int orte_init(int* pargc, char*** pargv, orte_proc_type_t flags)
{
int ret;
char *error = NULL;
if (0 < orte_initialized) {
/* track number of times we have been called */
orte_initialized++;
return ORTE_SUCCESS;
}
orte_initialized++;
/* initialize the opal layer */
if (ORTE_SUCCESS != (ret = opal_init(pargc, pargv))) {
error = "opal_init";
goto error;
}
/* ensure we know the type of proc for when we finalize */
orte_process_info.proc_type = flags;
/* setup the locks */
if (ORTE_SUCCESS != (ret = orte_locks_init())) {
error = "orte_locks_init";
goto error;
}
/* Register all MCA Params */
if (ORTE_SUCCESS != (ret = orte_register_params())) {
error = "orte_register_params";
goto error;
}
/* setup the orte_show_help system */
if (ORTE_SUCCESS != (ret = orte_show_help_init())) {
error = "opal_output_init";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORTE", ORTE_ERR_BASE, ORTE_ERR_MAX, orte_err2str);
/* Ensure the rest of the process info structure is initialized */
if (ORTE_SUCCESS != (ret = orte_proc_info())) {
error = "orte_proc_info";
goto error;
}
/* open the ESS and select the correct module for this environment */
if (ORTE_SUCCESS != (ret = orte_ess_base_open())) {
error = "orte_ess_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess_base_select())) {
error = "orte_ess_base_select";
goto error;
}
if (ORTE_PROC_IS_APP) {
#if !ORTE_DISABLE_FULL_SUPPORT && ORTE_ENABLE_PROGRESS_THREADS
#if OPAL_EVENT_HAVE_THREAD_SUPPORT
/* get a separate orte event base */
orte_event_base = opal_event_base_create();
/* setup the finalize event - we'll need it
* to break the thread out of the event lib
* when we want to stop it
*/
opal_event_set(orte_event_base, &orte_finalize_event, -1, OPAL_EV_WRITE, ignore_callback, NULL);
opal_event_set_priority(&orte_finalize_event, ORTE_ERROR_PRI);
#if 0
{
/* seems strange, but wake us up once a second just so we can check for new events */
opal_event_t *ev;
struct timeval tv = {1,0};
ev = opal_event_alloc();
opal_event_evtimer_set(orte_event_base,
ev, ignore_callback, ev);
opal_event_set_priority(ev, ORTE_INFO_PRI);
opal_event_evtimer_add(ev, &tv);
}
#endif
/* construct the thread object */
OBJ_CONSTRUCT(&orte_progress_thread, opal_thread_t);
/* fork off a thread to progress it */
orte_progress_thread.t_run = orte_progress_thread_engine;
if (OPAL_SUCCESS != (ret = opal_thread_start(&orte_progress_thread))) {
error = "orte progress thread start";
goto error;
}
#else
error = "event thread support is not configured";
ret = ORTE_ERROR;
goto error;
#endif
#else
/* set the event base to the opal one */
orte_event_base = opal_event_base;
#endif
} else {
/* set the event base to the opal one */
orte_event_base = opal_event_base;
}
/* initialize the RTE for this environment */
if (ORTE_SUCCESS != (ret = orte_ess.init())) {
error = "orte_ess_init";
goto error;
}
/* All done */
return ORTE_SUCCESS;
error:
if (ORTE_ERR_SILENT != ret) {
orte_show_help("help-orte-runtime",
"orte_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}
int
opal_init_util(int* pargc, char*** pargv)
{
int ret;
char *error = NULL;
if( ++opal_util_initialized != 1 ) {
if( opal_util_initialized < 1 ) {
return OPAL_ERROR;
}
return OPAL_SUCCESS;
}
/* JMS See note in runtime/opal.h -- this is temporary; to be
replaced with real hwloc information soon (in trunk/v1.5 and
beyond, only). This *used* to be a #define, so it's important
to define it very early. */
opal_cache_line_size = 128;
/* initialize the memory allocator */
opal_malloc_init();
/* initialize the output system */
opal_output_init();
/* initialize install dirs code */
if (OPAL_SUCCESS != (ret = opal_installdirs_base_open())) {
fprintf(stderr, "opal_installdirs_base_open() failed -- process will likely abort (%s:%d, returned %d instead of OPAL_INIT)\n",
__FILE__, __LINE__, ret);
return ret;
}
/* initialize the help system */
opal_show_help_init();
/* register handler for errnum -> string converstion */
if (OPAL_SUCCESS !=
(ret = opal_error_register("OPAL",
OPAL_ERR_BASE, OPAL_ERR_MAX, opal_err2str))) {
error = "opal_error_register";
goto return_error;
}
/* init the trace function */
opal_trace_init();
/* keyval lex-based parser */
if (OPAL_SUCCESS != (ret = opal_util_keyval_parse_init())) {
error = "opal_util_keyval_parse_init";
goto return_error;
}
if (OPAL_SUCCESS != (ret = opal_net_init())) {
error = "opal_net_init";
goto return_error;
}
/* Setup the parameter system */
if (OPAL_SUCCESS != (ret = mca_base_param_init())) {
error = "mca_base_param_init";
goto return_error;
}
/* register params for opal */
if (OPAL_SUCCESS != (ret = opal_register_params())) {
error = "opal_register_params";
goto return_error;
}
/* pretty-print stack handlers */
if (OPAL_SUCCESS != (ret = opal_util_register_stackhandlers())) {
error = "opal_util_register_stackhandlers";
goto return_error;
}
if (OPAL_SUCCESS != (ret = opal_util_init_sys_limits())) {
error = "opal_util_init_sys_limits";
goto return_error;
}
/* initialize the datatype engine */
if (OPAL_SUCCESS != (ret = opal_datatype_init ())) {
error = "opal_datatype_init";
goto return_error;
}
/* Initialize the data storage service. */
if (OPAL_SUCCESS != (ret = opal_dss_open())) {
error = "opal_dss_open";
goto return_error;
}
return OPAL_SUCCESS;
return_error:
opal_show_help( "help-opal-runtime.txt",
"opal_init:startup:internal-failure", true,
error, ret );
return ret;
}
int orcm_init(orcm_proc_type_t flags)
{
int ret;
char *error;
int spin;
opal_output_stream_t lds;
if (0 < orcm_initialized) {
/* track number of times we have been called */
orcm_initialized++;
return ORCM_SUCCESS;
}
orcm_initialized++;
if (NULL != getenv("ORCM_MCA_spin")) {
spin = 1;
/* spin until a debugger can attach */
while (0 != spin) {
ret = 0;
while (ret < 10000) {
ret++;
};
}
}
/* initialize the opal layer */
if (ORTE_SUCCESS != (ret = opal_init(NULL, NULL))) {
error = "opal_init";
goto error;
}
orcm_debug_verbosity = -1;
(void) mca_base_var_register ("orcm", "orcm", NULL, "debug_verbose",
"Verbosity level for ORCM debug messages (default: 1)",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_ALL,
&orcm_debug_verbosity);
if (0 <= orcm_debug_verbosity) {
/* get a debug output channel */
OBJ_CONSTRUCT(&lds, opal_output_stream_t);
lds.lds_want_stdout = true;
orcm_debug_output = opal_output_open(&lds);
OBJ_DESTRUCT(&lds);
/* set the verbosity */
opal_output_set_verbosity(orcm_debug_output, orcm_debug_verbosity);
}
/* ensure we know the type of proc for when we finalize */
orte_process_info.proc_type = flags;
/* setup the locks */
if (ORTE_SUCCESS != (ret = orte_locks_init())) {
error = "orte_locks_init";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORTE", ORTE_ERR_BASE, ORTE_ERR_MAX, orte_err2str);
/* Ensure the rest of the process info structure is initialized */
if (ORTE_SUCCESS != (ret = orte_proc_info())) {
error = "orte_proc_info";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORCM", ORCM_ERR_BASE, ORCM_ERR_MAX, orcm_err2str);
/* register handler for attr key -> string conversion */
if (ORTE_SUCCESS != (ret = orte_attr_register("orcm", ORCM_ATTR_KEY_BASE, ORCM_ATTR_KEY_MAX,
orcm_attr_key_print))) {
error = "register attr print";
goto error;
}
/* we don't need a progress thread as all our tools loop inside themselves,
* so define orte_event_base to be the base opal_event_base
*/
orte_event_base = opal_sync_event_base;
/* setup the globals */
orcm_clusters = OBJ_NEW(opal_list_t);
orcm_schedulers = OBJ_NEW(opal_list_t);
/* everyone must open the cfgi framework */
if (ORCM_SUCCESS != (ret = mca_base_framework_open(&orcm_cfgi_base_framework, 0))) {
error = "orcm_cfgi_base_open";
goto error;
}
if (ORCM_SUCCESS != (ret = orcm_cfgi_base_select())) {
error = "orcm_cfgi_select";
goto error;
}
/* everyone must open the sst framework */
if (ORCM_SUCCESS != (ret = mca_base_framework_open(&orcm_sst_base_framework, 0))) {
error = "orcm_sst_base_open";
goto error;
}
if (ORCM_SUCCESS != (ret = orcm_sst_base_select())) {
error = "orcm_sst_select";
goto error;
}
/* open the ESS and select the correct module for this environment - the
* orcm module is basically a no-op, but we need the framework defined
* as other parts of ORTE will want to call it
*/
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_ess_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess_base_select())) {
error = "orte_ess_base_select";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess.init())) {
error = "orte_ess_init";
goto error;
}
/* initialize us - we will register the ORTE-level MCA params in there */
if (ORTE_SUCCESS != (ret = orcm_sst.init())) {
error = "orte_init";
goto error;
}
/* setup the orte_show_help system - don't do this until the
* end as otherwise show_help messages won't appear
*/
if (ORTE_SUCCESS != (ret = orte_show_help_init())) {
error = "opal_output_init";
goto error;
}
/* initialize orcm datatype support */
if (ORCM_SUCCESS != (ret = orcm_dt_init())) {
error = "orcm_dt_init";
goto error;
}
/* flag that orte is initialized so things can work */
orte_initialized = true;
return ORCM_SUCCESS;
error:
if (ORCM_ERR_SILENT != ret) {
opal_show_help("help-orcm-runtime.txt",
"orcm_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}
int opal_init_test(void)
{
int ret;
char *error;
/* initialize the memory allocator */
opal_malloc_init();
/* initialize the output system */
opal_output_init();
/* initialize install dirs code */
if (OPAL_SUCCESS != (ret = mca_base_framework_open(&opal_installdirs_base_framework, 0))) {
fprintf(stderr, "opal_installdirs_base_open() failed -- process will likely abort (%s:%d, returned %d instead of OPAL_SUCCESS)\n",
__FILE__, __LINE__, ret);
return ret;
}
/* initialize the help system */
opal_show_help_init();
/* register handler for errnum -> string converstion */
if (OPAL_SUCCESS !=
(ret = opal_error_register("OPAL",
OPAL_ERR_BASE, OPAL_ERR_MAX, opal_err2str))) {
error = "opal_error_register";
goto return_error;
}
/* keyval lex-based parser */
if (OPAL_SUCCESS != (ret = opal_util_keyval_parse_init())) {
error = "opal_util_keyval_parse_init";
goto return_error;
}
if (OPAL_SUCCESS != (ret = opal_net_init())) {
error = "opal_net_init";
goto return_error;
}
/* Setup the parameter system */
if (OPAL_SUCCESS != (ret = mca_base_var_init())) {
error = "mca_base_var_init";
goto return_error;
}
/* register params for opal */
if (OPAL_SUCCESS != (ret = opal_register_params())) {
error = "opal_register_params";
goto return_error;
}
/* pretty-print stack handlers */
if (OPAL_SUCCESS != (ret = opal_util_register_stackhandlers())) {
error = "opal_util_register_stackhandlers";
goto return_error;
}
/* Initialize the data storage service. */
if (OPAL_SUCCESS != (ret = opal_dss_open())) {
error = "opal_dss_open";
goto return_error;
}
/* initialize the mca */
if (OPAL_SUCCESS != (ret = mca_base_open())) {
error = "mca_base_open";
goto return_error;
}
if (OPAL_SUCCESS != (ret = mca_base_framework_register(&opal_event_base_framework, 0))) {
error = "opal_event_register";
goto return_error;
}
return OPAL_SUCCESS;
return_error:
opal_show_help( "help-opal-runtime.txt",
"opal_init:startup:internal-failure", true,
error, ret );
return ret;
}
int
opal_init_util(void)
{
int ret;
char *error = NULL;
if( ++opal_initialized != 1 ) {
if( opal_initialized < 1 ) {
return OPAL_ERROR;
}
return OPAL_SUCCESS;
}
/* initialize the memory allocator */
opal_malloc_init();
/* initialize the output system */
opal_output_init();
/* register handler for errnum -> string converstion */
if (OPAL_SUCCESS != (ret = opal_error_register("OPAL",
OPAL_ERR_BASE, OPAL_ERR_MAX, opal_err2str))) {
error = "opal_error_register";
goto return_error;
}
/* initialize install dirs code */
if (OPAL_SUCCESS != (ret = opal_installdirs_base_open())) {
fprintf(stderr, "opal_installdirs_base_open() failed -- process will likely abort (%s:%d, returned %d instead of OPAL_INIT)\n",
__FILE__, __LINE__, ret);
return ret;
}
/* init the trace function */
opal_trace_init();
/* keyval lex-based parser */
if (OPAL_SUCCESS != (ret = opal_util_keyval_parse_init())) {
error = "opal_util_keyval_parse_init";
goto return_error;
}
/* Setup the parameter system */
if (OPAL_SUCCESS != (ret = mca_base_param_init())) {
error = "mca_base_param_init";
goto return_error;
}
/* register params for opal */
if (OPAL_SUCCESS != opal_register_params()) {
error = "opal_register_params";
goto return_error;
}
/* pretty-print stack handlers */
if (OPAL_SUCCESS != (ret = opal_util_register_stackhandlers ())) {
error = "util_register_stackhandlers() failed";
goto return_error;
}
return OPAL_SUCCESS;
return_error:
opal_show_help( "help-opal-runtime",
"opal_init:startup:internal-failure", true,
error, ret );
return ret;
}
int
opal_init_util(int* pargc, char*** pargv)
{
int ret;
char *error = NULL;
char hostname[512];
if( ++opal_util_initialized != 1 ) {
if( opal_util_initialized < 1 ) {
return OPAL_ERROR;
}
return OPAL_SUCCESS;
}
#if OPAL_NO_LIB_DESTRUCTOR
if (opal_init_called) {
/* can't use show_help here */
fprintf (stderr, "opal_init_util: attempted to initialize after finalize without compiler "
"support for either __attribute__(destructor) or linker support for -fini -- process "
"will likely abort\n");
return OPAL_ERR_NOT_SUPPORTED;
}
#endif
opal_init_called = true;
/* set the nodename right away so anyone who needs it has it. Note
* that we don't bother with fqdn and prefix issues here - we let
* the RTE later replace this with a modified name if the user
* requests it */
gethostname(hostname, 512);
opal_process_info.nodename = strdup(hostname);
/* initialize the memory allocator */
opal_malloc_init();
/* initialize the output system */
opal_output_init();
/* initialize install dirs code */
if (OPAL_SUCCESS != (ret = mca_base_framework_open(&opal_installdirs_base_framework, 0))) {
fprintf(stderr, "opal_installdirs_base_open() failed -- process will likely abort (%s:%d, returned %d instead of OPAL_SUCCESS)\n",
__FILE__, __LINE__, ret);
return ret;
}
/* initialize the help system */
opal_show_help_init();
/* register handler for errnum -> string converstion */
if (OPAL_SUCCESS !=
(ret = opal_error_register("OPAL",
OPAL_ERR_BASE, OPAL_ERR_MAX, opal_err2str))) {
error = "opal_error_register";
goto return_error;
}
/* keyval lex-based parser */
if (OPAL_SUCCESS != (ret = opal_util_keyval_parse_init())) {
error = "opal_util_keyval_parse_init";
goto return_error;
}
/* Setup the parameter system */
if (OPAL_SUCCESS != (ret = mca_base_var_init())) {
error = "mca_base_var_init";
goto return_error;
}
/* register params for opal */
if (OPAL_SUCCESS != (ret = opal_register_params())) {
error = "opal_register_params";
goto return_error;
}
if (OPAL_SUCCESS != (ret = opal_net_init())) {
error = "opal_net_init";
goto return_error;
}
/* pretty-print stack handlers */
if (OPAL_SUCCESS != (ret = opal_util_register_stackhandlers())) {
error = "opal_util_register_stackhandlers";
goto return_error;
}
/* set system resource limits - internally protected against
* doing so twice in cases where the launch agent did it for us
*/
if (OPAL_SUCCESS != (ret = opal_util_init_sys_limits(&error))) {
opal_show_help("help-opal-runtime.txt",
"opal_init:syslimit", false,
error);
return OPAL_ERR_SILENT;
}
/* initialize the arch string */
if (OPAL_SUCCESS != (ret = opal_arch_init ())) {
error = "opal_arch_init";
goto return_error;
}
/* initialize the datatype engine */
if (OPAL_SUCCESS != (ret = opal_datatype_init ())) {
error = "opal_datatype_init";
goto return_error;
}
/* Initialize the data storage service. */
if (OPAL_SUCCESS != (ret = opal_dss_open())) {
error = "opal_dss_open";
goto return_error;
}
/* initialize the mca */
if (OPAL_SUCCESS != (ret = mca_base_open())) {
error = "mca_base_open";
goto return_error;
}
return OPAL_SUCCESS;
return_error:
if (OPAL_ERR_SILENT != ret) {
opal_show_help( "help-opal-runtime.txt",
"opal_init:startup:internal-failure", true,
error, ret );
}
return ret;
}
int
opal_init_util(int* pargc, char*** pargv)
{
int ret;
char *error = NULL;
if( ++opal_util_initialized != 1 ) {
if( opal_util_initialized < 1 ) {
return OPAL_ERROR;
}
return OPAL_SUCCESS;
}
/* initialize the memory allocator */
opal_malloc_init();
/* initialize the output system */
opal_output_init();
/* initialize install dirs code */
if (OPAL_SUCCESS != (ret = mca_base_framework_open(&opal_installdirs_base_framework, 0))) {
fprintf(stderr, "opal_installdirs_base_open() failed -- process will likely abort (%s:%d, returned %d instead of OPAL_SUCCESS)\n",
__FILE__, __LINE__, ret);
return ret;
}
/* initialize the help system */
opal_show_help_init();
/* register handler for errnum -> string converstion */
if (OPAL_SUCCESS !=
(ret = opal_error_register("OPAL",
OPAL_ERR_BASE, OPAL_ERR_MAX, opal_err2str))) {
error = "opal_error_register";
goto return_error;
}
/* keyval lex-based parser */
if (OPAL_SUCCESS != (ret = opal_util_keyval_parse_init())) {
error = "opal_util_keyval_parse_init";
goto return_error;
}
if (OPAL_SUCCESS != (ret = opal_net_init())) {
error = "opal_net_init";
goto return_error;
}
/* Setup the parameter system */
if (OPAL_SUCCESS != (ret = mca_base_var_init())) {
error = "mca_base_var_init";
goto return_error;
}
/* register params for opal */
if (OPAL_SUCCESS != (ret = opal_register_params())) {
error = "opal_register_params";
goto return_error;
}
/* pretty-print stack handlers */
if (OPAL_SUCCESS != (ret = opal_util_register_stackhandlers())) {
error = "opal_util_register_stackhandlers";
goto return_error;
}
/* set system resource limits - internally protected against
* doing so twice in cases where the launch agent did it for us
*/
if (OPAL_SUCCESS != (ret = opal_util_init_sys_limits(&error))) {
opal_show_help("help-opal-runtime.txt",
"opal_init:syslimit", false,
error);
return OPAL_ERR_SILENT;
}
/* initialize the arch string */
if (OPAL_SUCCESS != (ret = opal_arch_init ())) {
error = "opal_arch_init";
goto return_error;
}
/* initialize the datatype engine */
if (OPAL_SUCCESS != (ret = opal_datatype_init ())) {
error = "opal_datatype_init";
goto return_error;
}
/* Initialize the data storage service. */
if (OPAL_SUCCESS != (ret = opal_dss_open())) {
error = "opal_dss_open";
goto return_error;
}
return OPAL_SUCCESS;
return_error:
if (OPAL_ERR_SILENT != ret) {
opal_show_help( "help-opal-runtime.txt",
"opal_init:startup:internal-failure", true,
error, ret );
}
return ret;
}
int orte_init(int* pargc, char*** pargv, orte_proc_type_t flags)
{
int ret;
char *error = NULL;
if (orte_initialized) {
return ORTE_SUCCESS;
}
/* initialize the opal layer */
if (ORTE_SUCCESS != (ret = opal_init(pargc, pargv))) {
ORTE_ERROR_LOG(ret);
return ret;
}
/* ensure we know the type of proc for when we finalize */
orte_process_info.proc_type = flags;
/* setup the locks */
if (ORTE_SUCCESS != (ret = orte_locks_init())) {
error = "orte_locks_init";
goto error;
}
/* Register all MCA Params */
if (ORTE_SUCCESS != (ret = orte_register_params())) {
error = "orte_register_params";
goto error;
}
/* setup the orte_show_help system */
if (ORTE_SUCCESS != (ret = orte_show_help_init())) {
ORTE_ERROR_LOG(ret);
error = "opal_output_init";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORTE", ORTE_ERR_BASE, ORTE_ERR_MAX, orte_err2str);
/* Ensure the rest of the process info structure is initialized */
if (ORTE_SUCCESS != (ret = orte_proc_info())) {
error = "orte_proc_info";
goto error;
}
/* open the ESS and select the correct module for this environment */
if (ORTE_SUCCESS != (ret = orte_ess_base_open())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_select";
goto error;
}
/* initialize the RTE for this environment */
if (ORTE_SUCCESS != (ret = orte_ess.init())) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_set_name";
goto error;
}
/* All done */
orte_initialized = true;
return ORTE_SUCCESS;
error:
if (ORTE_ERR_SILENT != OPAL_SOS_GET_ERROR_CODE(ret)) {
orte_show_help("help-orte-runtime",
"orte_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}
int orte_init(int* pargc, char*** pargv, orte_proc_type_t flags)
{
int ret;
char *error = NULL;
if (0 < orte_initialized) {
/* track number of times we have been called */
orte_initialized++;
return ORTE_SUCCESS;
}
orte_initialized++;
/* initialize the opal layer */
if (ORTE_SUCCESS != (ret = opal_init(pargc, pargv))) {
error = "opal_init";
goto error;
}
/* Convince OPAL to use our naming scheme */
opal_process_name_print = _process_name_print_for_opal;
opal_vpid_print = _vpid_print_for_opal;
opal_jobid_print = _jobid_print_for_opal;
opal_compare_proc = _process_name_compare;
opal_convert_string_to_process_name = _convert_string_to_process_name;
/* ensure we know the type of proc for when we finalize */
orte_process_info.proc_type = flags;
/* setup the locks */
if (ORTE_SUCCESS != (ret = orte_locks_init())) {
error = "orte_locks_init";
goto error;
}
/* Register all MCA Params */
if (ORTE_SUCCESS != (ret = orte_register_params())) {
error = "orte_register_params";
goto error;
}
/* setup the orte_show_help system */
if (ORTE_SUCCESS != (ret = orte_show_help_init())) {
error = "opal_output_init";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORTE", ORTE_ERR_BASE, ORTE_ERR_MAX, orte_err2str);
/* Ensure the rest of the process info structure is initialized */
if (ORTE_SUCCESS != (ret = orte_proc_info())) {
error = "orte_proc_info";
goto error;
}
/* we may have modified the local nodename according to
* request to retain/strip the FQDN and prefix, so update
* it here. The OPAL layer will strdup the hostname, so
* we have to free it first to avoid a memory leak */
if (NULL != opal_process_info.nodename) {
free(opal_process_info.nodename);
}
/* opal_finalize_util will call free on this pointer so set from strdup */
opal_process_info.nodename = strdup (orte_process_info.nodename);
/* setup the dstore framework */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&opal_dstore_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "opal_dstore_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = opal_dstore_base_select())) {
ORTE_ERROR_LOG(ret);
error = "opal_dstore_base_select";
goto error;
}
/* create the handle */
if (0 > (opal_dstore_internal = opal_dstore.open("INTERNAL", "hash", NULL))) {
error = "opal dstore internal";
ret = ORTE_ERR_FATAL;
goto error;
}
if (ORTE_PROC_IS_APP) {
if (0 > (opal_dstore_modex = opal_dstore.open("MODEX", "sm,hash", NULL))) {
error = "opal dstore modex";
ret = ORTE_ERR_FATAL;
goto error;
}
}
if (ORTE_PROC_IS_DAEMON || ORTE_PROC_IS_HNP) {
/* let the pmix server register params */
pmix_server_register();
}
/* open the ESS and select the correct module for this environment */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_ess_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess_base_select())) {
error = "orte_ess_base_select";
goto error;
}
if (!ORTE_PROC_IS_APP) {
/* ORTE tools "block" in their own loop over the event
* base, so no progress thread is required - apps will
* start their progress thread in ess_base_std_app.c
* at the appropriate point
*/
orte_event_base = opal_event_base;
}
/* initialize the RTE for this environment */
if (ORTE_SUCCESS != (ret = orte_ess.init())) {
error = "orte_ess_init";
goto error;
}
/* set the remaining opal_process_info fields. Note that
* the OPAL layer will have initialized these to NULL, and
* anyone between us would not have strdup'd the string, so
* we cannot free it here */
opal_process_info.job_session_dir = orte_process_info.job_session_dir;
opal_process_info.proc_session_dir = orte_process_info.proc_session_dir;
opal_process_info.num_local_peers = (int32_t)orte_process_info.num_local_peers;
opal_process_info.my_local_rank = (int32_t)orte_process_info.my_local_rank;
#if OPAL_HAVE_HWLOC
opal_process_info.cpuset = orte_process_info.cpuset;
#endif /* OPAL_HAVE_HWLOC */
#if OPAL_ENABLE_TIMING
opal_timing_set_jobid(ORTE_NAME_PRINT(ORTE_PROC_MY_NAME));
#endif
/* All done */
return ORTE_SUCCESS;
error:
if (ORTE_ERR_SILENT != ret) {
orte_show_help("help-orte-runtime",
"orte_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}
int orcm_init(orcm_proc_type_t flags)
{
int ret;
char *error, *envar;
int spin;
opal_output_stream_t lds;
if (0 < orcm_initialized) {
/* track number of times we have been called */
orcm_initialized++;
return ORCM_SUCCESS;
}
orcm_initialized++;
if (NULL != getenv("ORCM_MCA_spin")) {
spin = 1;
/* spin until a debugger can attach */
while (0 != spin) {
ret = 0;
while (ret < 10000) {
ret++;
};
}
}
/* prior to initializing the OPAL layer, check to see
* if the OPAL (and friends) install location has been
* moved. In order to avoid conflicts with any other
* OPAL-using software, the relocation point will have
* been expressed as a set of "ORCM_foo" envars. We
* therefore check for the ORCM_foo values, and name-shift
* any we find to OPAL_foo so that OPAL will find them.
* Since all ORCM tools will have already copied their
* local environment, these name-shifted vars will not
* appear in the environment of any launched processes */
if (NULL != (envar = getenv("ORCM_PREFIX"))) {
opal_unsetenv("ORCM_PREFIX", &environ);
opal_setenv("OPAL_PREFIX", envar, true, &environ);
}
if (NULL != (envar = getenv("ORCM_LIBDIR"))) {
opal_unsetenv("ORCM_LIBDIR", &environ);
opal_setenv("OPAL_LIBDIR", envar, true, &environ);
}
if (NULL != (envar = getenv("ORCM_DATADIR"))) {
opal_unsetenv("ORCM_DATADIR", &environ);
opal_setenv("OPAL_DATADIR", envar, true, &environ);
}
/* initialize the opal layer */
if (ORTE_SUCCESS != (ret = opal_init(NULL, NULL))) {
error = "opal_init";
goto error;
}
orcm_debug_verbosity = -1;
(void) mca_base_var_register ("orcm", "orcm", NULL, "debug_verbose",
"Verbosity level for ORCM debug messages (default: 1)",
MCA_BASE_VAR_TYPE_INT, NULL, 0, 0,
OPAL_INFO_LVL_9, MCA_BASE_VAR_SCOPE_ALL,
&orcm_debug_verbosity);
if (0 <= orcm_debug_verbosity) {
/* get a debug output channel */
OBJ_CONSTRUCT(&lds, opal_output_stream_t);
lds.lds_want_stdout = true;
orcm_debug_output = opal_output_open(&lds);
OBJ_DESTRUCT(&lds);
/* set the verbosity */
opal_output_set_verbosity(orcm_debug_output, orcm_debug_verbosity);
}
/* ensure we know the type of proc for when we finalize */
orte_process_info.proc_type = flags;
/* setup the locks */
if (ORTE_SUCCESS != (ret = orte_locks_init())) {
error = "orte_locks_init";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORTE", ORTE_ERR_BASE, ORTE_ERR_MAX, orte_err2str);
/* Ensure the rest of the process info structure is initialized */
if (ORTE_SUCCESS != (ret = orte_proc_info())) {
error = "orte_proc_info";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORCM", ORCM_ERR_BASE, ORCM_ERR_MAX, orcm_err2str);
/* register handler for attr key -> string conversion */
if (ORTE_SUCCESS != (ret = orte_attr_register("orcm", ORCM_ATTR_KEY_BASE, ORCM_ATTR_KEY_MAX,
orcm_attr_key_print))) {
error = "register attr print";
goto error;
}
/* we don't need a progress thread as all our tools loop inside themselves,
* so define orte_event_base to be the base opal_event_base
*/
orte_event_base = opal_sync_event_base;
/* setup the globals */
orcm_clusters = OBJ_NEW(opal_list_t);
orcm_schedulers = OBJ_NEW(opal_list_t);
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orcm_parser_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orcm_parser_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orcm_parser_base_select())) {
ORTE_ERROR_LOG(ret);
error = "orcm_parser_select";
goto error;
}
/* everyone must open the cfgi framework */
if (ORCM_SUCCESS != (ret = mca_base_framework_open(&orcm_cfgi_base_framework, 0))) {
error = "orcm_cfgi_base_open";
goto error;
}
if (ORCM_SUCCESS != (ret = orcm_cfgi_base_select())) {
error = "orcm_cfgi_select";
goto error;
}
envar = getenv("ORCM_MCA_logical_group_config_file");
if (ORCM_SUCCESS != (ret = orcm_logical_group_load_to_memory(envar))) {
error = "orcm_logical_group_load_to_memory";
goto error;
}
if (ORCM_SCHED == flags) {
if (NULL == (envar = getenv("ORCM_MCA_event_exec_path"))) {
asprintf(&orcm_event_exec_path, "%s/bin", opal_install_dirs.prefix);
} else {
orcm_event_exec_path = strdup(envar);
}
if (NULL == orcm_event_exec_path) {
error = "orcm_event_exec_path";
goto error;
}
}
/* everyone must open the sst framework */
if (ORCM_SUCCESS != (ret = mca_base_framework_open(&orcm_sst_base_framework, 0))) {
error = "orcm_sst_base_open";
goto error;
}
if (ORCM_SUCCESS != (ret = orcm_sst_base_select())) {
error = "orcm_sst_select";
goto error;
}
/* open the ESS and select the correct module for this environment - the
* orcm module is basically a no-op, but we need the framework defined
* as other parts of ORTE will want to call it
*/
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_ess_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess_base_select())) {
error = "orte_ess_base_select";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess.init())) {
error = "orte_ess_init";
goto error;
}
/* initialize us - we will register the ORTE-level MCA params in there */
if (ORTE_SUCCESS != (ret = orcm_sst.init())) {
error = "orte_init";
goto error;
}
/* setup the orte_show_help system - don't do this until the
* end as otherwise show_help messages won't appear
*/
if (ORTE_SUCCESS != (ret = orte_show_help_init())) {
error = "opal_output_init";
goto error;
}
/* initialize orcm datatype support */
if (ORCM_SUCCESS != (ret = orcm_dt_init())) {
error = "orcm_dt_init";
goto error;
}
/* flag that orte is initialized so things can work */
orte_initialized = true;
orte_help_want_aggregate = false;
return ORCM_SUCCESS;
error:
if (ORCM_ERR_SILENT != ret) {
opal_show_help("help-orcm-runtime.txt",
"orcm_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}
int orte_init(int* pargc, char*** pargv, orte_proc_type_t flags)
{
int ret;
char *error = NULL;
if (0 < orte_initialized) {
/* track number of times we have been called */
orte_initialized++;
return ORTE_SUCCESS;
}
orte_initialized++;
/* initialize the opal layer */
if (ORTE_SUCCESS != (ret = opal_init(pargc, pargv))) {
error = "opal_init";
goto error;
}
/* Convince OPAL to use our naming scheme */
opal_process_name_print = _process_name_print_for_opal;
opal_process_name_vpid = _process_name_vpid_for_opal;
opal_process_name_jobid = _process_name_jobid_for_opal;
opal_compare_proc = _process_name_compare;
/* ensure we know the type of proc for when we finalize */
orte_process_info.proc_type = flags;
/* setup the locks */
if (ORTE_SUCCESS != (ret = orte_locks_init())) {
error = "orte_locks_init";
goto error;
}
/* Register all MCA Params */
if (ORTE_SUCCESS != (ret = orte_register_params())) {
error = "orte_register_params";
goto error;
}
/* setup the orte_show_help system */
if (ORTE_SUCCESS != (ret = orte_show_help_init())) {
error = "opal_output_init";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORTE", ORTE_ERR_BASE, ORTE_ERR_MAX, orte_err2str);
/* Ensure the rest of the process info structure is initialized */
if (ORTE_SUCCESS != (ret = orte_proc_info())) {
error = "orte_proc_info";
goto error;
}
/* open the ESS and select the correct module for this environment */
if (ORTE_SUCCESS != (ret = mca_base_framework_open(&orte_ess_base_framework, 0))) {
ORTE_ERROR_LOG(ret);
error = "orte_ess_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess_base_select())) {
error = "orte_ess_base_select";
goto error;
}
if (!ORTE_PROC_IS_APP) {
/* ORTE tools "block" in their own loop over the event
* base, so no progress thread is required - apps will
* start their progress thread in ess_base_std_app.c
* at the appropriate point
*/
orte_event_base = opal_event_base;
}
/* initialize the RTE for this environment */
if (ORTE_SUCCESS != (ret = orte_ess.init())) {
error = "orte_ess_init";
goto error;
}
/* All done */
return ORTE_SUCCESS;
error:
if (ORTE_ERR_SILENT != ret) {
orte_show_help("help-orte-runtime",
"orte_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}
int orcm_init_util(void)
{
int ret, i;
char *error;
char *destdir, *tmp, *mcp, *new_mcp;
/* Setup OPAL */
if( ORTE_SUCCESS != (ret = opal_init(NULL, NULL)) ) {
error = "opal_init_util";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("OPENRCM", ORCM_ERR_BASE, ORCM_ERR_MAX, orcm_err2str);
/* register where the OPENRCM show_help files are located */
if (NULL != (destdir = getenv("ORCM_DESTDIR"))) {
asprintf(&tmp, "%s%s", destdir, ORCM_PKGHELPDIR);
} else {
tmp = strdup(ORCM_PKGHELPDIR);
}
if (ORTE_SUCCESS != (ret = opal_show_help_add_dir(tmp))) {
error = "register show_help_dir";
goto error;
}
free(tmp);
/* Add ORCM's component directory into the
mca_base_param_component_path */
i = mca_base_param_find("mca", NULL, "component_path");
if (i < 0) {
ret = ORCM_ERR_NOT_FOUND;
error = "Could not find mca_component_path";
goto error;
}
mca_base_param_lookup_string(i, &mcp);
if (NULL == mcp) {
ret = ORCM_ERR_NOT_FOUND;
error = "Could not find mca_component_path";
goto error;
}
if (NULL != destdir) {
asprintf(&new_mcp, "%s%s:%s", destdir, ORCM_PKGLIBDIR, mcp);
} else {
asprintf(&new_mcp, "%s:%s", ORCM_PKGLIBDIR, mcp);
}
mca_base_param_set_string(i, new_mcp);
free(new_mcp);
free(mcp);
orcm_util_initialized = true;
return ORCM_SUCCESS;
error:
if (ORCM_ERR_SILENT != ret) {
orte_show_help("help-openrcm-runtime.txt",
"orcm_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}
int
opal_init_util(int* pargc, char*** pargv)
{
int ret;
char *error = NULL;
char hostname[OPAL_MAXHOSTNAMELEN];
OPAL_TIMING_ENV_INIT(otmng);
if( ++opal_util_initialized != 1 ) {
if( opal_util_initialized < 1 ) {
return OPAL_ERROR;
}
return OPAL_SUCCESS;
}
OBJ_CONSTRUCT(&opal_init_util_domain, opal_finalize_domain_t);
(void) opal_finalize_domain_init (&opal_init_util_domain, "opal_init_util");
opal_finalize_set_domain (&opal_init_util_domain);
opal_thread_set_main();
opal_init_called = true;
/* register for */
opal_finalize_register_cleanup_arg (mca_base_framework_close_list, opal_init_util_frameworks);
/* set the nodename right away so anyone who needs it has it. Note
* that we don't bother with fqdn and prefix issues here - we let
* the RTE later replace this with a modified name if the user
* requests it */
gethostname(hostname, sizeof(hostname));
opal_process_info.nodename = strdup(hostname);
/* initialize the memory allocator */
opal_malloc_init();
OPAL_TIMING_ENV_NEXT(otmng, "opal_malloc_init");
/* initialize the output system */
opal_output_init();
/* initialize install dirs code */
if (OPAL_SUCCESS != (ret = mca_base_framework_open(&opal_installdirs_base_framework, 0))) {
fprintf(stderr, "opal_installdirs_base_open() failed -- process will likely abort (%s:%d, returned %d instead of OPAL_SUCCESS)\n",
__FILE__, __LINE__, ret);
return ret;
}
/* initialize the help system */
opal_show_help_init();
OPAL_TIMING_ENV_NEXT(otmng, "opal_show_help_init");
/* register handler for errnum -> string converstion */
if (OPAL_SUCCESS !=
(ret = opal_error_register("OPAL",
OPAL_ERR_BASE, OPAL_ERR_MAX, opal_err2str))) {
return opal_init_error ("opal_error_register", ret);
}
/* keyval lex-based parser */
if (OPAL_SUCCESS != (ret = opal_util_keyval_parse_init())) {
return opal_init_error ("opal_util_keyval_parse_init", ret);
}
// Disable PSM signal hijacking (see comment in function for more
// details)
opal_init_psm();
OPAL_TIMING_ENV_NEXT(otmng, "opal_init_psm");
/* Setup the parameter system */
if (OPAL_SUCCESS != (ret = mca_base_var_init())) {
return opal_init_error ("mca_base_var_init", ret);
}
OPAL_TIMING_ENV_NEXT(otmng, "opal_var_init");
/* read any param files that were provided */
if (OPAL_SUCCESS != (ret = mca_base_var_cache_files(false))) {
return opal_init_error ("failed to cache files", ret);
}
OPAL_TIMING_ENV_NEXT(otmng, "opal_var_cache");
/* register params for opal */
if (OPAL_SUCCESS != (ret = opal_register_params())) {
return opal_init_error ("opal_register_params", ret);
}
if (OPAL_SUCCESS != (ret = opal_net_init())) {
return opal_init_error ("opal_net_init", ret);
}
OPAL_TIMING_ENV_NEXT(otmng, "opal_net_init");
/* pretty-print stack handlers */
if (OPAL_SUCCESS != (ret = opal_util_register_stackhandlers())) {
return opal_init_error ("opal_util_register_stackhandlers", ret);
}
/* set system resource limits - internally protected against
* doing so twice in cases where the launch agent did it for us
*/
if (OPAL_SUCCESS != (ret = opal_util_init_sys_limits(&error))) {
opal_show_help("help-opal-runtime.txt",
"opal_init:syslimit", false,
error);
return OPAL_ERR_SILENT;
}
/* initialize the arch string */
if (OPAL_SUCCESS != (ret = opal_arch_init ())) {
return opal_init_error ("opal_arch_init", ret);
}
OPAL_TIMING_ENV_NEXT(otmng, "opal_arch_init");
/* initialize the datatype engine */
if (OPAL_SUCCESS != (ret = opal_datatype_init ())) {
return opal_init_error ("opal_datatype_init", ret);
}
OPAL_TIMING_ENV_NEXT(otmng, "opal_datatype_init");
/* Initialize the data storage service. */
if (OPAL_SUCCESS != (ret = opal_dss_open())) {
return opal_init_error ("opal_dss_open", ret);
}
OPAL_TIMING_ENV_NEXT(otmng, "opal_dss_open");
/* initialize the mca */
if (OPAL_SUCCESS != (ret = mca_base_open())) {
return opal_init_error ("mca_base_open", ret);
}
OPAL_TIMING_ENV_NEXT(otmng, "mca_base_open");
/* initialize if framework */
if (OPAL_SUCCESS != (ret = mca_base_framework_open(&opal_if_base_framework, 0))) {
fprintf(stderr, "opal_if_base_open() failed -- process will likely abort (%s:%d, returned %d instead of OPAL_SUCCESS)\n",
__FILE__, __LINE__, ret);
return ret;
}
OPAL_TIMING_ENV_NEXT(otmng, "opal_if_init");
return OPAL_SUCCESS;
}
int orte_init(int* pargc, char*** pargv, orte_proc_type_t flags)
{
int ret;
char *error = NULL;
if (0 < orte_initialized) {
/* track number of times we have been called */
orte_initialized++;
return ORTE_SUCCESS;
}
orte_initialized++;
/* initialize the opal layer */
if (ORTE_SUCCESS != (ret = opal_init(pargc, pargv))) {
error = "opal_init";
goto error;
}
/* ensure we know the type of proc for when we finalize */
orte_process_info.proc_type = flags;
/* setup the locks */
if (ORTE_SUCCESS != (ret = orte_locks_init())) {
error = "orte_locks_init";
goto error;
}
/* Register all MCA Params */
if (ORTE_SUCCESS != (ret = orte_register_params())) {
error = "orte_register_params";
goto error;
}
/* setup the orte_show_help system */
if (ORTE_SUCCESS != (ret = orte_show_help_init())) {
error = "opal_output_init";
goto error;
}
/* register handler for errnum -> string conversion */
opal_error_register("ORTE", ORTE_ERR_BASE, ORTE_ERR_MAX, orte_err2str);
/* Ensure the rest of the process info structure is initialized */
if (ORTE_SUCCESS != (ret = orte_proc_info())) {
error = "orte_proc_info";
goto error;
}
/* open the ESS and select the correct module for this environment */
if (ORTE_SUCCESS != (ret = orte_ess_base_open())) {
error = "orte_ess_base_open";
goto error;
}
if (ORTE_SUCCESS != (ret = orte_ess_base_select())) {
error = "orte_ess_base_select";
goto error;
}
#if ORTE_ENABLE_PROGRESS_THREADS
#if OPAL_EVENT_HAVE_THREAD_SUPPORT
/* get a separate orte event base */
orte_event_base = opal_event_base_create();
/* construct the thread object */
OBJ_CONSTRUCT(&orte_progress_thread, opal_thread_t);
/* fork off a thread to progress it */
orte_progress_thread.t_run = orte_progress_thread_engine;
if (OPAL_SUCCESS != (ret = opal_thread_start(&orte_progress_thread))) {
error = "orte progress thread start";
goto error;
}
#else
error = "event thread support is not configured";
ret = ORTE_ERROR;
goto error;
#endif
#else
/* set the event base to the opal one */
orte_event_base = opal_event_base;
#endif
/* initialize the RTE for this environment */
if (ORTE_SUCCESS != (ret = orte_ess.init())) {
error = "orte_ess_init";
goto error;
}
/* All done */
return ORTE_SUCCESS;
error:
if (ORTE_ERR_SILENT != ret) {
orte_show_help("help-orte-runtime",
"orte_init:startup:internal-failure",
true, error, ORTE_ERROR_NAME(ret), ret);
}
return ret;
}