ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность

Update how we pass the node regex so we pass _all_ nodes, even those without daemons. This allows the backend daemons to form a complete picture of the allocation. Include info on which nodes have daemons on them, and populate that info on the backend as well.

Просмотр исходного кода 
Set the daemons' state to "running" and mark them as "alive" by default when constructing the nidmap

Get the DVM running again

Fix direct modex by eliminating race condition caused by releasing data while sending it

Up the size limit before compressing

Signed-off-by: Ralph Castain <rhc@open-mpi.org>
Этот коммит содержится в:
Ralph Castain 2017-03-29 18:28:25 -07:00
родитель 9850832dbd
Коммит 92c996487c
27 изменённых файлов: 643 добавлений и 1025 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

10
opal/mca/pmix/pmix2x/pmix/src/server/pmix_server.c
Просмотреть файл

@ -990,11 +990,11 @@ static void _dmodex_req(int sd, short args, void *cbdata)
* may not be a contribution */
if (PMIX_SUCCESS == (rc = pmix_hash_fetch(&nptr->server->myremote, info->rank, "modex", &val)) &&
NULL != val) {
data = val->data.bo.bytes;
sz = val->data.bo.size;
/* protect the data */
val->data.bo.bytes = NULL;
val->data.bo.size = 0;
data = val->data.bo.bytes;
sz = val->data.bo.size;
/* protect the data */
val->data.bo.bytes = NULL;
val->data.bo.size = 0;
PMIX_VALUE_RELEASE(val);
}

12
opal/mca/pmix/pmix2x/pmix/src/server/pmix_server_ops.c
Просмотреть файл

@ -138,7 +138,6 @@ pmix_status_t pmix_server_commit(pmix_peer_t *peer, pmix_buffer_t *buf)
pmix_nspace_t *nptr;
pmix_rank_info_t *info;
pmix_dmdx_remote_t *dcd, *dcdnext;
pmix_buffer_t *pbkt;
pmix_value_t *val;
char *data;
size_t sz;
@ -236,16 +235,19 @@ pmix_status_t pmix_server_commit(pmix_peer_t *peer, pmix_buffer_t *buf)
if (dcd->cd->proc.rank == info->rank) {
/* we can now fulfill this request - collect the
* remote/global data from this proc */
pbkt = PMIX_NEW(pmix_buffer_t);
/* get any remote contribution - note that there
* may not be a contribution */
data = NULL;
sz = 0;
if (PMIX_SUCCESS == pmix_hash_fetch(&nptr->server->myremote, info->rank, "modex", &val) &&
NULL != val) {
PMIX_LOAD_BUFFER(pbkt, val->data.bo.bytes, val->data.bo.size);
data = val->data.bo.bytes;
sz = val->data.bo.size;
/* protect the data */
val->data.bo.bytes = NULL;
val->data.bo.size = 0;
PMIX_VALUE_RELEASE(val);
}
PMIX_UNLOAD_BUFFER(pbkt, data, sz);
PMIX_RELEASE(pbkt);
/* execute the callback */
dcd->cd->cbfunc(PMIX_SUCCESS, data, sz, dcd->cd->cbdata);
if (NULL != data) {

6
opal/mca/pmix/pmix2x/pmix/src/util/compress.h
Просмотреть файл

@ -9,7 +9,7 @@
* University of Stuttgart. All rights reserved.
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2015-2016 Intel, Inc. All rights reserved.
* Copyright (c) 2015-2017 Intel, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
@ -31,8 +31,8 @@
BEGIN_C_DECLS
/* define a limit for storing raw strings */
#define PMIX_STRING_LIMIT 512
/* define a limit of 128k for raw strings */
#define PMIX_STRING_LIMIT 131072
/* define a macro for quickly checking if a string exceeds the
* compression limit */

16
orte/mca/ess/alps/ess_alps_module.c
Просмотреть файл

@ -12,6 +12,7 @@
* Copyright (c) 2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2011-2013 Los Alamos National Security, LLC.
* All rights reserved.
* Copyright (c) 2017 Intel, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
@ -58,7 +59,6 @@ static int rte_init(void)
{
int ret;
char *error = NULL;
char **hosts = NULL;
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_framework.framework_output,
"ess:alps in rte_init"));
@ -90,23 +90,11 @@ static int rte_init(void)
* default procedure
*/
if (ORTE_PROC_IS_DAEMON) {
if (NULL != orte_node_regex) {
/* extract the nodes */
if (ORTE_SUCCESS != (ret =
orte_regex_extract_node_names(orte_node_regex, &hosts)) ||
NULL == hosts) {
error = "orte_regex_extract_node_names";
goto fn_fail;
}
}
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup(hosts))) {
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_orted_setup";
goto fn_fail;
}
if (NULL != hosts) {
opal_argv_free(hosts);
}
/*
* now synchronize with aprun.

4
orte/mca/ess/base/base.h
Просмотреть файл

@ -12,7 +12,7 @@
* Copyright (c) 2011-2012 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2012 Oak Ridge National Labs. All rights reserved.
* Copyright (c) 2013 Los Alamos National Security, LLC. All rights reserved.
* Copyright (c) 2013 Intel, Inc. All rights reserved.
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
@ -67,7 +67,7 @@ ORTE_DECLSPEC void orte_ess_base_app_abort(int status, bool report);
ORTE_DECLSPEC int orte_ess_base_tool_setup(void);
ORTE_DECLSPEC int orte_ess_base_tool_finalize(void);
ORTE_DECLSPEC int orte_ess_base_orted_setup(char **hosts);
ORTE_DECLSPEC int orte_ess_base_orted_setup(void);
ORTE_DECLSPEC int orte_ess_base_orted_finalize(void);
/* Detect whether or not this proc is bound - if not,

43
orte/mca/ess/base/ess_base_std_orted.c
Просмотреть файл

@ -103,7 +103,7 @@ static void setup_sighandler(int signal, opal_event_t *ev,
}
int orte_ess_base_orted_setup(char **hosts)
int orte_ess_base_orted_setup(void)
{
int ret = ORTE_ERROR;
int fd;
@ -113,7 +113,6 @@ int orte_ess_base_orted_setup(char **hosts)
orte_job_t *jdata;
orte_proc_t *proc;
orte_app_context_t *app;
orte_node_t *node;
char *param;
hwloc_obj_t obj;
unsigned i, j;
@ -218,12 +217,9 @@ int orte_ess_base_orted_setup(char **hosts)
* a specific module to use
*/
(void) mca_base_var_env_name("plm", &param);
plm_in_use = !!(getenv(param));
free (param);
if (plm_in_use) {
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_plm_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_plm_base_open";
@ -332,11 +328,6 @@ int orte_ess_base_orted_setup(char **hosts)
app = OBJ_NEW(orte_app_context_t);
opal_pointer_array_set_item(jdata->apps, 0, app);
jdata->num_apps++;
/* create and store a node object where we are */
node = OBJ_NEW(orte_node_t);
node->name = strdup(orte_process_info.nodename);
node->index = ORTE_PROC_MY_NAME->vpid;
opal_pointer_array_set_item(orte_node_pool, ORTE_PROC_MY_NAME->vpid, node);
/* create and store a proc object for us */
proc = OBJ_NEW(orte_proc_t);
@ -345,19 +336,6 @@ int orte_ess_base_orted_setup(char **hosts)
proc->pid = orte_process_info.pid;
proc->state = ORTE_PROC_STATE_RUNNING;
opal_pointer_array_set_item(jdata->procs, proc->name.vpid, proc);
/* record that the daemon (i.e., us) is on this node
* NOTE: we do not add the proc object to the node's
* proc array because we are not an application proc.
* Instead, we record it in the daemon field of the
* node object
*/
OBJ_RETAIN(proc); /* keep accounting straight */
node->daemon = proc;
ORTE_FLAG_SET(node, ORTE_NODE_FLAG_DAEMON_LAUNCHED);
node->state = ORTE_NODE_STATE_UP;
/* now point our proc node field to the node */
OBJ_RETAIN(node); /* keep accounting straight */
proc->node = node;
/* record that the daemon job is running */
jdata->num_procs = 1;
jdata->state = ORTE_JOB_STATE_RUNNING;
@ -514,7 +492,6 @@ int orte_ess_base_orted_setup(char **hosts)
orte_topo_signature = opal_hwloc_base_get_topo_signature(opal_hwloc_topology);
t->sig = strdup(orte_topo_signature);
opal_pointer_array_add(orte_node_topologies, t);
node->topology = t;
if (15 < opal_output_get_verbosity(orte_ess_base_framework.framework_output)) {
opal_output(0, "%s Topology Info:", ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
opal_dss.dump(0, opal_hwloc_topology, OPAL_HWLOC_TOPO);
@ -526,12 +503,25 @@ int orte_ess_base_orted_setup(char **hosts)
* after we enable_comm as that function determines our
* own port, which we need in order to construct the nidmap
*/
if (NULL != hosts) {
if (NULL != orte_node_regex) {
if (ORTE_SUCCESS != (ret = orte_util_nidmap_parse(orte_node_regex))) {
ORTE_ERROR_LOG(ret);
error = "construct nidmap";
goto error;
}
}
if (orte_static_ports) {
if (NULL == orte_node_regex) {
/* we didn't get the node info */
error = "cannot construct daemon map for static ports - no node map info";
goto error;
}
/* extract the node info from the environment and
* build a nidmap from it - this will update the
* routing plan as well
*/
if (ORTE_SUCCESS != (ret = orte_util_build_daemon_nidmap(hosts))) {
if (ORTE_SUCCESS != (ret = orte_util_build_daemon_nidmap())) {
ORTE_ERROR_LOG(ret);
error = "construct daemon map from static ports";
goto error;
@ -635,6 +625,7 @@ int orte_ess_base_orted_setup(char **hosts)
}
return ORTE_SUCCESS;
error:
orte_show_help("help-orte-runtime.txt",
"orte_init:startup:internal-failure",

13
orte/mca/ess/env/ess_env_module.c поставляемый
Просмотреть файл

@ -10,7 +10,7 @@
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2011-2012 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2013-2015 Intel, Inc. All rights reserved.
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
@ -98,7 +98,6 @@ static int rte_init(void)
{
int ret;
char *error = NULL;
char **hosts = NULL;
/* run the prolog */
if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
@ -112,19 +111,11 @@ static int rte_init(void)
/* if I am a daemon, complete my setup using the
* default procedure
*/
if (NULL != orte_node_regex) {
/* extract the nodes */
if (ORTE_SUCCESS != (ret = orte_regex_extract_node_names(orte_node_regex, &hosts))) {
error = "orte_regex_extract_node_names";
goto error;
}
}
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup(hosts))) {
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_orted_setup";
goto error;
}
opal_argv_free(hosts);
return ORTE_SUCCESS;
error:

13
orte/mca/ess/lsf/ess_lsf_module.c
Просмотреть файл

@ -10,7 +10,7 @@
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2007-2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2013 Intel, Inc. All rights reserved.
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved.
* Copyright (c) 2016 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* $COPYRIGHT$
@ -68,7 +68,6 @@ static int rte_init(void)
{
int ret;
char *error = NULL;
char **hosts = NULL;
/* run the prolog */
if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
@ -83,19 +82,11 @@ static int rte_init(void)
* default procedure
*/
if (ORTE_PROC_IS_DAEMON) {
if (NULL != orte_node_regex) {
/* extract the nodes */
if (ORTE_SUCCESS != (ret = orte_regex_extract_node_names(orte_node_regex, &hosts))) {
error = "orte_regex_extract_node_names";
goto error;
}
}
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup(hosts))) {
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_orted_setup";
goto error;
}
opal_argv_free(hosts);
return ORTE_SUCCESS;
}

17
orte/mca/ess/slurm/ess_slurm_module.c
Просмотреть файл

@ -10,7 +10,7 @@
* Copyright (c) 2004-2005 The Regents of the University of California.
* All rights reserved.
* Copyright (c) 2008-2011 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2013 Intel, Inc. All rights reserved.
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
@ -62,7 +62,6 @@ static int rte_init(void)
{
int ret;
char *error = NULL;
char **hosts = NULL;
/* run the prolog */
if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
@ -77,23 +76,11 @@ static int rte_init(void)
* default procedure
*/
if (ORTE_PROC_IS_DAEMON) {
if (NULL != orte_node_regex) {
/* extract the nodes */
if (ORTE_SUCCESS != (ret =
orte_regex_extract_node_names(orte_node_regex, &hosts)) ||
NULL == hosts) {
error = "orte_regex_extract_node_names";
goto error;
}
}
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup(hosts))) {
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_orted_setup";
goto error;
}
if (NULL != hosts) {
opal_argv_free(hosts);
}
return ORTE_SUCCESS;
}

14
orte/mca/ess/tm/ess_tm_module.c
Просмотреть файл

@ -67,7 +67,6 @@ static int rte_init(void)
{
int ret;
char *error = NULL;
char **hosts = NULL;
/* run the prolog */
if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
@ -82,21 +81,11 @@ static int rte_init(void)
* default procedure
*/
if (ORTE_PROC_IS_DAEMON) {
if (NULL != orte_node_regex) {
/* extract the nodes */
if (ORTE_SUCCESS != (ret =
orte_regex_extract_node_names(orte_node_regex, &hosts)) ||
NULL == hosts) {
error = "orte_regex_extract_node_names";
goto error;
}
}
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup(hosts))) {
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_orted_setup";
goto error;
}
opal_argv_free(hosts);
return ORTE_SUCCESS;
}
@ -194,4 +183,3 @@ static int tm_set_name(void)
return ORTE_SUCCESS;
}

22
orte/mca/grpcomm/direct/grpcomm_direct.c
Просмотреть файл

@ -270,7 +270,7 @@ static void xcast_recv(int status, orte_process_name_t* sender,
opal_list_t coll;
orte_grpcomm_signature_t *sig;
orte_rml_tag_t tag;
char *rtmod;
char *rtmod, *nidmap;
size_t inlen, cmplen;
uint8_t *packed_data, *cmpdata;
@ -392,7 +392,8 @@ static void xcast_recv(int status, orte_process_name_t* sender,
}
opal_dss.copy_payload(relay, data);
} else if (ORTE_DAEMON_ADD_LOCAL_PROCS == command ||
ORTE_DAEMON_DVM_NIDMAP_CMD == command) {
ORTE_DAEMON_DVM_NIDMAP_CMD == command ||
ORTE_DAEMON_DVM_ADD_PROCS == command) {
/* setup our internal relay buffer */
relay = OBJ_NEW(opal_buffer_t);
/* repack the command */
@ -400,14 +401,25 @@ static void xcast_recv(int status, orte_process_name_t* sender,
ORTE_ERROR_LOG(ret);
goto relay;
}
/* see if any daemons were launched */
/* unpack the nidmap string - may be NULL */
cnt = 1;
if (OPAL_SUCCESS != (ret = opal_dss.unpack(data, &nidmap, &cnt, OPAL_STRING))) {
ORTE_ERROR_LOG(ret);
goto relay;
}
if (NULL != nidmap) {
if (ORTE_SUCCESS != (ret = orte_util_nidmap_parse(nidmap))) {
ORTE_ERROR_LOG(ret);
goto relay;
}
free(nidmap);
}
/* see if they included info on node capabilities */
cnt = 1;
if (OPAL_SUCCESS != (ret = opal_dss.unpack(data, &flag, &cnt, OPAL_INT8))) {
ORTE_ERROR_LOG(ret);
goto relay;
}
/* add it to our relay buffer as we will need it later */
opal_dss.pack(relay, &flag, 1, OPAL_INT8);
if (0 != flag) {
/* update our local nidmap, if required - the decode function
* knows what to do

93
orte/mca/odls/base/odls_base_default_fns.c
Просмотреть файл

@ -113,6 +113,7 @@ int orte_odls_base_default_get_add_procs_data(opal_buffer_t *buffer,
int8_t flag;
void *nptr;
uint32_t key;
char *nidmap;
/* get the job data pointer */
if (NULL == (jdata = orte_get_job_data_object(job))) {
@ -127,19 +128,32 @@ int orte_odls_base_default_get_add_procs_data(opal_buffer_t *buffer,
return ORTE_SUCCESS;
}
/* if we launched new daemons... */
if (orte_get_attribute(&jdata->attributes, ORTE_JOB_LAUNCHED_DAEMONS, NULL, OPAL_BOOL)) {
/* flag that we did */
/* if we couldn't provide the allocation regex on the orted
* cmd line, then we need to provide all the info here */
if (!orte_nidmap_communicated) {
if (ORTE_SUCCESS != (rc = orte_util_nidmap_create(&nidmap))) {
ORTE_ERROR_LOG(rc);
return rc;
}
orte_nidmap_communicated = true;
} else {
nidmap = NULL;
}
opal_dss.pack(buffer, &nidmap, 1, OPAL_STRING);
if (NULL != nidmap) {
free(nidmap);
}
/* if we haven't already done so, provide the info on the
* capabilities of each node */
if (!orte_node_info_communicated ||
orte_get_attribute(&jdata->attributes, ORTE_JOB_LAUNCHED_DAEMONS, NULL, OPAL_BOOL)) {
flag = 1;
opal_dss.pack(buffer, &flag, 1, OPAL_INT8);
/* include a nodemap of the daemons */
if (ORTE_SUCCESS != (rc = orte_util_encode_nodemap(buffer))) {
ORTE_ERROR_LOG(rc);
return rc;
}
/* if we are not using static ports, we need to send the wireup info */
if (!orte_static_ports && !orte_fwd_mpirun_port) {
/* pack a flag indicating wiring info is provided */
flag = 1;
@ -176,41 +190,52 @@ int orte_odls_base_default_get_add_procs_data(opal_buffer_t *buffer,
* copy of all active jobs so the grpcomm collectives can
* properly work should a proc from one of the other jobs
* interact with this one */
OBJ_CONSTRUCT(&jobdata, opal_buffer_t);
numjobs = 0;
rc = opal_hash_table_get_first_key_uint32(orte_job_data, &key, (void **)&jptr, &nptr);
while (OPAL_SUCCESS == rc) {
/* skip the one we are launching now */
if (NULL != jptr && jptr != jdata &&
ORTE_PROC_MY_NAME->jobid != jptr->jobid) {
/* pack the job struct */
if (ORTE_SUCCESS != (rc = opal_dss.pack(&jobdata, &jptr, 1, ORTE_JOB))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&jobdata);
return rc;
if (orte_get_attribute(&jdata->attributes, ORTE_JOB_LAUNCHED_DAEMONS, NULL, OPAL_BOOL)) {
flag = 1;
opal_dss.pack(buffer, &flag, 1, OPAL_INT8);
OBJ_CONSTRUCT(&jobdata, opal_buffer_t);
numjobs = 0;
rc = opal_hash_table_get_first_key_uint32(orte_job_data, &key, (void **)&jptr, &nptr);
while (OPAL_SUCCESS == rc) {
/* skip the one we are launching now */
if (NULL != jptr && jptr != jdata &&
ORTE_PROC_MY_NAME->jobid != jptr->jobid) {
/* pack the job struct */
if (ORTE_SUCCESS != (rc = opal_dss.pack(&jobdata, &jptr, 1, ORTE_JOB))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&jobdata);
return rc;
}
++numjobs;
}
++numjobs;
rc = opal_hash_table_get_next_key_uint32(orte_job_data, &key, (void **)&jptr, nptr, &nptr);
}
rc = opal_hash_table_get_next_key_uint32(orte_job_data, &key, (void **)&jptr, nptr, &nptr);
}
/* pack the number of jobs */
if (ORTE_SUCCESS != (rc = opal_dss.pack(buffer, &numjobs, 1, OPAL_INT32))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&jobdata);
return rc;
}
if (0 < numjobs) {
/* pack the jobdata buffer */
wireup = &jobdata;
if (ORTE_SUCCESS != (rc = opal_dss.pack(buffer, &wireup, 1, OPAL_BUFFER))) {
/* pack the number of jobs */
if (ORTE_SUCCESS != (rc = opal_dss.pack(buffer, &numjobs, 1, OPAL_INT32))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&jobdata);
return rc;
}
OBJ_DESTRUCT(&jobdata);
if (0 < numjobs) {
/* pack the jobdata buffer */
wireup = &jobdata;
if (ORTE_SUCCESS != (rc = opal_dss.pack(buffer, &wireup, 1, OPAL_BUFFER))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&jobdata);
return rc;
}
OBJ_DESTRUCT(&jobdata);
}
} else {
flag = 0;
opal_dss.pack(buffer, &flag, 1, OPAL_INT8);
}
orte_node_info_communicated = true;
} else {
/* include a sentinel */
/* mark that we didn't */
flag = 0;
opal_dss.pack(buffer, &flag, 1, OPAL_INT8);
/* and that we didn't launch daemons */
flag = 0;
opal_dss.pack(buffer, &flag, 1, OPAL_INT8);
}

71
orte/mca/plm/alps/plm_alps_module.c
Просмотреть файл

@ -306,41 +306,42 @@ static void launch_daemons(int fd, short args, void *cbdata)
opal_argv_append(&argc, &argv, "-e");
opal_argv_append(&argc, &argv, "OMPI_NO_USE_CRAY_PMI=1");
/* create nodelist */
nodelist_argv = NULL;
nodelist_argc = 0;
for (nnode=0; nnode < map->nodes->size; nnode++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, nnode))) {
continue;
}
/* if the daemon already exists on this node, then
* don't include it
*/
if (ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_DAEMON_LAUNCHED)) {
continue;
}
/* otherwise, add it to the list of nodes upon which
* we need to launch a daemon
*/
opal_argv_append(&nodelist_argc, &nodelist_argv, node->name);
}
if (0 == opal_argv_count(nodelist_argv)) {
orte_show_help("help-plm-alps.txt", "no-hosts-in-list", true);
rc = ORTE_ERR_FAILED_TO_START;
goto cleanup;
}
nodelist_flat = opal_argv_join(nodelist_argv, ',');
opal_argv_free(nodelist_argv);
/* if we are using all allocated nodes, then alps
* doesn't need a nodelist, or if running without a batch scheduler
*/
if ((map->num_new_daemons < orte_num_allocated_nodes) || (orte_num_allocated_nodes == 0)) {
/* create nodelist */
nodelist_argv = NULL;
nodelist_argc = 0;
for (nnode=0; nnode < map->nodes->size; nnode++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, nnode))) {
continue;
}
/* if the daemon already exists on this node, then
* don't include it
*/
if (ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_DAEMON_LAUNCHED)) {
continue;
}
/* otherwise, add it to the list of nodes upon which
* we need to launch a daemon
*/
opal_argv_append(&nodelist_argc, &nodelist_argv, node->name);
}
if (0 == opal_argv_count(nodelist_argv)) {
orte_show_help("help-plm-alps.txt", "no-hosts-in-list", true);
rc = ORTE_ERR_FAILED_TO_START;
goto cleanup;
}
nodelist_flat = opal_argv_join(nodelist_argv, ',');
opal_argv_free(nodelist_argv);
opal_argv_append(&argc, &argv, "-L");
opal_argv_append(&argc, &argv, nodelist_flat);
free(nodelist_flat);
}
@ -351,20 +352,10 @@ static void launch_daemons(int fd, short args, void *cbdata)
/* add the daemon command (as specified by user) */
orte_plm_base_setup_orted_cmd(&argc, &argv);
/* ensure that mpirun is
* on the list. Since alps won't be launching a daemon on it,
* it won't have been placed on the list, so create a new
* version here that includes it */
asprintf(&ltmp, "%s,%s", orte_process_info.nodename, nodelist_flat);
free(nodelist_flat);
nodelist_flat = ltmp;
/* Add basic orted command line options, including debug flags */
orte_plm_base_orted_append_basic_args(&argc, &argv,
NULL,
&proc_vpid_index,
nodelist_flat);
free(nodelist_flat);
&proc_vpid_index);
/* tell the new daemons the base of the name list so they can compute
* their own name on the other end

83
orte/mca/plm/base/plm_base_launch_support.c
Просмотреть файл

@ -1037,20 +1037,6 @@ void orte_plm_base_daemon_callback(int status, orte_process_name_t* sender,
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(&daemon->name), nodename));
/* look this node up, if necessary */
if (!orte_plm_globals.daemon_nodes_assigned_at_launch) {
OPAL_OUTPUT_VERBOSE((5, orte_plm_base_framework.framework_output,
"%s plm:base:orted_report_launch attempting to assign daemon %s to node %s",
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
ORTE_NAME_PRINT(&dname), nodename));
/* to "relocate" the daemon, we just update the name of
* the node object pointed to by this daemon */
free(daemon->node->name);
daemon->node->name = strdup(nodename);
/* mark that it was verified */
ORTE_FLAG_SET(daemon->node, ORTE_NODE_FLAG_LOC_VERIFIED);
}
/* mark the daemon as launched */
ORTE_FLAG_SET(daemon->node, ORTE_NODE_FLAG_DAEMON_LAUNCHED);
@ -1312,8 +1298,7 @@ int orte_plm_base_setup_orted_cmd(int *argc, char ***argv)
*/
int orte_plm_base_orted_append_basic_args(int *argc, char ***argv,
char *ess,
int *proc_vpid_index,
char *nodes)
int *proc_vpid_index)
{
char *param = NULL;
const char **tmp_value, **tmp_value2;
@ -1321,7 +1306,6 @@ int orte_plm_base_orted_append_basic_args(int *argc, char ***argv,
char *tmp_force = NULL;
int i, j, cnt, rc;
orte_job_t *jdata;
char *rml_uri;
unsigned long num_procs;
bool ignore;
@ -1411,39 +1395,32 @@ int orte_plm_base_orted_append_basic_args(int *argc, char ***argv,
opal_argv_append(argc, argv, param);
free(param);
/* pass the uri of the hnp */
if (ORTE_PROC_IS_HNP) {
rml_uri = orte_rml.get_contact_info();
} else {
rml_uri = orte_rml.get_contact_info();
opal_argv_append(argc, argv, "-"OPAL_MCA_CMD_LINE_ID);
opal_argv_append(argc, argv, "orte_parent_uri");
opal_argv_append(argc, argv, rml_uri);
free(rml_uri);
rml_uri = strdup(orte_process_info.my_hnp_uri);
}
opal_argv_append(argc, argv, "-"OPAL_MCA_CMD_LINE_ID);
opal_argv_append(argc, argv, "orte_hnp_uri");
opal_argv_append(argc, argv, rml_uri);
free(rml_uri);
/* pass the node list if one was given*/
/* convert the nodes with daemons to a regex */
param = NULL;
if (NULL != nodes) {
/* convert the nodes to a regex */
if (ORTE_SUCCESS != (rc = orte_regex_create(nodes, &param))) {
ORTE_ERROR_LOG(rc);
return rc;
}
} else if (NULL != orte_node_regex) {
param = strdup(orte_node_regex);
if (ORTE_SUCCESS != (rc = orte_util_nidmap_create(&param))) {
ORTE_ERROR_LOG(rc);
return rc;
}
if (NULL != param) {
/* if this is too long, then we'll have to do it with
* a phone home operation instead */
if (strlen(param) < ORTE_MAX_REGEX_CMD_LENGTH) {
opal_argv_append(argc, argv, "-"OPAL_MCA_CMD_LINE_ID);
opal_argv_append(argc, argv, "orte_node_regex");
opal_argv_append(argc, argv, param);
free(param);
/* mark that the nidmap has been communicated */
orte_nidmap_communicated = true;
}
free(param);
if (!orte_static_ports && !orte_fwd_mpirun_port) {
/* if we are using static ports, or we are forwarding
* mpirun's port, then we would have built all the
* connection info and so there is nothing to be passed.
* Otherwise, we have to pass the HNP uri so we can
* phone home */
opal_argv_append(argc, argv, "-"OPAL_MCA_CMD_LINE_ID);
opal_argv_append(argc, argv, "orte_hnp_uri");
opal_argv_append(argc, argv, orte_process_info.my_hnp_uri);
}
/* if requested, pass our port */
@ -1994,7 +1971,7 @@ int orte_plm_base_setup_virtual_machine(orte_job_t *jdata)
if (orte_hnp_is_allocated) {
node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, 0);
OBJ_RETAIN(node);
opal_list_append(&nodes, &node->super);
opal_list_prepend(&nodes, &node->super);
}
for (i=0; i < jdata->apps->size; i++) {
if (NULL == (app = (orte_app_context_t*)opal_pointer_array_get_item(jdata->apps, i))) {
@ -2028,15 +2005,11 @@ int orte_plm_base_setup_virtual_machine(orte_job_t *jdata)
}
/* ensure we are not on the list */
for (item = opal_list_get_first(&nodes);
item != opal_list_get_end(&nodes);
item = opal_list_get_next(item)) {
node = (orte_node_t*)item;
if (0 == node->index) {
opal_list_remove_item(&nodes, item);
OBJ_RELEASE(item);
break;
}
item = opal_list_get_first(&nodes);
node = (orte_node_t*)item;
if (0 == node->index) {
opal_list_remove_item(&nodes, item);
OBJ_RELEASE(item);
}
/* if we didn't get anything, then we are the only node in the

3
orte/mca/plm/base/plm_private.h
Просмотреть файл

@ -114,8 +114,7 @@ ORTE_DECLSPEC void orte_plm_base_recv(int status, orte_process_name_t* sender,
*/
ORTE_DECLSPEC int orte_plm_base_orted_append_basic_args(int *argc, char ***argv,
char *ess_module,
int *proc_vpid_index,
char *nodes);
int *proc_vpid_index);
/*
* Proxy functions for use by daemons and application procs

11
orte/mca/plm/lsf/plm_lsf_module.c
Просмотреть файл

@ -160,7 +160,6 @@ static void launch_daemons(int fd, short args, void *cbdata)
int rc;
char** env = NULL;
char **nodelist_argv;
char *nodelist;
int nodelist_argc;
char *vpid_string;
int i;
@ -257,19 +256,11 @@ static void launch_daemons(int fd, short args, void *cbdata)
/* add the daemon command (as specified by user) */
orte_plm_base_setup_orted_cmd(&argc, &argv);
/* we need mpirun to be the first node on this list - since we
* aren't launching mpirun via TM, it won't be there now */
opal_argv_prepend_nosize(&nodelist_argv, orte_process_info.nodename);
nodelist = opal_argv_join(nodelist_argv, ',');
opal_argv_free(nodelist_argv);
/* Add basic orted command line options */
orte_plm_base_orted_append_basic_args(&argc, &argv,
"lsf",
&proc_vpid_index,
nodelist);
free(nodelist);
&proc_vpid_index);
/* tell the new daemons the base of the name list so they can compute
* their own name on the other end

35
orte/mca/plm/rsh/plm_rsh_module.c
Просмотреть файл

@ -328,8 +328,7 @@ static void rsh_wait_daemon(orte_proc_t *daemon, void* cbdata)
static int setup_launch(int *argcptr, char ***argvptr,
char *nodename,
int *node_name_index1,
int *proc_vpid_index, char *prefix_dir,
char *nodelist)
int *proc_vpid_index, char *prefix_dir)
{
int argc;
char **argv;
@ -613,8 +612,7 @@ static int setup_launch(int *argcptr, char ***argvptr,
*/
orte_plm_base_orted_append_basic_args(&argc, &argv,
"env",
proc_vpid_index,
nodelist);
proc_vpid_index);
/* ensure that only the ssh plm is selected on the remote daemon */
opal_argv_append_nosize(&argv, "-"OPAL_MCA_CMD_LINE_ID);
@ -828,8 +826,9 @@ static int remote_spawn(opal_buffer_t *launch)
}
/* setup the launch */
if (ORTE_SUCCESS != (rc = setup_launch(&argc, &argv, orte_process_info.nodename, &node_name_index1,
&proc_vpid_index, prefix, NULL))) {
if (ORTE_SUCCESS != (rc = setup_launch(&argc, &argv,
orte_process_info.nodename, &node_name_index1,
&proc_vpid_index, prefix))) {
ORTE_ERROR_LOG(rc);
OBJ_DESTRUCT(&coll);
goto cleanup;
@ -1030,7 +1029,6 @@ static void launch_daemons(int fd, short args, void *cbdata)
int port, *portptr;
orte_namelist_t *child;
char *rtmod;
char *nlistflat;
/* if we are launching debugger daemons, then just go
* do it - no new daemons will be launched
@ -1199,33 +1197,12 @@ static void launch_daemons(int fd, short args, void *cbdata)
orte_routed.get_routing_list(rtmod, &coll);
}
/* create a list of all nodes involved so we can pass it along */
char **nodelist = NULL;
orte_node_t *n2;
for (nnode=0; nnode < map->nodes->size; nnode++) {
if (NULL != (n2 = (orte_node_t*)opal_pointer_array_get_item(map->nodes, nnode))) {
opal_argv_append_nosize(&nodelist, n2->name);
}
}
/* we need mpirun to be the first node on this list */
if (NULL == nodelist || 0 != strcmp(nodelist[0], orte_process_info.nodename)) {
opal_argv_prepend_nosize(&nodelist, orte_process_info.nodename);
}
nlistflat = opal_argv_join(nodelist, ',');
opal_argv_free(nodelist);
/* setup the launch */
if (ORTE_SUCCESS != (rc = setup_launch(&argc, &argv, node->name, &node_name_index1,
&proc_vpid_index, prefix_dir, nlistflat))) {
&proc_vpid_index, prefix_dir))) {
ORTE_ERROR_LOG(rc);
if (NULL != nlistflat) {
free(nlistflat);
}
goto cleanup;
}
if (NULL != nlistflat) {
free(nlistflat);
}
/*
* Iterate through each of the nodes

13
orte/mca/plm/slurm/plm_slurm_module.c
Просмотреть файл

@ -323,6 +323,7 @@ static void launch_daemons(int fd, short args, void *cbdata)
goto cleanup;
}
nodelist_flat = opal_argv_join(nodelist_argv, ',');
opal_argv_free(nodelist_argv);
/* if we are using all allocated nodes, then srun doesn't
* require any further arguments
@ -336,6 +337,7 @@ static void launch_daemons(int fd, short args, void *cbdata)
opal_argv_append(&argc, &argv, tmp);
free(tmp);
}
free(nodelist_flat);
/* tell srun how many tasks to run */
asprintf(&tmp, "--ntasks=%lu", (unsigned long)map->num_new_daemons);
@ -353,18 +355,9 @@ static void launch_daemons(int fd, short args, void *cbdata)
/* add the daemon command (as specified by user) */
orte_plm_base_setup_orted_cmd(&argc, &argv);
/* we need mpirun to be the first node on this list - since we
* aren't launching mpirun via srun, it won't be there now */
opal_argv_prepend_nosize(&nodelist_argv, orte_process_info.nodename);
free(nodelist_flat);
nodelist_flat = opal_argv_join(nodelist_argv, ',');
opal_argv_free(nodelist_argv);
/* Add basic orted command line options, including debug flags */
orte_plm_base_orted_append_basic_args(&argc, &argv,
"slurm", &proc_vpid_index,
nodelist_flat);
free(nodelist_flat);
"slurm", &proc_vpid_index);
/* tell the new daemons the base of the name list so they can compute
* their own name on the other end

27
orte/mca/plm/tm/plm_tm_module.c
Просмотреть файл

@ -171,7 +171,6 @@ static void launch_daemons(int fd, short args, void *cbdata)
char **env = NULL;
char *var;
char **argv = NULL;
char **nodeargv;
int argc = 0;
int rc;
orte_std_cntr_t i;
@ -180,7 +179,6 @@ static void launch_daemons(int fd, short args, void *cbdata)
tm_task_id *tm_task_ids = NULL;
bool failed_launch = true;
mode_t current_umask;
char *nodelist;
char* vpid_string;
orte_job_t *daemons, *jdata;
orte_state_caddy_t *state = (orte_state_caddy_t*)cbdata;
@ -260,32 +258,9 @@ static void launch_daemons(int fd, short args, void *cbdata)
/* add the daemon command (as specified by user) */
orte_plm_base_setup_orted_cmd(&argc, &argv);
/* create a list of nodes in this launch */
nodeargv = NULL;
for (i = 0; i < map->nodes->size; i++) {
if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(map->nodes, i))) {
continue;
}
/* if this daemon already exists, don't launch it! */
if (ORTE_FLAG_TEST(node, ORTE_NODE_FLAG_DAEMON_LAUNCHED)) {
continue;
}
/* add to list */
opal_argv_append_nosize(&nodeargv, node->name);
}
/* we need mpirun to be the first node on this list - since we
* aren't launching mpirun via TM, it won't be there now */
opal_argv_prepend_nosize(&nodeargv, orte_process_info.nodename);
nodelist = opal_argv_join(nodeargv, ',');
opal_argv_free(nodeargv);
/* Add basic orted command line options */
orte_plm_base_orted_append_basic_args(&argc, &argv, "tm",
&proc_vpid_index,
nodelist);
&proc_vpid_index);
free(nodelist);
if (0 < opal_output_get_verbosity(orte_plm_base_framework.framework_output)) {

22
orte/mca/ras/alps/ras_alps_module.c
Просмотреть файл

@ -365,25 +365,6 @@ ras_alps_getline(FILE *fp)
return NULL;
}
static int compare_nodes (opal_list_item_t **a, opal_list_item_t **b)
{
orte_node_t *nodea = (orte_node_t *) *a;
orte_node_t *nodeb = (orte_node_t *) *b;
int32_t launcha, launchb, *ldptr;
ldptr = &launcha;
if (!orte_get_attribute(&nodea->attributes, ORTE_NODE_LAUNCH_ID, (void**)&ldptr, OPAL_INT32)) {
return 0;
}
ldptr = &launchb;
if (!orte_get_attribute(&nodeb->attributes, ORTE_NODE_LAUNCH_ID, (void**)&ldptr, OPAL_INT32)) {
return 0;
}
return (launcha > launchb) ? 1 : -1;
}
#if ALPS_APPINFO_VERSION > 0 && ALPS_APPINFO_VERSION < 3
typedef placeNodeList_t orte_ras_alps_placeNodeList_t;
#else
@ -602,8 +583,6 @@ orte_ras_alps_read_appinfo_file(opal_list_t *nodes, char *filename,
break; /* Extended details ignored */
}
opal_list_sort (nodes, compare_nodes);
free(cpBuf); /* Free the buffer */
return ORTE_SUCCESS;
@ -617,4 +596,3 @@ orte_ras_alps_finalize(void)
"ras:alps:finalize: success (nothing to do)");
return ORTE_SUCCESS;
}

84
orte/mca/state/dvm/state_dvm.c
Просмотреть файл

@ -243,6 +243,7 @@ static void vm_ready(int fd, short args, void *cbdata)
opal_byte_object_t bo, *boptr;
int8_t flag;
int32_t numbytes;
char *nidmap;
/* if this is my job, then we are done */
if (ORTE_PROC_MY_NAME->jobid == caddy->jdata->jobid) {
@ -250,50 +251,65 @@ static void vm_ready(int fd, short args, void *cbdata)
* do this here so we don't have to do it for every
* job we are going to launch */
buf = OBJ_NEW(opal_buffer_t);
/* pack the "load nidmap" cmd */
if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &command, 1, ORTE_DAEMON_CMD))) {
ORTE_ERROR_LOG(rc);
OBJ_RELEASE(buf);
return;
opal_dss.pack(buf, &command, 1, ORTE_DAEMON_CMD);
/* if we couldn't provide the allocation regex on the orted
* cmd line, then we need to provide all the info here */
if (!orte_nidmap_communicated) {
if (ORTE_SUCCESS != (rc = orte_util_nidmap_create(&nidmap))) {
ORTE_ERROR_LOG(rc);
OBJ_RELEASE(buf);
return;
}
orte_nidmap_communicated = true;
} else {
nidmap = NULL;
}
/* flag that daemons were launched so we will update the nidmap */
flag = 1;
opal_dss.pack(buf, &flag, 1, OPAL_INT8);
/* construct a nodemap with everything in it */
if (ORTE_SUCCESS != (rc = orte_util_encode_nodemap(buf))) {
ORTE_ERROR_LOG(rc);
OBJ_RELEASE(buf);
return;
opal_dss.pack(buf, &nidmap, 1, OPAL_STRING);
if (NULL != nidmap) {
free(nidmap);
}
if (!orte_static_ports && !orte_fwd_mpirun_port) {
/* pack a flag indicating wiring info is provided */
/* provide the info on the capabilities of each node */
if (!orte_node_info_communicated) {
flag = 1;
opal_dss.pack(buf, &flag, 1, OPAL_INT8);
/* get wireup info for daemons per the selected routing module */
wireup = OBJ_NEW(opal_buffer_t);
if (ORTE_SUCCESS != (rc = orte_rml_base_get_contact_info(ORTE_PROC_MY_NAME->jobid, wireup))) {
if (ORTE_SUCCESS != (rc = orte_util_encode_nodemap(buf))) {
ORTE_ERROR_LOG(rc);
OBJ_RELEASE(wireup);
OBJ_RELEASE(buf);
return;
}
/* put it in a byte object for xmission */
opal_dss.unload(wireup, (void**)&bo.bytes, &numbytes);
/* pack the byte object - zero-byte objects are fine */
bo.size = numbytes;
boptr = &bo;
if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &boptr, 1, OPAL_BYTE_OBJECT))) {
ORTE_ERROR_LOG(rc);
orte_node_info_communicated = true;
if (!orte_static_ports && !orte_fwd_mpirun_port) {
/* pack a flag indicating wiring info is provided */
flag = 1;
opal_dss.pack(buf, &flag, 1, OPAL_INT8);
/* get wireup info for daemons per the selected routing module */
wireup = OBJ_NEW(opal_buffer_t);
if (ORTE_SUCCESS != (rc = orte_rml_base_get_contact_info(ORTE_PROC_MY_NAME->jobid, wireup))) {
ORTE_ERROR_LOG(rc);
OBJ_RELEASE(wireup);
OBJ_RELEASE(buf);
return;
}
/* put it in a byte object for xmission */
opal_dss.unload(wireup, (void**)&bo.bytes, &numbytes);
/* pack the byte object - zero-byte objects are fine */
bo.size = numbytes;
boptr = &bo;
if (ORTE_SUCCESS != (rc = opal_dss.pack(buf, &boptr, 1, OPAL_BYTE_OBJECT))) {
ORTE_ERROR_LOG(rc);
OBJ_RELEASE(wireup);
OBJ_RELEASE(buf);
return;
}
/* release the data since it has now been copied into our buffer */
if (NULL != bo.bytes) {
free(bo.bytes);
}
OBJ_RELEASE(wireup);
OBJ_RELEASE(buf);
return;
} else {
flag = 0;
opal_dss.pack(buf, &flag, 1, OPAL_INT8);
}
/* release the data since it has now been copied into our buffer */
if (NULL != bo.bytes) {
free(bo.bytes);
}
OBJ_RELEASE(wireup);
} else {
flag = 0;
opal_dss.pack(buf, &flag, 1, OPAL_INT8);

2
orte/runtime/orte_globals.c
Просмотреть файл

@ -108,6 +108,8 @@ bool orte_display_allocation = false;
bool orte_display_devel_allocation = false;
bool orte_soft_locations = false;
int orted_pmi_version = 0;
bool orte_nidmap_communicated = false;
bool orte_node_info_communicated = false;
/* launch agents */
char *orte_launch_agent = NULL;

2
orte/runtime/orte_globals.h
Просмотреть файл

@ -491,6 +491,8 @@ ORTE_DECLSPEC extern bool orte_display_allocation;
ORTE_DECLSPEC extern bool orte_display_devel_allocation;
ORTE_DECLSPEC extern bool orte_soft_locations;
ORTE_DECLSPEC extern bool orte_hnp_connected;
ORTE_DECLSPEC extern bool orte_nidmap_communicated;
ORTE_DECLSPEC extern bool orte_node_info_communicated;
/* launch agents */
ORTE_DECLSPEC extern char *orte_launch_agent;

806
orte/util/nidmap.c
Просмотреть файл

Разница между файлами не показана из-за своего большого размера Загрузить разницу

12
orte/util/nidmap.h
Просмотреть файл

@ -37,10 +37,16 @@
BEGIN_C_DECLS
#define ORTE_MAX_REGEX_CMD_LENGTH 1024
#define ORTE_MAX_NODE_PREFIX 50
#define ORTE_CONTIG_NODE_CMD 0x01
#define ORTE_NON_CONTIG_NODE_CMD 0x02
ORTE_DECLSPEC int orte_util_nidmap_create(char **regex);
ORTE_DECLSPEC int orte_util_nidmap_parse(char *regex);
/* create a regular expression describing the nodes in the
* allocation */
ORTE_DECLSPEC int orte_util_encode_nodemap(opal_buffer_t *buffer);
@ -49,11 +55,7 @@ ORTE_DECLSPEC int orte_util_encode_nodemap(opal_buffer_t *buffer);
* into the orte_node_pool array */
ORTE_DECLSPEC int orte_util_decode_daemon_nodemap(opal_buffer_t *buffer);
ORTE_DECLSPEC int orte_util_build_daemon_nidmap(char **nodes);
ORTE_DECLSPEC int orte_util_encode_topologies(opal_buffer_t *buffer);
ORTE_DECLSPEC int orte_util_decode_topologies(opal_buffer_t *buffer);
ORTE_DECLSPEC int orte_util_build_daemon_nidmap(void);
END_C_DECLS

226
orte/util/regex.c
Просмотреть файл

@ -63,230 +63,6 @@
static int regex_parse_node_ranges(char *base, char *ranges, int num_digits, char *suffix, char ***names);
static int regex_parse_node_range(char *base, char *range, int num_digits, char *suffix, char ***names);
int orte_regex_create(char *nodelist, char **regexp)
{
char *node;
char prefix[ORTE_MAX_NODE_PREFIX];
int i, j, len, startnum, nodenum, numdigits;
bool found, fullname;
char *suffix, *sfx;
orte_regex_node_t *ndreg;
orte_regex_range_t *range;
opal_list_t nodeids;
opal_list_item_t *item, *itm2;
char **regexargs = NULL, *tmp, *tmp2;
char *cptr;
/* define the default */
*regexp = NULL;
cptr = strchr(nodelist, ',');
if (NULL == cptr) {
/* if there is only one node, don't bother */
*regexp = strdup(nodelist);
return ORTE_SUCCESS;
}
/* setup the list of results */
OBJ_CONSTRUCT(&nodeids, opal_list_t);
/* cycle thru the array of nodenames */
node = nodelist;
while (NULL != (cptr = strchr(node, ',')) || 0 < strlen(node)) {
if (NULL != cptr) {
*cptr = '\0';
}
/* determine this node's prefix by looking for first non-alpha char */
fullname = false;
len = strlen(node);
startnum = -1;
memset(prefix, 0, ORTE_MAX_NODE_PREFIX);
numdigits = 0;
for (i=0, j=0; i < len; i++) {
if (!isalpha(node[i])) {
/* found a non-alpha char */
if (!isdigit(node[i])) {
/* if it is anything but a digit, we just use
* the entire name
*/
fullname = true;
break;
}
/* count the size of the numeric field - but don't
* add the digits to the prefix
*/
numdigits++;
if (startnum < 0) {
/* okay, this defines end of the prefix */
startnum = i;
}
continue;
}
if (startnum < 0) {
prefix[j++] = node[i];
}
}
if (fullname || startnum < 0) {
/* can't compress this name - just add it to the list */
ndreg = OBJ_NEW(orte_regex_node_t);
ndreg->prefix = strdup(node);
opal_list_append(&nodeids, &ndreg->super);
/* move to the next posn */
if (NULL == cptr) {
break;
}
node = cptr + 1;
continue;
}
/* convert the digits and get any suffix */
nodenum = strtol(&node[startnum], &sfx, 10);
if (NULL != sfx) {
suffix = strdup(sfx);
} else {
suffix = NULL;
}
/* is this nodeid already on our list? */
found = false;
for (item = opal_list_get_first(&nodeids);
!found && item != opal_list_get_end(&nodeids);
item = opal_list_get_next(item)) {
ndreg = (orte_regex_node_t*)item;
if (0 < strlen(prefix) && NULL == ndreg->prefix) {
continue;
}
if (0 == strlen(prefix) && NULL != ndreg->prefix) {
continue;
}
if (0 < strlen(prefix) && NULL != ndreg->prefix
&& 0 != strcmp(prefix, ndreg->prefix)) {
continue;
}
if (NULL == suffix && NULL != ndreg->suffix) {
continue;
}
if (NULL != suffix && NULL == ndreg->suffix) {
continue;
}
if (NULL != suffix && NULL != ndreg->suffix &&
0 != strcmp(suffix, ndreg->suffix)) {
continue;
}
if (numdigits != ndreg->num_digits) {
continue;
}
/* found a match - flag it */
found = true;
/* get the last range on this nodeid - we do this
* to preserve order
*/
range = (orte_regex_range_t*)opal_list_get_last(&ndreg->ranges);
if (NULL == range) {
/* first range for this nodeid */
range = OBJ_NEW(orte_regex_range_t);
range->start = nodenum;
range->cnt = 1;
opal_list_append(&ndreg->ranges, &range->super);
break;
}
/* see if the node number is out of sequence */
if (nodenum != (range->start + range->cnt)) {
/* start a new range */
range = OBJ_NEW(orte_regex_range_t);
range->start = nodenum;
range->cnt = 1;
opal_list_append(&ndreg->ranges, &range->super);
break;
}
/* everything matches - just increment the cnt */
range->cnt++;
break;
}
if (!found) {
/* need to add it */
ndreg = OBJ_NEW(orte_regex_node_t);
if (0 < strlen(prefix)) {
ndreg->prefix = strdup(prefix);
}
if (NULL != suffix) {
ndreg->suffix = strdup(suffix);
}
ndreg->num_digits = numdigits;
opal_list_append(&nodeids, &ndreg->super);
/* record the first range for this nodeid - we took
* care of names we can't compress above
*/
range = OBJ_NEW(orte_regex_range_t);
range->start = nodenum;
range->cnt = 1;
opal_list_append(&ndreg->ranges, &range->super);
}
if (NULL != suffix) {
free(suffix);
}
/* move to the next posn */
if (NULL == cptr) {
break;
}
node = cptr + 1;
}
/* begin constructing the regular expression */
while (NULL != (item = opal_list_remove_first(&nodeids))) {
ndreg = (orte_regex_node_t*)item;
/* if no ranges, then just add the name */
if (0 == opal_list_get_size(&ndreg->ranges)) {
if (NULL != ndreg->prefix) {
/* solitary node */
asprintf(&tmp, "%s", ndreg->prefix);
opal_argv_append_nosize(&regexargs, tmp);
free(tmp);
}
OBJ_RELEASE(ndreg);
continue;
}
/* start the regex for this nodeid with the prefix */
if (NULL != ndreg->prefix) {
asprintf(&tmp, "%s[%d:", ndreg->prefix, ndreg->num_digits);
} else {
asprintf(&tmp, "[%d:", ndreg->num_digits);
}
/* add the ranges */
while (NULL != (itm2 = opal_list_remove_first(&ndreg->ranges))) {
range = (orte_regex_range_t*)itm2;
if (1 == range->cnt) {
asprintf(&tmp2, "%s%d,", tmp, range->start);
} else {
asprintf(&tmp2, "%s%d-%d,", tmp, range->start, range->start + range->cnt - 1);
}
free(tmp);
tmp = tmp2;
OBJ_RELEASE(range);
}
/* replace the final comma */
tmp[strlen(tmp)-1] = ']';
if (NULL != ndreg->suffix) {
/* add in the suffix, if provided */
asprintf(&tmp2, "%s%s", tmp, ndreg->suffix);
free(tmp);
tmp = tmp2;
}
opal_argv_append_nosize(&regexargs, tmp);
free(tmp);
OBJ_RELEASE(ndreg);
}
/* assemble final result */
*regexp = opal_argv_join(regexargs, ',');
/* cleanup */
opal_argv_free(regexargs);
OBJ_DESTRUCT(&nodeids);
return ORTE_SUCCESS;
}
int orte_regex_extract_node_names(char *regexp, char ***names)
{
int i, j, k, len, ret;
@ -592,7 +368,7 @@ static int regex_parse_node_range(char *base, char *range, int num_digits, char
static void range_construct(orte_regex_range_t *ptr)
{
ptr->start = 0;
ptr->vpid = 0;
ptr->cnt = 0;
}
OBJ_CLASS_INSTANCE(orte_regex_range_t,

8
orte/util/regex.h
Просмотреть файл

@ -36,8 +36,7 @@ BEGIN_C_DECLS
typedef struct {
opal_list_item_t super;
int start;
int endpt;
int vpid;
int cnt;
int slots;
orte_topology_t *t;
@ -54,11 +53,6 @@ typedef struct {
} orte_regex_node_t;
ORTE_DECLSPEC OBJ_CLASS_DECLARATION(orte_regex_node_t);
/* NOTE: this is a destructive call for the nodes param - the
* function will search and replace all commas with '\0'
*/
ORTE_DECLSPEC int orte_regex_create(char *nodes, char **regexp);
ORTE_DECLSPEC int orte_regex_extract_node_names(char *regexp, char ***names);
END_C_DECLS