink->daemon.jobid = sender->jobid;
            sink->daemon.vpid = sender->vpid;
        }
        if (ORTE_IOF_STDDIAG & stream) {
            ORTE_IOF_SINK_DEFINE(&sink, &origin, -1, ORTE_IOF_STDDIAG,
                                 NULL, &mca_iof_hnp_component.sinks);
            sink->daemon.jobid = sender->jobid;
            sink->daemon.vpid = sender->vpid;
        }
        goto CLEAN_RETURN;
    }
    
    if (ORTE_IOF_CLOSE & stream) {
        OPAL_OUTPUT_VERBOSE((1, orte_iof_base.iof_output,
                             "%s received close cmd from remote tool %s for proc %s",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                             ORTE_NAME_PRINT(sender),
                             ORTE_NAME_PRINT(&origin)));
        /* a tool is requesting that we no longer forward a copy of the
         * specified stream(s) from the specified process(es) - remove the sink
         */
        item = opal_list_get_first(&mca_iof_hnp_component.sinks);
        while (item != opal_list_get_end(&mca_iof_hnp_component.sinks)) {
            next = opal_list_get_next(item);
            sink = (orte_iof_sink_t*)item;
            /* if the target isn't set, then this sink is for another purpose - ignore it */
            if (ORTE_JOBID_INVALID == sink->daemon.jobid) {
                continue;
            }
            /* if this sink is the designated one, then remove it from list */
            if ((stream & sink->tag) &&
                sink->name.jobid == origin.jobid &&
                (ORTE_VPID_WILDCARD == sink->name.vpid ||
                 ORTE_VPID_WILDCARD == origin.vpid ||
                 sink->name.vpid == origin.vpid)) {
//.........这里部分代码省略.........

void orte_grpcomm_base_store_modex(opal_buffer_t *rbuf, void *cbdata)
{
    int rc, j, cnt;
    int32_t num_recvd_entries;
    orte_process_name_t pname;
    orte_grpcomm_collective_t *modex = (orte_grpcomm_collective_t*)cbdata;

    OPAL_OUTPUT_VERBOSE((2, orte_grpcomm_base_framework.framework_output,
                         "%s STORING MODEX DATA",
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));

    /* unpack the process name */
    cnt=1;
    while (ORTE_SUCCESS == (rc = opal_dss.unpack(rbuf, &pname, &cnt, ORTE_NAME))) {
        /* unpack the number of entries for this proc */
        cnt=1;
        if (ORTE_SUCCESS != (rc = opal_dss.unpack(rbuf, &num_recvd_entries, &cnt, OPAL_INT32))) {
            ORTE_ERROR_LOG(rc);
            goto cleanup;
        }
        
        OPAL_OUTPUT_VERBOSE((5, orte_grpcomm_base_framework.framework_output,
                             "%s grpcomm:base:update_modex_entries: adding %d entries for proc %s",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), num_recvd_entries,
                             ORTE_NAME_PRINT(&pname)));
        
        /*
         * Extract the attribute names and values
         */
        for (j = 0; j < num_recvd_entries; j++) {
            opal_value_t *kv;
            cnt = 1;
            if (ORTE_SUCCESS != (rc = opal_dss.unpack(rbuf, &kv, &cnt, OPAL_VALUE))) {
                ORTE_ERROR_LOG(rc);
                goto cleanup;
            }
            OPAL_OUTPUT_VERBOSE((10, orte_grpcomm_base_framework.framework_output,
                                 "%s STORING MODEX DATA FROM %s FOR %s",
                                 ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                                 ORTE_NAME_PRINT(&pname), kv->key));
            /* if this is me, dump the data - we already have it in the db */
            if (ORTE_PROC_MY_NAME->jobid == pname.jobid &&
                ORTE_PROC_MY_NAME->vpid == pname.vpid) {
                OBJ_RELEASE(kv);
            } else {
                /* store it in the database */
                if (ORTE_SUCCESS != (rc = opal_db.store_pointer((opal_identifier_t*)&pname, kv))) {
                    ORTE_ERROR_LOG(rc);
                    goto cleanup;
                }
                /* do not release the kv - the db holds that pointer */
            }
        }

        OPAL_OUTPUT_VERBOSE((5, orte_grpcomm_base_framework.framework_output,
                             "%s store:peer:modex: completed modex entry for proc %s",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                             ORTE_NAME_PRINT(&pname)));
    }

 cleanup:
    if (NULL == cbdata) {
        return;
    }
    /* cleanup the list, but don't release the
     * collective object as it was passed into us
     */
    opal_list_remove_item(&orte_grpcomm_base.active_colls, &modex->super);
    /* notify that the modex is complete */
    if (NULL != modex->cbfunc) {
        OPAL_OUTPUT_VERBOSE((2, orte_grpcomm_base_framework.framework_output,
                             "%s CALLING MODEX RELEASE",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
        modex->cbfunc(NULL, modex->cbdata);
    } else {
        OPAL_OUTPUT_VERBOSE((2, orte_grpcomm_base_framework.framework_output,
                             "%s store:peer:modex NO MODEX RELEASE CBFUNC",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
    }
    /* flag the collective as complete */
    modex->active = false;
}

static int parse_cli(int argc, int start, char **argv)
{
    int i, j, k;
    bool ignore;
    char *no_dups[] = {
        "grpcomm",
        "odls",
        "rml",
        "routed",
        NULL
    };
    bool takeus = false;

    opal_output_verbose(1, orte_schizo_base_framework.framework_output,
                        "%s schizo:ompi: parse_cli",
                        ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));

    /* if they gave us a list of personalities,
     * see if we are included */
    if (NULL != orte_schizo_base.personalities) {
        for (i=0; NULL != orte_schizo_base.personalities[i]; i++) {
            if (0 == strcmp(orte_schizo_base.personalities[i], "ompi")) {
                takeus = true;
                break;
            }
        }
        if (!takeus) {
            return ORTE_ERR_TAKE_NEXT_OPTION;
        }
    } else {
        /* attempt to auto-detect CLI options that
         * we recognize */
    }

    for (i = 0; i < (argc-start); ++i) {
        if (0 == strcmp("-mca",  argv[i]) ||
            0 == strcmp("--mca", argv[i]) ) {
            /* ignore this one */
            if (0 == strcmp(argv[i+1], "mca_base_env_list")) {
                i += 2;
                continue;
            }
            /* It would be nice to avoid increasing the length
             * of the orted cmd line by removing any non-ORTE
             * params. However, this raises a problem since
             * there could be OPAL directives that we really
             * -do- want the orted to see - it's only the OMPI
             * related directives we could ignore. This becomes
             * a very complicated procedure, however, since
             * the OMPI mca params are not cleanly separated - so
             * filtering them out is nearly impossible.
             *
             * see if this is already present so we at least can
             * avoid growing the cmd line with duplicates
             */
            ignore = false;
            if (NULL != orted_cmd_line) {
                for (j=0; NULL != orted_cmd_line[j]; j++) {
                    if (0 == strcmp(argv[i+1], orted_cmd_line[j])) {
                        /* already here - if the value is the same,
                         * we can quitely ignore the fact that they
                         * provide it more than once. However, some
                         * frameworks are known to have problems if the
                         * value is different. We don't have a good way
                         * to know this, but we at least make a crude
                         * attempt here to protect ourselves.
                         */
                        if (0 == strcmp(argv[i+2], orted_cmd_line[j+1])) {
                            /* values are the same */
                            ignore = true;
                            break;
                        } else {
                            /* values are different - see if this is a problem */
                            for (k=0; NULL != no_dups[k]; k++) {
                                if (0 == strcmp(no_dups[k], argv[i+1])) {
                                    /* print help message
                                     * and abort as we cannot know which one is correct
                                     */
                                    orte_show_help("help-orterun.txt", "orterun:conflicting-params",
                                                   true, orte_basename, argv[i+1],
                                                   argv[i+2], orted_cmd_line[j+1]);
                                    return ORTE_ERR_BAD_PARAM;
                                }
                            }
                            /* this passed muster - just ignore it */
                            ignore = true;
                            break;
                        }
                    }
                }
            }
            if (!ignore) {
                opal_argv_append_nosize(&orted_cmd_line, argv[i]);
                opal_argv_append_nosize(&orted_cmd_line, argv[i+1]);
                opal_argv_append_nosize(&orted_cmd_line, argv[i+2]);
            }
            i += 2;
        }
    }
    return ORTE_SUCCESS;
//.........这里部分代码省略.........

static int setup_fork(orte_job_t *jdata,
                      orte_app_context_t *app)
{
    int i;
    char *param;
    bool oversubscribed;
    orte_node_t *node;
    char **envcpy, **nps, **firstranks;
    char *npstring, *firstrankstring;
    char *num_app_ctx;
    bool takeus = false;

    opal_output_verbose(1, orte_schizo_base_framework.framework_output,
                        "%s schizo:ompi: setup_fork",
                        ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));

    if (NULL != orte_schizo_base.personalities) {
    /* see if we are included */
        for (i=0; NULL != jdata->personality[i]; i++) {
            if (0 == strcmp(jdata->personality[i], "ompi")) {
                takeus = true;
                break;
            }
        }
        if (!takeus) {
            return ORTE_ERR_TAKE_NEXT_OPTION;
        }
    }

    /* see if the mapper thinks we are oversubscribed */
    oversubscribed = false;
    if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, ORTE_PROC_MY_NAME->vpid))) {
        ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
        return ORTE_ERR_NOT_FOUND;
    }
    if (ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_OVERSUBSCRIBED)) {
        oversubscribed = true;
    }

    /* setup base environment: copy the current environ and merge
       in the app context environ */
    if (NULL != app->env) {
        /* manually free original context->env to avoid a memory leak */
        char **tmp = app->env;
        envcpy = opal_environ_merge(orte_launch_environ, app->env);
        if (NULL != tmp) {
            opal_argv_free(tmp);
        }
    } else {
        envcpy = opal_argv_copy(orte_launch_environ);
    }
    app->env = envcpy;

    /* special case handling for --prefix: this is somewhat icky,
       but at least some users do this.  :-\ It is possible that
       when using --prefix, the user will also "-x PATH" and/or
       "-x LD_LIBRARY_PATH", which would therefore clobber the
       work that was done in the prior pls to ensure that we have
       the prefix at the beginning of the PATH and
       LD_LIBRARY_PATH.  So examine the context->env and see if we
       find PATH or LD_LIBRARY_PATH.  If found, that means the
       prior work was clobbered, and we need to re-prefix those
       variables. */
    param = NULL;
    orte_get_attribute(&app->attributes, ORTE_APP_PREFIX_DIR, (void**)&param, OPAL_STRING);
    for (i = 0; NULL != param && NULL != app->env && NULL != app->env[i]; ++i) {
        char *newenv;

        /* Reset PATH */
        if (0 == strncmp("PATH=", app->env[i], 5)) {
            asprintf(&newenv, "%s/bin:%s", param, app->env[i] + 5);
            opal_setenv("PATH", newenv, true, &app->env);
            free(newenv);
        }

        /* Reset LD_LIBRARY_PATH */
        else if (0 == strncmp("LD_LIBRARY_PATH=", app->env[i], 16)) {
            asprintf(&newenv, "%s/lib:%s", param, app->env[i] + 16);
            opal_setenv("LD_LIBRARY_PATH", newenv, true, &app->env);
            free(newenv);
        }
    }
    if (NULL != param) {
        free(param);
    }

    /* pass my contact info to the local proc so we can talk */
    opal_setenv("OMPI_MCA_orte_local_daemon_uri", orte_process_info.my_daemon_uri, true, &app->env);

    /* pass the hnp's contact info to the local proc in case it
     * needs it
     */
    if (NULL != orte_process_info.my_hnp_uri) {
        opal_setenv("OMPI_MCA_orte_hnp_uri", orte_process_info.my_hnp_uri, true, &app->env);
    }

    /* setup yield schedule - do not override any user-supplied directive! */
    if (oversubscribed) {
        opal_setenv("OMPI_MCA_mpi_yield_when_idle", "1", false, &app->env);
    } else {
//.........这里部分代码省略.........

int orte_util_add_hostfile_nodes(opal_list_t *nodes,
                                 char *hostfile)
{
    opal_list_t exclude, adds;
    opal_list_item_t *item, *itm;
    int rc;
    orte_node_t *nd, *node;
    bool found;

    OPAL_OUTPUT_VERBOSE((1, orte_ras_base_framework.framework_output,
                         "%s hostfile: checking hostfile %s for nodes",
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), hostfile));

    OBJ_CONSTRUCT(&exclude, opal_list_t);
    OBJ_CONSTRUCT(&adds, opal_list_t);

    /* parse the hostfile and add any new contents to the list */
    if (ORTE_SUCCESS != (rc = hostfile_parse(hostfile, &adds, &exclude, false))) {
        goto cleanup;
    }
    
    /* check for any relative node directives */
    for (item = opal_list_get_first(&adds);
         item != opal_list_get_end(&adds);
         item = opal_list_get_next(item)) {
        node=(orte_node_t*)item;
        
        if ('+' == node->name[0]) {
            orte_show_help("help-hostfile.txt", "hostfile:relative-syntax",
                           true, node->name);
            rc = ORTE_ERR_SILENT;
            goto cleanup;
        }
    }
    
    /* remove from the list of nodes those that are in the exclude list */
    while (NULL != (item = opal_list_remove_first(&exclude))) {
        nd = (orte_node_t*)item;
        /* check for matches on nodes */
        for (itm = opal_list_get_first(&adds);
             itm != opal_list_get_end(&adds);
             itm = opal_list_get_next(itm)) {
            node = (orte_node_t*)itm;
            if (0 == strcmp(nd->name, node->name)) {
                /* match - remove it */
                opal_list_remove_item(&adds, itm);
                OBJ_RELEASE(itm);
                break;
            }
        }
        OBJ_RELEASE(item);
    }
    
    /* transfer across all unique nodes */
    while (NULL != (item = opal_list_remove_first(&adds))) {
        nd = (orte_node_t*)item;
        found = false;
        for (itm = opal_list_get_first(nodes);
             itm != opal_list_get_end(nodes);
             itm = opal_list_get_next(itm)) {
            node = (orte_node_t*)itm;
            if (0 == strcmp(nd->name, node->name)) {
                found = true;
                break;
            }
        }
        if (!found) {
            opal_list_append(nodes, &nd->super);
            OPAL_OUTPUT_VERBOSE((1, orte_ras_base_framework.framework_output,
                                 "%s hostfile: adding node %s slots %d",
                                 ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), nd->name, nd->slots));
        } else {
            OBJ_RELEASE(item);
        }
    }

cleanup:
    OPAL_LIST_DESTRUCT(&exclude);
    OPAL_LIST_DESTRUCT(&adds);

    return rc;
}

/* Parse the provided hostfile and filter the nodes that are
 * on the input list, removing those that
 * are not found in the hostfile
 */
int orte_util_filter_hostfile_nodes(opal_list_t *nodes,
                                    char *hostfile,
                                    bool remove)
{
    opal_list_t newnodes, exclude;
    opal_list_item_t *item1, *item2, *next, *item3;
    orte_node_t *node_from_list, *node_from_file, *node_from_pool, *node3;
    int rc = ORTE_SUCCESS;
    char *cptr;
    int num_empty, nodeidx;
    bool want_all_empty = false;
    opal_list_t keep;
    bool found;

    OPAL_OUTPUT_VERBOSE((1, orte_ras_base_framework.framework_output,
                        "%s hostfile: filtering nodes through hostfile %s",
                        ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), hostfile));

    /* parse the hostfile and create local list of findings */
    OBJ_CONSTRUCT(&newnodes, opal_list_t);
    OBJ_CONSTRUCT(&exclude, opal_list_t);
    if (ORTE_SUCCESS != (rc = hostfile_parse(hostfile, &newnodes, &exclude, false))) {
        OBJ_DESTRUCT(&newnodes);
        OBJ_DESTRUCT(&exclude);
        return rc;
    }
    
    /* if the hostfile was empty, then treat it as a no-op filter */
    if (0 == opal_list_get_size(&newnodes)) {
        OBJ_DESTRUCT(&newnodes);
        OBJ_DESTRUCT(&exclude);
        /* indicate that the hostfile was empty */
        return ORTE_ERR_TAKE_NEXT_OPTION;
    }

    /* remove from the list of newnodes those that are in the exclude list
     * since we could have added duplicate names above due to the */
    while (NULL != (item1 = opal_list_remove_first(&exclude))) {
        node_from_file = (orte_node_t*)item1;
        /* check for matches on nodes */
        for (item2 = opal_list_get_first(&newnodes);
             item2 != opal_list_get_end(&newnodes);
             item2 = opal_list_get_next(item2)) {
            orte_node_t *node = (orte_node_t*)item2;
            if (0 == strcmp(node_from_file->name, node->name)) {
                /* match - remove it */
                opal_output(0, "HOST %s ON EXCLUDE LIST - REMOVING", node->name);
                opal_list_remove_item(&newnodes, item2);
                OBJ_RELEASE(item2);
                break;
            }
        }
        OBJ_RELEASE(item1);
    }
    
    /* now check our nodes and keep or mark those that match. We can
     * destruct our hostfile list as we go since this won't be needed
     */
    OBJ_CONSTRUCT(&keep, opal_list_t);
    while (NULL != (item2 = opal_list_remove_first(&newnodes))) {
        node_from_file = (orte_node_t*)item2;
        
        next = opal_list_get_next(item2);
        
        /* see if this is a relative node syntax */
        if ('+' == node_from_file->name[0]) {
            /* see if we specified empty nodes */
            if ('e' == node_from_file->name[1] ||
                'E' == node_from_file->name[1]) {
                /* request for empty nodes - do they want
                 * all of them?
                 */
                if (NULL != (cptr = strchr(node_from_file->name, ':'))) {
                    /* the colon indicates a specific # are requested */
                    cptr++; /* step past : */
                    num_empty = strtol(cptr, NULL, 10);
                } else {
                    /* want them all - set num_empty to max */
                    num_empty = INT_MAX;
                    want_all_empty = true;
                }
                /* search the list of nodes provided to us and find those
                 * that are empty
                 */
                item1 = opal_list_get_first(nodes);
                while (0 < num_empty && item1 != opal_list_get_end(nodes)) {
                    node_from_list = (orte_node_t*)item1;
                    next = opal_list_get_next(item1);  /* keep our place */
                    if (0 == node_from_list->slots_inuse) {
                        /* check to see if this node is explicitly called
                         * out later - if so, don't use it here
                         */
                        for (item3 = opal_list_get_first(&newnodes);
                             item3 != opal_list_get_end(&newnodes);
                             item3 = opal_list_get_next(item3)) {
                            node3 = (orte_node_t*)item3;
//.........这里部分代码省略.........

int orte_plm_proxy_spawn(orte_job_t *jdata)
{
    opal_buffer_t buf;
    orte_plm_cmd_flag_t command;
    orte_std_cntr_t count;
    orte_process_name_t *target;
    int rc;
    
    OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
                         "%s plm:base:proxy spawn child job",
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));

    /* setup the buffer */
    OBJ_CONSTRUCT(&buf, opal_buffer_t);
    
    /* tell the HNP we are sending a launch request */
    command = ORTE_PLM_LAUNCH_JOB_CMD;
    if (ORTE_SUCCESS != (rc = opal_dss.pack(&buf, &command, 1, ORTE_PLM_CMD))) {
        ORTE_ERROR_LOG(rc);
        goto CLEANUP;
    }
    
    /* pack the jdata object */
    if (ORTE_SUCCESS != (rc = opal_dss.pack(&buf, &jdata, 1, ORTE_JOB))) {
        ORTE_ERROR_LOG(rc);
        goto CLEANUP;
        
    }
    
    /* identify who gets this command - the HNP or the local orted */
    if (jdata->controls & ORTE_JOB_CONTROL_LOCAL_SPAWN) {
        /* for now, this is unsupported */
        opal_output(0, "LOCAL DAEMON SPAWN IS CURRENTLY UNSUPPORTED");
        target = ORTE_PROC_MY_HNP;
        /* target = ORTE_PROC_MY_DAEMON; */
    } else {
        target = ORTE_PROC_MY_HNP;
    }
    
    OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
                         "%s plm:base:proxy sending spawn cmd to %s",
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                         ORTE_NAME_PRINT(target)));
    
    /* tell the target to launch the job */
    if (0 > (rc = orte_rml.send_buffer(target, &buf, ORTE_RML_TAG_PLM, 0))) {
        ORTE_ERROR_LOG(rc);
        goto CLEANUP;
    }
    OBJ_DESTRUCT(&buf);
    
    
    OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
                         "%s plm:base:proxy waiting for response",
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
    
    /* wait for the target's response */
    OBJ_CONSTRUCT(&buf, opal_buffer_t);
    if (0 > (rc = orte_rml.recv_buffer(ORTE_NAME_WILDCARD, &buf, ORTE_RML_TAG_PLM_PROXY, 0))) {
        ORTE_ERROR_LOG(rc);
        goto CLEANUP;
    }

    /* get the new jobid back in case the caller wants it */
    count = 1;
    if (ORTE_SUCCESS != (rc = opal_dss.unpack(&buf, &(jdata->jobid), &count, ORTE_JOBID))) {
        ORTE_ERROR_LOG(rc);
        goto CLEANUP;
    }
    if (ORTE_JOBID_INVALID == jdata->jobid) {
        /* something went wrong on far end - go no further */
        rc = ORTE_ERR_FAILED_TO_START;
        goto CLEANUP;
    }
    
    /* good to go! */
    
CLEANUP:
    OBJ_DESTRUCT(&buf);
    
    return rc;
}

static int hostfile_parse_line(int token, opal_list_t* updates,
                               opal_list_t* exclude, bool keep_all)
{
    int rc;
    orte_node_t* node;
    bool got_max = false;
    char* value;
    char** argv;
    char* node_name = NULL;
    char* node_alias = NULL;
    char* username = NULL;
    int cnt;
    int number_of_slots = 0;
    char buff[64];

    if (ORTE_HOSTFILE_STRING == token ||
        ORTE_HOSTFILE_HOSTNAME == token ||
        ORTE_HOSTFILE_INT == token ||
        ORTE_HOSTFILE_IPV4 == token ||
        ORTE_HOSTFILE_IPV6 == token) {

        if(ORTE_HOSTFILE_INT == token) {
            snprintf(buff, 64, "%d", orte_util_hostfile_value.ival);
            value = buff;
        } else {
            value = orte_util_hostfile_value.sval;
        }
        argv = opal_argv_split (value, '@');
        
        cnt = opal_argv_count (argv);
        if (1 == cnt) {
            node_name = strdup(argv[0]);
        } else if (2 == cnt) {
            username = strdup(argv[0]);
            node_name = strdup(argv[1]);
        } else {
            opal_output(0, "WARNING: Unhandled [email protected]\n"); /* XXX */
        }
        opal_argv_free (argv);

        /* if the first letter of the name is '^', then this is a node
         * to be excluded. Remove the ^ character so the nodename is
         * usable, and put it on the exclude list
         */
        if ('^' == node_name[0]) {
            int i, len;
            len = strlen(node_name);
            for (i=1; i < len; i++) {
                node_name[i-1] = node_name[i];
            }
            node_name[len-1] = '\0';  /* truncate */
            
            OPAL_OUTPUT_VERBOSE((3, orte_ras_base_framework.framework_output,
                                 "%s hostfile: node %s is being excluded",
                                 ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node_name));
            
            /* convert this into something globally unique */
            if (strcmp(node_name, "localhost") == 0 || opal_ifislocal(node_name)) {
                /* Nodename has been allocated, that is for sure */
                if (orte_show_resolved_nodenames &&
                    0 != strcmp(node_name, orte_process_info.nodename)) {
                    node_alias = strdup(node_name);
                }
                free (node_name);
                node_name = strdup(orte_process_info.nodename);
            }
            
            /* Do we need to make a new node object?  First check to see
               if it's already in the exclude list */
            if (NULL == (node = hostfile_lookup(exclude, node_name, keep_all))) {
                node = OBJ_NEW(orte_node_t);
                node->name = node_name;
                if (NULL != username) {
                    node->username = strdup(username);
                }
            }
            /* Note that we need to add this back to the exclude list.
               If it was found, we just removed it (in hostfile_lookup()),
               so this puts it back. If it was not found, then we have to
               add it to the exclude list anyway. */
            opal_list_append(exclude, &node->super);
            return ORTE_SUCCESS;
        }
        
        /* this is not a node to be excluded, so we need to process it and
         * add it to the "include" list. See if this host is actually us.
         */
        if (strcmp(node_name, "localhost") == 0 || opal_ifislocal(node_name)) {
            /* Nodename has been allocated, that is for sure */
            if (orte_show_resolved_nodenames &&
                0 != strcmp(node_name, orte_process_info.nodename)) {
                node_alias = strdup(node_name);
            }
            free (node_name);
            node_name = strdup(orte_process_info.nodename);
        }

        OPAL_OUTPUT_VERBOSE((3, orte_ras_base_framework.framework_output,
                             "%s hostfile: node %s is being included - keep all is %s",
                             ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), node_name,
//.........这里部分代码省略.........

int orte_rmaps_base_set_mapping_policy(orte_mapping_policy_t *policy,
                                       char **device, char *inspec)
{
    char *ck;
    char *ptr;
    orte_mapping_policy_t tmp;
    int rc;
    size_t len;
    char *spec;
    char *pch;

    /* set defaults */
    tmp = 0;
    if (NULL != device) {
        *device = NULL;
    }

    opal_output_verbose(5, orte_rmaps_base_framework.framework_output,
                        "%s rmaps:base set policy with %s",
                        ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
                        (NULL == inspec) ? "NULL" : inspec);

    if (NULL == inspec) {
        ORTE_SET_MAPPING_POLICY(tmp, ORTE_MAPPING_BYSOCKET);
    } else {
        spec = strdup(inspec);  // protect the input string
        /* see if a colon was included - if so, then we have a policy + modifier */
        ck = strchr(spec, ':');
        if (NULL != ck) {
            /* if the colon is the first character of the string, then we
             * just have modifiers on the default mapping policy */
            if (ck == spec) {
                ck++;
                opal_output_verbose(5, orte_rmaps_base_framework.framework_output,
                                    "%s rmaps:base only modifiers %s provided - assuming bysocket mapping",
                                    ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), ck);
                ORTE_SET_MAPPING_POLICY(tmp, ORTE_MAPPING_BYSOCKET);
                if (ORTE_ERR_SILENT == (rc = check_modifiers(ck, &tmp)) &&
                    ORTE_ERR_BAD_PARAM != rc) {
                    free(spec);
                    return ORTE_ERR_SILENT;
                }
                free(spec);
                goto setpolicy;
            }
            /* split the string */
            *ck = '\0';
            ck++;
            opal_output_verbose(5, orte_rmaps_base_framework.framework_output,
                                "%s rmaps:base policy %s modifiers %s provided",
                                ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), spec, ck);
            /* if the policy is "dist", then we set the policy to that value
             * and save the second argument as the device
             */
            if (0 == strncasecmp(spec, "ppr", strlen(spec))) {
                /* we have to allow additional modifiers here - e.g., specifying
                 * #pe's/proc or oversubscribe - so check for modifiers
                 */
                if (NULL == (ptr = strrchr(ck, ':'))) {
                    /* this is an error - there had to be at least one
                     * colon to delimit the number from the object type
                     */
                    orte_show_help("help-orte-rmaps-base.txt", "invalid-pattern", true, inspec);
                    free(spec);
                    return ORTE_ERR_SILENT;
                }
                ptr++; // move past the colon
                /* check the remaining string for modifiers - may be none, so
                 * don't emit an error message if the modifier isn't recognized
                 */
                if (ORTE_ERR_SILENT == (rc = check_modifiers(ptr, &tmp)) &&
                    ORTE_ERR_BAD_PARAM != rc) {
                    free(spec);
                    return ORTE_ERR_SILENT;
                }
                /* if we found something, then we need to adjust the string */
                if (ORTE_SUCCESS == rc) {
                    ptr--;
                    *ptr = '\0';
                }
                /* now get the pattern */
                orte_rmaps_base.ppr = strdup(ck);
                ORTE_SET_MAPPING_POLICY(tmp, ORTE_MAPPING_PPR);
                ORTE_SET_MAPPING_DIRECTIVE(tmp, ORTE_MAPPING_GIVEN);
                free(spec);
                goto setpolicy;
            }
            if (ORTE_SUCCESS != (rc = check_modifiers(ck, &tmp)) &&
                ORTE_ERR_TAKE_NEXT_OPTION != rc) {
                if (ORTE_ERR_BAD_PARAM == rc) {
                    orte_show_help("help-orte-rmaps-base.txt", "unrecognized-modifier", true, inspec);
                }
                free(spec);
                return rc;
            }
        }
        len = strlen(spec);
        if (0 == strncasecmp(spec, "slot", len)) {
            ORTE_SET_MAPPING_POLICY(tmp, ORTE_MAPPING_BYSLOT);
        } else if (0 == strncasecmp(spec, "node", len)) {
//.........这里部分代码省略.........

//.........这里部分代码省略.........
                    orte_show_help("help-orte-rmaps-base.txt", "mismatch-binding", true,
                                   orte_rmaps_base.cpus_per_rank, "use-hwthreads-as-cpus",
                                   opal_hwloc_base_print_binding(opal_hwloc_binding_policy),
                                   "bind-to hwthread");
                    return ORTE_ERR_SILENT;
                }
            } else if (OPAL_BIND_TO_CORE != OPAL_GET_BINDING_POLICY(opal_hwloc_binding_policy) &&
                       OPAL_BIND_TO_NONE != OPAL_GET_BINDING_POLICY(opal_hwloc_binding_policy)) {
                orte_show_help("help-orte-rmaps-base.txt", "mismatch-binding", true,
                               orte_rmaps_base.cpus_per_rank, "cores as cpus",
                               opal_hwloc_base_print_binding(opal_hwloc_binding_policy),
                               "bind-to core");
                return ORTE_ERR_SILENT;
            }
        } else {
            if (opal_hwloc_use_hwthreads_as_cpus) {
                OPAL_SET_BINDING_POLICY(opal_hwloc_binding_policy, OPAL_BIND_TO_HWTHREAD);
            } else {
                OPAL_SET_BINDING_POLICY(opal_hwloc_binding_policy, OPAL_BIND_TO_CORE);
            }
        }
        /* we also need to ensure we are mapping to a high-enough level to have
         * multiple cpus beneath it - by default, we'll go to the NUMA level */
        if (ORTE_MAPPING_GIVEN & ORTE_GET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping)) {
            if (ORTE_GET_MAPPING_POLICY(orte_rmaps_base.mapping) == ORTE_MAPPING_BYHWTHREAD ||
              (ORTE_GET_MAPPING_POLICY(orte_rmaps_base.mapping) == ORTE_MAPPING_BYCORE &&
              !opal_hwloc_use_hwthreads_as_cpus)) {
                orte_show_help("help-orte-rmaps-base.txt", "mapping-too-low-init", true);
                return ORTE_ERR_SILENT;
            }
        } else {
            opal_output_verbose(5, orte_rmaps_base_framework.framework_output,
                                "%s rmaps:base pe/rank set - setting mapping to BYNUMA",
                                ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
            ORTE_SET_MAPPING_POLICY(orte_rmaps_base.mapping, ORTE_MAPPING_BYNUMA);
            ORTE_SET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping, ORTE_MAPPING_GIVEN);
        }
    }

    if (orte_rmaps_base_pernode) {
        /* there is no way to resolve this conflict, so if something else was
         * given, we have no choice but to error out
         */
        if (ORTE_MAPPING_GIVEN & ORTE_GET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping)) {
            orte_show_help("help-orte-rmaps-base.txt", "redefining-policy", true, "mapping",
                           "bynode", orte_rmaps_base_print_mapping(orte_rmaps_base.mapping));
            return ORTE_ERR_SILENT;
        }
        /* ensure we set the mapping policy to ppr */
        ORTE_SET_MAPPING_POLICY(orte_rmaps_base.mapping, ORTE_MAPPING_PPR);
        ORTE_SET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping, ORTE_MAPPING_GIVEN);
        /* define the ppr */
        orte_rmaps_base.ppr = strdup("1:node");
    }

    if (0 < orte_rmaps_base_n_pernode) {
        /* there is no way to resolve this conflict, so if something else was
         * given, we have no choice but to error out
         */
        if (ORTE_MAPPING_GIVEN & ORTE_GET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping)) {
            orte_show_help("help-orte-rmaps-base.txt", "redefining-policy", true, "mapping",
                           "bynode", orte_rmaps_base_print_mapping(orte_rmaps_base.mapping));
            return ORTE_ERR_SILENT;
        }
        /* ensure we set the mapping policy to ppr */
        ORTE_SET_MAPPING_POLICY(orte_rmaps_base.mapping, ORTE_MAPPING_PPR);

int orte_ess_base_orted_setup(char **hosts)
{
    int ret = ORTE_ERROR;
    int fd;
    char log_file[PATH_MAX];
    char *jobidstring;
    char *error = NULL;
    char *plm_to_use;
    orte_job_t *jdata;
    orte_proc_t *proc;
    orte_app_context_t *app;
    orte_node_t *node;

#ifndef __WINDOWS__
    /* setup callback for SIGPIPE */
    setup_sighandler(SIGPIPE, &epipe_handler, epipe_signal_callback);
    /* Set signal handlers to catch kill signals so we can properly clean up
     * after ourselves. 
     */
    setup_sighandler(SIGTERM, &term_handler, shutdown_signal);
    setup_sighandler(SIGINT, &int_handler, shutdown_signal);
    
    /** setup callbacks for signals we should ignore */
    setup_sighandler(SIGUSR1, &sigusr1_handler, signal_callback);
    setup_sighandler(SIGUSR2, &sigusr2_handler, signal_callback);
#endif  /* __WINDOWS__ */
    
    signals_set = true;
    
#if OPAL_HAVE_HWLOC
    {
        hwloc_obj_t obj;
        unsigned i, j;
        
        /* get the local topology */
        if (NULL == opal_hwloc_topology) {
            if (OPAL_SUCCESS != opal_hwloc_base_get_topology()) {
                error = "topology discovery";
                goto error;
            }
        }
        
        /* remove the hostname from the topology. Unfortunately, hwloc
         * decided to add the source hostname to the "topology", thus
         * rendering it unusable as a pure topological description. So
         * we remove that information here.
         */
        obj = hwloc_get_root_obj(opal_hwloc_topology);
        for (i=0; i < obj->infos_count; i++) {
            if (NULL == obj->infos[i].name ||
                NULL == obj->infos[i].value) {
                continue;
            }
            if (0 == strncmp(obj->infos[i].name, "HostName", strlen("HostName"))) {
                free(obj->infos[i].name);
                free(obj->infos[i].value);
                /* left justify the array */
                for (j=i; j < obj->infos_count-1; j++) {
                    obj->infos[j] = obj->infos[j+1];
                }
                obj->infos[obj->infos_count-1].name = NULL;
                obj->infos[obj->infos_count-1].value = NULL;
                obj->infos_count--;
                break;
            }
        }
        
        if (4 < opal_output_get_verbosity(orte_ess_base_output)) {
            opal_output(0, "%s Topology Info:", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
            opal_dss.dump(0, opal_hwloc_topology, OPAL_HWLOC_TOPO);
        }
    }
#endif
    
    /* open and setup the opal_pstat framework so we can provide
     * process stats if requested
     */
    if (ORTE_SUCCESS != (ret = opal_pstat_base_open())) {
        ORTE_ERROR_LOG(ret);
        error = "opal_pstat_base_open";
        goto error;
    }
    if (ORTE_SUCCESS != (ret = opal_pstat_base_select())) {
        ORTE_ERROR_LOG(ret);
        error = "opal_pstat_base_select";
        goto error;
    }
    
    /* open and setup the state machine */
    if (ORTE_SUCCESS != (ret = orte_state_base_open())) {
        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;
    }
    
//.........这里部分代码省略.........

int orte_util_get_ordered_host_list(opal_list_t *nodes,
                                    char *hostfile)
{
    opal_list_t exclude;
    opal_list_item_t *item, *itm, *item2, *item1;
    char *cptr;
    int num_empty, i, nodeidx, startempty=0;
    bool want_all_empty=false;
    orte_node_t *node_from_pool, *newnode;
    int rc;
    
    OPAL_OUTPUT_VERBOSE((1, orte_ras_base_framework.framework_output,
                         "%s hostfile: creating ordered list of hosts from hostfile %s",
                         ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), hostfile));
    
    OBJ_CONSTRUCT(&exclude, opal_list_t);
    
    /* parse the hostfile and add the contents to the list, keeping duplicates */
    if (ORTE_SUCCESS != (rc = hostfile_parse(hostfile, nodes, &exclude, true))) {
        goto cleanup;
    }
    
    /* parse the nodes to process any relative node directives */
    item2 = opal_list_get_first(nodes);
    while (item2 != opal_list_get_end(nodes)) {
        orte_node_t *node=(orte_node_t*)item2;
        
        /* save the next location in case this one gets removed */
        item1 = opal_list_get_next(item2);
        
        if ('+' != node->name[0]) {
            item2 = item1;
            continue;
        }
        
        /* see if we specified empty nodes */
        if ('e' == node->name[1] ||
            'E' == node->name[1]) {
            /* request for empty nodes - do they want
             * all of them?
             */
            if (NULL != (cptr = strchr(node->name, ':'))) {
                /* the colon indicates a specific # are requested */
                cptr++; /* step past : */
                num_empty = strtol(cptr, NULL, 10);
            } else {
                /* want them all - set num_empty to max */
                num_empty = INT_MAX;
                want_all_empty = true;
            }
            /* insert empty nodes into newnodes list in place of the current item.
             * since item1 is the next item, we insert in front of it
             */
            if (!orte_hnp_is_allocated && 0 == startempty) {
               startempty = 1;
            }
            for (i=startempty; 0 < num_empty && i < orte_node_pool->size; i++) {
                if (NULL == (node_from_pool = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, i))) {
                    continue;
                }
                if (0 == node_from_pool->slots_inuse) {
                    newnode = OBJ_NEW(orte_node_t);
                    newnode->name = strdup(node_from_pool->name);
                    /* if the slot count here is less than the
                     * total slots avail on this node, set it
                     * to the specified count - this allows people
                     * to subdivide an allocation
                     */
                    if (node->slots < node_from_pool->slots) {
                        newnode->slots = node->slots;
                    } else {
                        newnode->slots = node_from_pool->slots;
                    }
                    opal_list_insert_pos(nodes, item1, &newnode->super);
                    /* track number added */
                    --num_empty;
                }
            }
            /* bookmark where we stopped in case they ask for more */
            startempty = i;
            /* did they get everything they wanted? */
            if (!want_all_empty && 0 < num_empty) {
                orte_show_help("help-hostfile.txt", "hostfile:not-enough-empty",
                               true, num_empty);
                rc = ORTE_ERR_SILENT;
                goto cleanup;
            }            
            /* since we have expanded the provided node, remove
             * it from list
             */
            opal_list_remove_item(nodes, item2);
            OBJ_RELEASE(item2);
        } else if ('n' == node->name[1] ||
                   'N' == node->name[1]) {
            /* they want a specific relative node #, so
             * look it up on global pool
             */
            nodeidx = strtol(&node->name[2], NULL, 10);
            /* if the HNP is not allocated, then we need to
             * adjust the index as the node pool is offset
//.........这里部分代码省略.........