ess/alps: add initial alps ess component
Note this alps ess component has nothing to do with the old CNOS alps component used on Cray Seastar/Portals3 (Cray XT) systems. To work properly, changes need to be made to the open method of the ess/pmi component to keep it from selecting, and thus initializing, the opal/pmix/cray component.
Этот коммит содержится в:
Howard Pritchard
родитель
32bf0e7b7e
Коммит
a753c3ece0
50
orte/mca/ess/alps/Makefile.am
Обычный файл
50
orte/mca/ess/alps/Makefile.am
Обычный файл
@ -0,0 +1,50 @@
|
||||
#
|
||||
# Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
||||
# University Research and Technology
|
||||
# Corporation. All rights reserved.
|
||||
# Copyright (c) 2004-2005 The University of Tennessee and The University
|
||||
# of Tennessee Research Foundation. All rights
|
||||
# reserved.
|
||||
# Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
||||
# University of Stuttgart. All rights reserved.
|
||||
# Copyright (c) 2004-2005 The Regents of the University of California.
|
||||
# All rights reserved.
|
||||
# Copyright (c) 2008-2010 Cisco Systems, Inc. All rights reserved.
|
||||
# $COPYRIGHT$
|
||||
#
|
||||
# Additional copyrights may follow
|
||||
#
|
||||
# $HEADER$
|
||||
#
|
||||
|
||||
sources = \
|
||||
ess_alps.h \
|
||||
ess_alps_component.c \
|
||||
ess_alps_module.c \
|
||||
ess_alps_utils.c
|
||||
|
||||
# Make the output library in this directory, and name it either
|
||||
# mca_<type>_<name>.la (for DSO builds) or libmca_<type>_<name>.la
|
||||
# (for static builds).
|
||||
|
||||
if MCA_BUILD_orte_ess_alps_DSO
|
||||
component_noinst =
|
||||
component_install = mca_ess_alps.la
|
||||
else
|
||||
component_noinst = libmca_ess_alps.la
|
||||
component_install =
|
||||
endif
|
||||
|
||||
mcacomponentdir = $(ompilibdir)
|
||||
mcacomponent_LTLIBRARIES = $(component_install)
|
||||
mca_ess_alps_la_SOURCES = $(sources)
|
||||
mca_ess_alps_la_CPPFLAGS = $(ess_alps_CPPFLAGS) -fno-ident
|
||||
mca_ess_alps_la_LDFLAGS = -module -avoid-version $(ess_alps_LDFLAGS)
|
||||
mca_ess_alps_la_LIBADD = $(ess_alps_LDFLAGS)
|
||||
|
||||
noinst_LTLIBRARIES = $(component_noinst)
|
||||
libmca_ess_alps_la_SOURCES =$(sources)
|
||||
libmca_ess_alps_la_CPPFLAGS = $(ess_alps_CPPFLAGS)
|
||||
libmca_ess_alps_la_LDFLAGS = -module -avoid-version $(ess_alps_LDFLAGS)
|
||||
libmca_ess_alps_la_LIBADD = $(ess_alps_LIBS)
|
||||
|
||||
52
orte/mca/ess/alps/configure.m4
Обычный файл
52
orte/mca/ess/alps/configure.m4
Обычный файл
@ -0,0 +1,52 @@
|
||||
# -*- shell-script -*-
|
||||
#
|
||||
# Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
||||
# University Research and Technology
|
||||
# Corporation. All rights reserved.
|
||||
# Copyright (c) 2004-2005 The University of Tennessee and The University
|
||||
# of Tennessee Research Foundation. All rights
|
||||
# reserved.
|
||||
# Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
||||
# University of Stuttgart. All rights reserved.
|
||||
# Copyright (c) 2004-2005 The Regents of the University of California.
|
||||
# All rights reserved.
|
||||
# Copyright (c) 2009-2012 Cisco Systems, Inc. All rights reserved.
|
||||
# Copyright (c) 2011 Los Alamos National Security, LLC.
|
||||
# All rights reserved.
|
||||
# $COPYRIGHT$
|
||||
#
|
||||
# Additional copyrights may follow
|
||||
#
|
||||
# $HEADER$
|
||||
#
|
||||
|
||||
# MCA_ess_alps_CONFIG([action-if-found], [action-if-not-found])
|
||||
# -----------------------------------------------------------
|
||||
AC_DEFUN([MCA_orte_ess_alps_CONFIG],[
|
||||
AC_CONFIG_FILES([orte/mca/ess/alps/Makefile])
|
||||
|
||||
ess_alps_lli_happy="no"
|
||||
ess_alps_util_happy="no"
|
||||
|
||||
PKG_CHECK_MODULES([CRAY_ALPS_LLI], [cray-alpslli],
|
||||
[ess_alps_CPPFLAGS=$CRAY_ALPS_LLI_CFLAGS
|
||||
ess_alps_LDFLAGS=$CRAY_ALPS_LLI_LIBS
|
||||
ess_alps_LIBS=$CRAY_ALPS_LLI_LIBS
|
||||
ess_alps_lli_happy="yes"],
|
||||
[AC_MSG_RESULT([no])])
|
||||
|
||||
PKG_CHECK_MODULES([CRAY_ALPS_UTIL], [cray-alpsutil],
|
||||
[ess_alps_CPPFLAGS="$ess_alps_CPPFLAGS $CRAY_ALPS_UTIL_CFLAGS"
|
||||
ess_alps_LDFLAGS="$ess_alps_LDFLAGS $CRAY_ALPS_UTIL_LIBS"
|
||||
ess_alps_LIBS="$ess_alps_LIBS $CRAY_ALPS_LLI_LIBS"
|
||||
ess_alps_util_happy="yes"],
|
||||
[AC_MSG_RESULT([no])])
|
||||
|
||||
AS_IF([test "$ess_alps_lli_happy" = "yes" -a "$ess_alps_util_happy" = "yes"],
|
||||
[$1
|
||||
AC_SUBST([ess_alps_CPPFLAGS])
|
||||
AC_SUBST([ess_alps_LDFLAGS])
|
||||
AC_SUBST([ess_alps_LIBS])],
|
||||
[$2])
|
||||
|
||||
])dnl
|
||||
56
orte/mca/ess/alps/ess_alps.h
Обычный файл
56
orte/mca/ess/alps/ess_alps.h
Обычный файл
@ -0,0 +1,56 @@
|
||||
/*
|
||||
* Copyright (c) 2004-2008 The Trustees of Indiana University and Indiana
|
||||
* University Research and Technology
|
||||
* Corporation. All rights reserved.
|
||||
* Copyright (c) 2004-2005 The University of Tennessee and The University
|
||||
* of Tennessee Research Foundation. All rights
|
||||
* reserved.
|
||||
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
||||
* University of Stuttgart. All rights reserved.
|
||||
* Copyright (c) 2004-2005 The Regents of the University of California.
|
||||
* All rights reserved.
|
||||
* $COPYRIGHT$
|
||||
*
|
||||
* Additional copyrights may follow
|
||||
*
|
||||
* $HEADER$
|
||||
*/
|
||||
|
||||
#ifndef ORTE_ESS_ALPS_H
|
||||
#define ORTE_ESS_ALPS_H
|
||||
|
||||
#include "orte_config.h"
|
||||
#include "opal/mca/mca.h"
|
||||
#include "orte/mca/ess/ess.h"
|
||||
|
||||
#include "alps/alps.h"
|
||||
#include "alps/alps_toolAssist.h"
|
||||
#include "alps/libalpsutil.h"
|
||||
#include "alps/libalpslli.h"
|
||||
|
||||
BEGIN_C_DECLS
|
||||
|
||||
/*
|
||||
* Module open / close
|
||||
*/
|
||||
int orte_ess_alps_component_open(void);
|
||||
int orte_ess_alps_component_close(void);
|
||||
int orte_ess_alps_component_query(mca_base_module_t **module, int *priority);
|
||||
|
||||
/*
|
||||
* alps component internal utility functions
|
||||
*/
|
||||
|
||||
int orte_ess_alps_get_first_rank_on_node(int *first_rank);
|
||||
int orte_ess_alps_sync_start(void);
|
||||
int orte_ess_alps_sync_complete(void);
|
||||
|
||||
/*
|
||||
* ODLS Alps module
|
||||
*/
|
||||
extern orte_ess_base_module_t orte_ess_alps_module;
|
||||
ORTE_MODULE_DECLSPEC extern orte_ess_base_component_t mca_ess_alps_component;
|
||||
|
||||
END_C_DECLS
|
||||
|
||||
#endif /* ORTE_ESS_ALPS_H */
|
||||
123
orte/mca/ess/alps/ess_alps_component.c
Обычный файл
123
orte/mca/ess/alps/ess_alps_component.c
Обычный файл
@ -0,0 +1,123 @@
|
||||
/*
|
||||
* Copyright (c) 2004-2008 The Trustees of Indiana University and Indiana
|
||||
* University Research and Technology
|
||||
* Corporation. All rights reserved.
|
||||
* Copyright (c) 2004-2005 The University of Tennessee and The University
|
||||
* of Tennessee Research Foundation. All rights
|
||||
* reserved.
|
||||
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
||||
* University of Stuttgart. All rights reserved.
|
||||
* Copyright (c) 2004-2005 The Regents of the University of California.
|
||||
* All rights reserved.
|
||||
* Copyright (c) 2011-2012 Los Alamos National Security, LLC.
|
||||
* All rights reserved.
|
||||
* $COPYRIGHT$
|
||||
*
|
||||
* Additional copyrights may follow
|
||||
*
|
||||
* $HEADER$
|
||||
*
|
||||
* These symbols are in a file by themselves to provide nice linker
|
||||
* semantics. Since linkers generally pull in symbols by object
|
||||
* files, keeping these symbols as the only symbols in this file
|
||||
* prevents utility programs such as "ompi_info" from having to import
|
||||
* entire components just to query their version and parameters.
|
||||
*/
|
||||
|
||||
#include "orte_config.h"
|
||||
#include "orte/constants.h"
|
||||
#include "orte/runtime/orte_globals.h"
|
||||
#include "orte/util/proc_info.h"
|
||||
#include "orte/mca/ess/ess.h"
|
||||
#include "orte/mca/ess/base/base.h"
|
||||
#include "orte/mca/ess/alps/ess_alps.h"
|
||||
|
||||
#include <sys/syscall.h>
|
||||
|
||||
|
||||
/*
|
||||
* Instantiate the public struct with all of our public information
|
||||
* and pointers to our public functions in it
|
||||
*/
|
||||
orte_ess_base_component_t mca_ess_alps_component = {
|
||||
/* First, the mca_component_t struct containing meta information
|
||||
about the component itself */
|
||||
{
|
||||
ORTE_ESS_BASE_VERSION_3_0_0,
|
||||
|
||||
/* Component name and version */
|
||||
"alps",
|
||||
ORTE_MAJOR_VERSION,
|
||||
ORTE_MINOR_VERSION,
|
||||
ORTE_RELEASE_VERSION,
|
||||
|
||||
/* Component open and close functions */
|
||||
orte_ess_alps_component_open,
|
||||
orte_ess_alps_component_close,
|
||||
orte_ess_alps_component_query
|
||||
},
|
||||
{
|
||||
/* The component is not checkpoint ready */
|
||||
MCA_BASE_METADATA_PARAM_NONE
|
||||
}
|
||||
};
|
||||
|
||||
int
|
||||
orte_ess_alps_component_open(void)
|
||||
{
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
int orte_ess_alps_component_query(mca_base_module_t **module, int *priority)
|
||||
{
|
||||
int rc = ORTE_SUCCESS;
|
||||
const char proc_job_file[]="/proc/job";
|
||||
FILE *fd = NULL, *fd_task_is_app = NULL;
|
||||
char task_is_app_fname[PATH_MAX];
|
||||
|
||||
/*
|
||||
* don't use the alps ess component if an app proc
|
||||
*/
|
||||
|
||||
if (ORTE_PROC_IS_APP) {
|
||||
*priority = 0;
|
||||
*module = NULL;
|
||||
return ORTE_ERROR;
|
||||
}
|
||||
|
||||
/*
|
||||
* make sure we're in a Cray PAGG container, and that we are also on
|
||||
* a compute node (i.e. we are thought of as an application task by
|
||||
* the cray job kernel module - the thing that creates the PAGG)
|
||||
*/
|
||||
|
||||
/* disqualify ourselves if not running in a Cray PAGG container */
|
||||
fd = fopen(proc_job_file, "r");
|
||||
if (fd == NULL) {
|
||||
*priority = 0;
|
||||
*module = NULL;
|
||||
rc = ORTE_ERROR;
|
||||
} else {
|
||||
snprintf(task_is_app_fname,sizeof(task_is_app_fname),
|
||||
"/proc/self/task/%ld/task_is_app",syscall(SYS_gettid));
|
||||
fd_task_is_app = fopen(task_is_app_fname, "r");
|
||||
if (fd_task_is_app != NULL) { /* okay we're in a PAGG container,
|
||||
and we are an app task (not just a process
|
||||
running on a mom node, for example),
|
||||
so we should give cray pmi a shot. */
|
||||
*priority = 35; /* take precendence over base */
|
||||
*module = (mca_base_module_t *) &orte_ess_alps_module;
|
||||
fclose(fd_task_is_app);
|
||||
}
|
||||
fclose(fd);
|
||||
}
|
||||
|
||||
return rc;
|
||||
}
|
||||
|
||||
int
|
||||
orte_ess_alps_component_close(void)
|
||||
{
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
|
||||
229
orte/mca/ess/alps/ess_alps_module.c
Обычный файл
229
orte/mca/ess/alps/ess_alps_module.c
Обычный файл
@ -0,0 +1,229 @@
|
||||
/*
|
||||
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
||||
* University Research and Technology
|
||||
* Corporation. All rights reserved.
|
||||
* Copyright (c) 2004-2011 The University of Tennessee and The University
|
||||
* of Tennessee Research Foundation. All rights
|
||||
* reserved.
|
||||
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
||||
* University of Stuttgart. All rights reserved.
|
||||
* Copyright (c) 2004-2005 The Regents of the University of California.
|
||||
* All rights reserved.
|
||||
* Copyright (c) 2011 Cisco Systems, Inc. All rights reserved.
|
||||
* Copyright (c) 2011-2013 Los Alamos National Security, LLC.
|
||||
* All rights reserved.
|
||||
* $COPYRIGHT$
|
||||
*
|
||||
* Additional copyrights may follow
|
||||
*
|
||||
* $HEADER$
|
||||
*
|
||||
*/
|
||||
|
||||
#include "orte_config.h"
|
||||
#include "orte/constants.h"
|
||||
|
||||
#include "orte/util/show_help.h"
|
||||
#include "opal/util/argv.h"
|
||||
|
||||
#include "orte/util/proc_info.h"
|
||||
#include "orte/mca/errmgr/base/base.h"
|
||||
#include "orte/util/name_fns.h"
|
||||
#include "orte/util/nidmap.h"
|
||||
#include "orte/util/regex.h"
|
||||
#include "orte/runtime/orte_globals.h"
|
||||
|
||||
#include "orte/mca/ess/ess.h"
|
||||
#include "orte/mca/ess/base/base.h"
|
||||
#include "orte/mca/ess/alps/ess_alps.h"
|
||||
|
||||
#include <errno.h>
|
||||
|
||||
static int alps_set_name(void);
|
||||
static int rte_init(void);
|
||||
static int rte_finalize(void);
|
||||
|
||||
orte_ess_base_module_t orte_ess_alps_module = {
|
||||
rte_init,
|
||||
rte_finalize,
|
||||
orte_ess_base_app_abort,
|
||||
NULL /* ft_event */
|
||||
};
|
||||
|
||||
/* Local variables */
|
||||
static orte_vpid_t starting_vpid = 0;
|
||||
|
||||
|
||||
static int rte_init(void)
|
||||
{
|
||||
int ret, i;
|
||||
char *error = NULL;
|
||||
char **hosts = NULL;
|
||||
|
||||
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_framework.framework_output,
|
||||
"ess:alps in rte_init"));
|
||||
|
||||
/*
|
||||
* shouldn't have been able to open this ess component if
|
||||
* process is app proc
|
||||
*/
|
||||
|
||||
if (ORTE_PROC_IS_APP) {
|
||||
error = "mpi rank invoking alps rte_init";
|
||||
ret = ORTE_ERR_NOT_SUPPORTED;
|
||||
goto fn_fail;
|
||||
}
|
||||
|
||||
/* run the prolog */
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_std_prolog())) {
|
||||
error = "orte_ess_base_std_prolog";
|
||||
goto fn_fail;
|
||||
}
|
||||
|
||||
if (ORTE_SUCCESS != (ret = alps_set_name())) {
|
||||
error = "alps_set_name";
|
||||
goto fn_fail;
|
||||
}
|
||||
|
||||
/*
|
||||
* if I am a daemon, complete my setup using the
|
||||
* 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;
|
||||
}
|
||||
|
||||
/* find our host in the list */
|
||||
for (i=0; NULL != hosts[i]; i++) {
|
||||
if (0 == strncmp(hosts[i], orte_process_info.nodename,
|
||||
strlen(hosts[i]))) {
|
||||
/* correct our vpid - this is probably not necessary with aprun*/
|
||||
ORTE_PROC_MY_NAME->vpid = starting_vpid + i;
|
||||
OPAL_OUTPUT_VERBOSE((1, orte_ess_base_framework.framework_output,
|
||||
"ess:alps reset name to %s",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME)));
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_setup(hosts))) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
error = "orte_ess_base_orted_setup";
|
||||
goto fn_fail;
|
||||
}
|
||||
if (NULL != hosts) {
|
||||
opal_argv_free(hosts);
|
||||
}
|
||||
|
||||
/*
|
||||
* now synchronize with aprun.
|
||||
*/
|
||||
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_alps_sync_start())) {
|
||||
error = "orte_ess_alps_sync";
|
||||
goto fn_fail;
|
||||
}
|
||||
|
||||
ret = ORTE_SUCCESS;
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
if (ORTE_PROC_IS_TOOL) {
|
||||
/* otherwise, if I am a tool proc, use that procedure */
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_tool_setup())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
error = "orte_ess_base_tool_setup";
|
||||
goto fn_fail;
|
||||
}
|
||||
/* as a tool, I don't need a nidmap - so just return now */
|
||||
ret = ORTE_SUCCESS;
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
fn_exit:
|
||||
return ret;
|
||||
|
||||
fn_fail:
|
||||
if (ORTE_ERR_SILENT != ret && !orte_report_silent_errors) {
|
||||
orte_show_help("help-orte-runtime.txt",
|
||||
"orte_init:startup:internal-failure",
|
||||
true, error, ORTE_ERROR_NAME(ret), ret);
|
||||
}
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
static int rte_finalize(void)
|
||||
{
|
||||
int ret = ORTE_SUCCESS;
|
||||
|
||||
/* if I am a daemon, finalize using the default procedure */
|
||||
if (ORTE_PROC_IS_DAEMON) {
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_orted_finalize())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
/* notify alps that we're done */
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_alps_sync_complete())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
}
|
||||
|
||||
} else if (ORTE_PROC_IS_TOOL) {
|
||||
/* otherwise, if I am a tool proc, use that procedure */
|
||||
if (ORTE_SUCCESS != (ret = orte_ess_base_tool_finalize())) {
|
||||
ORTE_ERROR_LOG(ret);
|
||||
}
|
||||
}
|
||||
|
||||
fn_exit:
|
||||
return ret;
|
||||
}
|
||||
|
||||
static int alps_set_name(void)
|
||||
{
|
||||
int rc;
|
||||
int rank;
|
||||
orte_jobid_t jobid;
|
||||
|
||||
if (NULL == orte_ess_base_jobid) {
|
||||
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
|
||||
return ORTE_ERR_NOT_FOUND;
|
||||
}
|
||||
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_jobid(&jobid, orte_ess_base_jobid))) {
|
||||
ORTE_ERROR_LOG(rc);
|
||||
return rc;
|
||||
}
|
||||
|
||||
if (NULL == orte_ess_base_vpid) {
|
||||
ORTE_ERROR_LOG(ORTE_ERR_NOT_FOUND);
|
||||
return ORTE_ERR_NOT_FOUND;
|
||||
}
|
||||
if (ORTE_SUCCESS != (rc = orte_util_convert_string_to_vpid(&starting_vpid,
|
||||
orte_ess_base_vpid))) {
|
||||
ORTE_ERROR_LOG(rc);
|
||||
return(rc);
|
||||
}
|
||||
|
||||
ORTE_PROC_MY_NAME->jobid = jobid;
|
||||
|
||||
if (ORTE_SUCCESS != (rc = orte_ess_alps_get_first_rank_on_node(&rank))) {
|
||||
ORTE_ERROR_LOG(rc);
|
||||
return(rc);
|
||||
}
|
||||
|
||||
ORTE_PROC_MY_NAME->vpid = (orte_vpid_t)rank + starting_vpid;
|
||||
|
||||
/* get the num procs as provided in the cmd line param */
|
||||
if (ORTE_SUCCESS != (rc = orte_ess_env_get())) {
|
||||
ORTE_ERROR_LOG(rc);
|
||||
return rc;
|
||||
}
|
||||
|
||||
return ORTE_SUCCESS;
|
||||
}
|
||||
241
orte/mca/ess/alps/ess_alps_utils.c
Обычный файл
241
orte/mca/ess/alps/ess_alps_utils.c
Обычный файл
@ -0,0 +1,241 @@
|
||||
/*
|
||||
* Copyright (c) 2004-2005 The Trustees of Indiana University and Indiana
|
||||
* University Research and Technology
|
||||
* Corporation. All rights reserved.
|
||||
* Copyright (c) 2004-2011 The University of Tennessee and The University
|
||||
* of Tennessee Research Foundation. All rights
|
||||
* reserved.
|
||||
* Copyright (c) 2004-2005 High Performance Computing Center Stuttgart,
|
||||
* University of Stuttgart. All rights reserved.
|
||||
* Copyright (c) 2004-2005 The Regents of the University of California.
|
||||
* All rights reserved.
|
||||
* Copyright (c) 2011 Cisco Systems, Inc. All rights reserved.
|
||||
* Copyright (c) 2011-2013 Los Alamos National Security, LLC.
|
||||
* All rights reserved.
|
||||
* $COPYRIGHT$
|
||||
*
|
||||
* Additional copyrights may follow
|
||||
*
|
||||
* $HEADER$
|
||||
*
|
||||
*/
|
||||
|
||||
#include "orte_config.h"
|
||||
#include "orte/constants.h"
|
||||
|
||||
#include "orte/util/show_help.h"
|
||||
#include "opal/util/argv.h"
|
||||
|
||||
#include "orte/util/proc_info.h"
|
||||
#include "orte/mca/errmgr/base/base.h"
|
||||
#include "orte/util/name_fns.h"
|
||||
#include "orte/util/nidmap.h"
|
||||
#include "orte/util/regex.h"
|
||||
#include "orte/runtime/orte_globals.h"
|
||||
|
||||
#include "orte/mca/ess/ess.h"
|
||||
#include "orte/mca/ess/base/base.h"
|
||||
#include "orte/mca/ess/alps/ess_alps.h"
|
||||
|
||||
/*
|
||||
* use the Alps placement file to obtain
|
||||
* the global rank of the "first" local rank
|
||||
* on the node.
|
||||
*/
|
||||
|
||||
int
|
||||
orte_ess_alps_get_first_rank_on_node(int *first_rank)
|
||||
{
|
||||
int alps_status = 0;
|
||||
uint64_t apid;
|
||||
size_t alps_count;
|
||||
int ret = ORTE_SUCCESS;
|
||||
int lli_ret = 0, place_ret;
|
||||
alpsAppLayout_t orted_layout;
|
||||
|
||||
if (first_rank == NULL) {
|
||||
ret = ORTE_ERR_BAD_PARAM;
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
/*
|
||||
* First get our apid
|
||||
*/
|
||||
|
||||
lli_ret = alps_app_lli_lock();
|
||||
if (0 != ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_lock returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
lli_ret = alps_app_lli_put_request(ALPS_APP_LLI_ALPS_REQ_APID, NULL, 0);
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != lli_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_put_request - APID returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
goto fn_exit_w_lock;
|
||||
}
|
||||
|
||||
lli_ret = alps_app_lli_get_response (&alps_status, &alps_count);
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != alps_status) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_get_response returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), alps_status));
|
||||
ret = ORTE_ERR_FILE_READ_FAILURE;
|
||||
goto fn_exit_w_lock;
|
||||
}
|
||||
|
||||
lli_ret = alps_app_lli_get_response_bytes (&apid, sizeof(apid));
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != lli_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_get_response_bytes returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_READ_FAILURE;
|
||||
goto fn_exit_w_lock;
|
||||
}
|
||||
|
||||
place_ret = alps_get_placement_info(apid,
|
||||
&orted_layout,
|
||||
NULL,
|
||||
NULL,
|
||||
NULL,
|
||||
NULL,
|
||||
NULL,
|
||||
NULL,
|
||||
NULL,
|
||||
NULL,
|
||||
NULL);
|
||||
if (1 != place_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_get_placement_info returned %d (%s)",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), place_ret, strerror(errno)));
|
||||
ret = ORTE_ERROR;
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
OPAL_OUTPUT_VERBOSE((2, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_get_placement_info returned %d first pe on node is %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), place_ret, orted_layout.firstPe));
|
||||
*first_rank = orted_layout.firstPe;
|
||||
|
||||
fn_exit_w_lock:
|
||||
lli_ret = alps_app_lli_unlock();
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != lli_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_unlock returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
}
|
||||
|
||||
fn_exit:
|
||||
return ret;
|
||||
}
|
||||
|
||||
/*
|
||||
* Function to check in with apshepherd to say we are a parallel application
|
||||
*/
|
||||
int
|
||||
orte_ess_alps_sync_start(void)
|
||||
{
|
||||
int ret = ORTE_SUCCESS;
|
||||
int lli_ret = 0;
|
||||
int alps_status = 0;
|
||||
size_t alps_count;
|
||||
|
||||
lli_ret = alps_app_lli_lock();
|
||||
if (0 != ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_lock returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
lli_ret = alps_app_lli_put_request(ALPS_APP_LLI_ALPS_REQ_START, NULL, 0);
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != lli_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_put_request returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
goto fn_exit_w_lock;
|
||||
}
|
||||
|
||||
lli_ret = alps_app_lli_get_response (&alps_status, &alps_count);
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != alps_status) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_get_response returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), alps_status));
|
||||
ret = ORTE_ERR_FILE_READ_FAILURE;
|
||||
goto fn_exit_w_lock;
|
||||
}
|
||||
|
||||
fn_exit_w_lock:
|
||||
lli_ret = alps_app_lli_unlock();
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != lli_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_unlock returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
}
|
||||
|
||||
fn_exit:
|
||||
return ret;
|
||||
}
|
||||
|
||||
/*
|
||||
* Function to check in with apshepherd to say we are a parallel application
|
||||
*/
|
||||
|
||||
int
|
||||
orte_ess_alps_sync_complete(void)
|
||||
{
|
||||
int ret = ORTE_SUCCESS;
|
||||
int lli_ret = 0;
|
||||
int alps_status = 0;
|
||||
size_t alps_count;
|
||||
|
||||
lli_ret = alps_app_lli_lock();
|
||||
if (0 != ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_lock returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
goto fn_exit;
|
||||
}
|
||||
|
||||
lli_ret = alps_app_lli_put_request(ALPS_APP_LLI_ALPS_REQ_EXITING, NULL, 0);
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != lli_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_put_request returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
goto fn_exit_w_lock;
|
||||
}
|
||||
|
||||
lli_ret = alps_app_lli_get_response (&alps_status, &alps_count);
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != alps_status) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_get_response returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), alps_status));
|
||||
ret = ORTE_ERR_FILE_READ_FAILURE;
|
||||
goto fn_exit_w_lock;
|
||||
}
|
||||
|
||||
fn_exit_w_lock:
|
||||
lli_ret = alps_app_lli_unlock();
|
||||
if (ALPS_APP_LLI_ALPS_STAT_OK != lli_ret) {
|
||||
OPAL_OUTPUT_VERBOSE((20, orte_ess_base_framework.framework_output,
|
||||
"%s ess:alps: alps_app_lli_unlock returned %d",
|
||||
ORTE_NAME_PRINT(ORTE_PROC_MY_NAME), lli_ret));
|
||||
ret = ORTE_ERR_FILE_WRITE_FAILURE;
|
||||
}
|
||||
|
||||
fn_exit:
|
||||
return ret;
|
||||
}
|
||||
|
||||
|
||||
Загрузка…
x
Ссылка в новой задаче
Block a user