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

Merge pull request #3595 from rhc54/topic/dyn

Просмотр исходного кода 
Update the connect/accept support
Этот коммит содержится в:
Ralph Castain 2017-05-27 11:32:08 -07:00 коммит произвёл GitHub
родитель a6f61132d7 9f60cd0fe7
Коммит c99978a767
14 изменённых файлов: 115 добавлений и 20 удалений
Показать статистику Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы

9
ompi/dpm/dpm.c
Просмотреть файл

@ -15,7 +15,7 @@
* Copyright (c) 2009 Sun Microsystems, Inc. All rights reserved.
* Copyright (c) 2011-2015 Los Alamos National Security, LLC. All rights
* reserved.
* Copyright (c) 2013-2016 Intel, Inc. All rights reserved.
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved.
* Copyright (c) 2014-2017 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
* $COPYRIGHT$
@ -40,6 +40,7 @@
#include "opal/util/argv.h"
#include "opal/util/opal_getcwd.h"
#include "opal/util/proc.h"
#include "opal/util/show_help.h"
#include "opal/dss/dss.h"
#include "opal/mca/hwloc/base/base.h"
#include "opal/mca/pmix/pmix.h"
@ -112,6 +113,12 @@ int ompi_dpm_connect_accept(ompi_communicator_t *comm, int root,
if (NULL == opal_pmix.publish || NULL == opal_pmix.connect ||
NULL == opal_pmix.unpublish ||
(NULL == opal_pmix.lookup && NULL == opal_pmix.lookup_nb)) {
/* print a nice message explaining we don't have support */
opal_show_help("help-mpi-runtime.txt", "noconxcpt", true);
return OMPI_ERR_NOT_SUPPORTED;
}
if (!ompi_rte_connect_accept_support(port_string)) {
/* they will have printed the help message */
return OMPI_ERR_NOT_SUPPORTED;
}

5
ompi/mca/rte/orte/rte_orte.h
Просмотреть файл

@ -1,7 +1,7 @@
/*
* Copyright (c) 2012-2013 Los Alamos National Security, LLC.
* All rights reserved.
* Copyright (c) 2013-2015 Intel, Inc. All rights reserved
* Copyright (c) 2013-2017 Intel, Inc. All rights reserved.
* Copyright (c) 2014 Cisco Systems, Inc. All rights reserved.
* Copyright (c) 2014-2016 Research Organization for Information Science
* and Technology (RIST). All rights reserved.
@ -116,6 +116,9 @@ static inline orte_process_name_t * OMPI_CAST_RTE_NAME(opal_process_name_t * nam
}
#endif
/* check dynamics support */
OMPI_DECLSPEC bool ompi_rte_connect_accept_support(const char *port);
END_C_DECLS
#endif /* MCA_OMPI_RTE_ORTE_H */

45
ompi/mca/rte/orte/rte_orte_module.c
Просмотреть файл

@ -39,6 +39,7 @@
#include "orte/mca/routed/routed.h"
#include "orte/util/name_fns.h"
#include "orte/util/session_dir.h"
#include "orte/util/show_help.h"
#include "orte/runtime/orte_globals.h"
#include "orte/runtime/orte_wait.h"
#include "orte/runtime/orte_data_server.h"
@ -198,3 +199,47 @@ void ompi_rte_wait_for_debugger(void)
opal_pmix.deregister_evhandler(handler, NULL, NULL);
}
}
bool ompi_rte_connect_accept_support(const char *port)
{
char *ptr, *tmp;
orte_process_name_t name;
/* were we launched by mpirun, or are we calling
* without a defined port? */
if (NULL == orte_process_info.my_hnp_uri ||
0 == strlen(port)) {
return true;
}
/* is the job family in the port different than my own? */
tmp = strdup(port); // protect input
if (NULL == (ptr = strchr(tmp, ':'))) {
/* this port didn't come from us! */
orte_show_help("help-orterun.txt", "orterun:malformedport", true);
free(tmp);
return false;
}
*ptr = '\0';
if (ORTE_SUCCESS != orte_util_convert_string_to_process_name(&name, tmp)) {
free(tmp);
orte_show_help("help-orterun.txt", "orterun:malformedport", true);
return false;
}
free(tmp);
if (ORTE_JOB_FAMILY(ORTE_PROC_MY_NAME->jobid) == ORTE_JOB_FAMILY(name.jobid)) {
/* same job family, so our infrastructure is adequate */
return true;
}
/* if the job family of the port is different than our own
* and we were launched by mpirun, then we require ompi-server
* support */
if (NULL == orte_data_server_uri) {
/* print a pretty help message */
orte_show_help("help-orterun.txt", "orterun:server-unavailable", true);
return false;
}
return true;
}

11
ompi/runtime/help-mpi-runtime.txt
Просмотреть файл

@ -12,6 +12,7 @@
# All rights reserved.
# Copyright (c) 2007-2015 Cisco Systems, Inc. All rights reserved.
# Copyright (c) 2013 NVIDIA Corporation. All rights reserved.
# Copyright (c) 2017 Intel, Inc. All rights reserved.
# $COPYRIGHT$
#
# Additional copyrights may follow
@ -93,3 +94,13 @@ Open MPI with --enable-heterogeneous.
[no cuda support]
The user requested CUDA support with the --mca mpi_cuda_support 1 flag
but the library was not compiled with any support.
#
[noconxcpt]
The user has called an operation involving MPI_Connect and/or MPI_Accept,
but this environment lacks the necessary infrastructure support for
that operation. Open MPI relies on the PMIx_Publish/Lookup (or one of
its predecessors) APIs for this operation.
This typically happens when launching outside of mpirun where the underlying
resource manager does not provide publish/lookup support. One way of solving
the problem is to simply use mpirun to start the application.

3
opal/mca/pmix/base/base.h
Просмотреть файл

@ -1,5 +1,5 @@
/*
* Copyright (c) 2014-2016 Intel, Inc. All rights reserved.
* Copyright (c) 2014-2017 Intel, Inc. All rights reserved.
* $COPYRIGHT$
*
* Additional copyrights may follow
@ -57,6 +57,7 @@ OPAL_DECLSPEC void opal_pmix_base_set_evbase(opal_event_base_t *evbase);
typedef struct {
opal_event_base_t *evbase;
int timeout;
} opal_pmix_base_t;
extern opal_pmix_base_t opal_pmix_base;

7
opal/mca/pmix/base/pmix_base_fns.c
Просмотреть файл

@ -193,7 +193,12 @@ int opal_pmix_base_exchange(opal_value_t *indat,
info = OBJ_NEW(opal_value_t);
info->key = strdup(OPAL_PMIX_TIMEOUT);
info->type = OPAL_INT;
info->data.integer = timeout;
if (0 < opal_pmix_base.timeout) {
/* the user has overridden the default */
info->data.integer = opal_pmix_base.timeout;
} else {
info->data.integer = timeout;
}
opal_list_append(&mlist, &info->super);
/* if a non-blocking version of lookup isn't

6
opal/mca/pmix/base/pmix_base_frame.c
Просмотреть файл

@ -47,6 +47,12 @@ static int opal_pmix_base_frame_register(mca_base_register_flag_t flags)
(void) mca_base_var_register("opal", "pmix", "base", "collect_data", "Collect all data during modex",
MCA_BASE_VAR_TYPE_BOOL, NULL, 0, 0, OPAL_INFO_LVL_9,
MCA_BASE_VAR_SCOPE_READONLY, &opal_pmix_collect_all_data);
opal_pmix_base.timeout = -1;
(void) mca_base_var_register("opal", "pmix", "base", "exchange_timeout",
"Time (in seconds) to wait for a data exchange to complete",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0, OPAL_INFO_LVL_3,
MCA_BASE_VAR_SCOPE_READONLY, &opal_pmix_base.timeout);
return OPAL_SUCCESS;
}

4
orte/orted/help-orted.txt
Просмотреть файл

@ -67,7 +67,9 @@ A request has timed out and will therefore fail:
Operation: %s
Your job may terminate as a result of this problem. You may want to
adjust the MCA parameter pmix_server_max_wait and try again.
adjust the MCA parameter pmix_server_max_wait and try again. If this
occurred during a connect/accept operation, you can adjust that time
using the pmix_base_exchange_timeout parameter.
#
[noroom]
A request for an asynchronous runtime operation cannot be fulfilled

8
orte/orted/pmix/pmix_server.c
Просмотреть файл

@ -138,14 +138,6 @@ void pmix_server_register_params(void)
OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_ALL,
&orte_pmix_server_globals.timeout);
/* register the URI of the UNIVERSAL data server */
orte_pmix_server_globals.server_uri = NULL;
(void) mca_base_var_register ("orte", "pmix", NULL, "server_uri",
"URI of a session-level keyval server for publish/lookup operations",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0,
OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_ALL,
&orte_pmix_server_globals.server_uri);
/* whether or not to wait for the universal server */
orte_pmix_server_globals.wait_for_server = false;
(void) mca_base_var_register ("orte", "pmix", NULL, "wait_for_server",

11
orte/orted/pmix/pmix_server_pub.c
Просмотреть файл

@ -59,13 +59,13 @@ static int init_server(void)
/* if the universal server wasn't specified, then we use
* our own HNP for that purpose */
if (NULL == orte_pmix_server_globals.server_uri) {
if (NULL == orte_data_server_uri) {
orte_pmix_server_globals.server = *ORTE_PROC_MY_HNP;
} else {
if (0 == strncmp(orte_pmix_server_globals.server_uri, "file", strlen("file")) ||
0 == strncmp(orte_pmix_server_globals.server_uri, "FILE", strlen("FILE"))) {
if (0 == strncmp(orte_data_server_uri, "file", strlen("file")) ||
0 == strncmp(orte_data_server_uri, "FILE", strlen("FILE"))) {
/* it is a file - get the filename */
filename = strchr(orte_pmix_server_globals.server_uri, ':');
filename = strchr(orte_data_server_uri, ':');
if (NULL == filename) {
/* filename is not correctly formatted */
orte_show_help("help-orterun.txt", "orterun:ompi-server-filename-bad", true,
@ -121,7 +121,6 @@ static int init_server(void)
* as a background job - e.g., in scripts
*/
if (orte_pmix_server_globals.wait_for_server) {
opal_output(0, "WAIT");
/* ping the server */
struct timeval timeout;
timeout.tv_sec = orte_pmix_server_globals.timeout;
@ -141,8 +140,6 @@ static int init_server(void)
}
}
opal_output(0, "SERVER READY");
return ORTE_SUCCESS;
}

1
orte/runtime/orte_globals.c
Просмотреть файл

@ -77,6 +77,7 @@ char *orte_coll_transport = NULL;
int orte_mgmt_conduit = -1;
int orte_coll_conduit = -1;
bool orte_no_vm = false;
char *orte_data_server_uri = NULL;
/* ORTE OOB port flags */
bool orte_static_ports = false;

1
orte/runtime/orte_globals.h
Просмотреть файл

@ -457,6 +457,7 @@ ORTE_DECLSPEC extern bool orte_coprocessors_detected;
ORTE_DECLSPEC extern opal_hash_table_t *orte_coprocessors;
ORTE_DECLSPEC extern char *orte_topo_signature;
ORTE_DECLSPEC extern bool orte_no_vm;
ORTE_DECLSPEC extern char *orte_data_server_uri;
/* ORTE OOB port flags */
ORTE_DECLSPEC extern bool orte_static_ports;

9
orte/runtime/orte_mca_params.c
Просмотреть файл

@ -788,5 +788,14 @@ int orte_register_params(void)
OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_READONLY,
&orte_fwd_mpirun_port);
/* register the URI of the UNIVERSAL data server */
orte_data_server_uri = NULL;
(void) mca_base_var_register ("orte", "pmix", NULL, "server_uri",
"URI of a session-level keyval server for publish/lookup operations",
MCA_BASE_VAR_TYPE_STRING, NULL, 0, 0,
OPAL_INFO_LVL_3, MCA_BASE_VAR_SCOPE_ALL,
&orte_data_server_uri);
return ORTE_SUCCESS;
}

15
orte/tools/orterun/help-orterun.txt
Просмотреть файл

@ -458,6 +458,21 @@ Error received: %s
Please check to ensure that the requested server matches the actual server
information, and that the server is in operation.
#
[orterun:server-unavailable]
The user has called an operation involving MPI_Connect and/or MPI_Accept
that spans multiple invocations of mpirun. This requires the support of
the ompi-server tool, which must be executing somewhere that can be
accessed by all participants.
Please ensure the tool is running, and provide each mpirun with the MCA
parameter "pmix_server_uri" pointing to it.
#
[orterun:malformedport]
An operation involving MPI_Connect and/or MPI_Accept was called with
an unrecognized port string. This typically happens when passing the
string on a cmd line and failing to properly quote it to protect
against the special characters it includes
#
[orterun:ompi-server-pid-bad]
%s was unable to parse the PID of the %s to be used as the ompi-server.
The option we were given was: