Skip to content

Instantly share code, notes, and snippets.

@rhc54

rhc54/gist:9c615b86f2d43db1c15911390667ab10

Created July 19, 2023 20:13
Show Gist options
  • Save rhc54/9c615b86f2d43db1c15911390667ab10 to your computer and use it in GitHub Desktop.
Save rhc54/9c615b86f2d43db1c15911390667ab10 to your computer and use it in GitHub Desktop.
Download ZIP
OMPI main - patch for PMIx v3.x
Raw
gistfile1.txt
diff --git a/VERSION b/VERSION
index 2178439c11..9bd94561b2 100644
--- a/VERSION
+++ b/VERSION
@@ -25,7 +25,7 @@ mpi_standard_subversion=1
# OMPI required dependency versions.
# List in x.y.z format.
-pmix_min_version=4.1.2
+pmix_min_version=3.2.4
prte_min_version=2.0.2
hwloc_min_version=1.11.0
event_min_version=2.0.21
diff --git a/ompi/communicator/comm_cid.c b/ompi/communicator/comm_cid.c
index 07970e8354..1e0128cbfe 100644
--- a/ompi/communicator/comm_cid.c
+++ b/ompi/communicator/comm_cid.c
@@ -303,6 +303,7 @@ static volatile int64_t ompi_comm_cid_lowest_id = INT64_MAX;
static int ompi_comm_cid_epoch = INT_MAX;
#endif /* OPAL_ENABLE_FT_MPI */
+#if PMIX_NUMERIC_VERSION > 0x00040000
static int ompi_comm_ext_cid_new_block (ompi_communicator_t *newcomm, ompi_communicator_t *comm,
ompi_comm_extended_cid_block_t *new_block,
const void *arg0, const void *arg1, bool send_first, int mode,
@@ -410,10 +411,11 @@ fn_exit:
free (name_array);
name_array = NULL;
}
-
return ret;
}
+#endif
+#if PMIX_NUMERIC_VERSION > 0x00040000
static int ompi_comm_nextcid_ext_nb (ompi_communicator_t *newcomm, ompi_communicator_t *comm,
ompi_communicator_t *bridgecomm, const void *arg0, const void *arg1,
bool send_first, int mode, ompi_request_t **req)
@@ -474,6 +476,7 @@ static int ompi_comm_nextcid_ext_nb (ompi_communicator_t *newcomm, ompi_communic
/* nothing more to do here */
return OMPI_SUCCESS;
}
+#endif
int ompi_comm_nextcid_nb (ompi_communicator_t *newcomm, ompi_communicator_t *comm,
ompi_communicator_t *bridgecomm, const void *arg0, const void *arg1,
@@ -482,11 +485,13 @@ int ompi_comm_nextcid_nb (ompi_communicator_t *newcomm, ompi_communicator_t *com
ompi_comm_cid_context_t *context;
ompi_comm_request_t *request;
+#if PMIX_NUMERIC_VERSION > 0x00040000
if (mca_pml_base_supports_extended_cid() && NULL == comm) {
return ompi_comm_nextcid_ext_nb (newcomm, comm, bridgecomm, arg0, arg1, send_first, mode, req);
}
+#endif
- /* old CID algorighm */
+ /* old CID algorithm */
/* if we got here and comm is NULL then that means the app is invoking MPI-4 Sessions or later
functions but the pml does not support these functions so return not supported */
diff --git a/ompi/errhandler/errhandler.c b/ompi/errhandler/errhandler.c
index 76b902932f..7c98d78fda 100644
--- a/ompi/errhandler/errhandler.c
+++ b/ompi/errhandler/errhandler.c
@@ -463,6 +463,7 @@ typedef struct ompi_errhandler_event_s {
static void *ompi_errhandler_event_cb(int fd, int flags, void *context) {
ompi_errhandler_event_t *event = (ompi_errhandler_event_t*) context;
int status = event->status;
+#if PMIX_NUMERIC_VERSION > 0x00040000
opal_process_name_t source = event->source;
opal_process_name_t prc;
int rc;
@@ -510,6 +511,7 @@ static void *ompi_errhandler_event_cb(int fd, int flags, void *context) {
status );
}
#endif /* OPAL_ENABLE_FT_MPI */
+#endif // PMIx
opal_event_del(&event->super);
free(event);
/* our default action is to abort */
diff --git a/ompi/instance/instance.c b/ompi/instance/instance.c
index 8b67db9ccd..70b5d8c9b4 100644
--- a/ompi/instance/instance.c
+++ b/ompi/instance/instance.c
@@ -71,12 +71,13 @@ enum {
opal_atomic_int32_t ompi_instance_count = 0;
+#if PMIX_NUMERIC_VERSION > 0x00040000
static const char *ompi_instance_builtin_psets[] = {
"mpi://WORLD",
"mpi://SELF",
"mpix://SHARED",
};
-
+#endif
static const int32_t ompi_instance_builtin_count = 3;
/** finalization functions that need to be called on teardown */
@@ -432,6 +433,7 @@ static int ompi_mpi_instance_init_common (int argc, char **argv)
return ret;
}
+#if PMIX_NUMERIC_VERSION > 0x00040000
/* Register the ULFM errhandler callback */
/* we want to go first */
PMIX_INFO_LOAD(&info[0], PMIX_EVENT_HDLR_PREPEND, NULL, PMIX_BOOL);
@@ -455,6 +457,7 @@ static int ompi_mpi_instance_init_common (int argc, char **argv)
ret = opal_pmix_convert_status(rc);
return ret;
}
+#endif
/* initialize MPI_INFO_ENV */
if (OMPI_SUCCESS != (ret = ompi_mpiinfo_init_env(0, NULL, &ompi_mpi_info_env.info))) {
@@ -971,6 +974,7 @@ int ompi_mpi_instance_finalize (ompi_instance_t **instance)
return ret;
}
+#if PMIX_NUMERIC_VERSION > 0x00040000
static void ompi_instance_get_num_psets_complete (pmix_status_t status,
pmix_info_t *info,
size_t ninfo,
@@ -1067,11 +1071,13 @@ static void ompi_instance_refresh_pmix_psets (const char *key)
opal_mutex_unlock (&instance_lock);
}
-
+#endif
int ompi_instance_get_num_psets (ompi_instance_t *instance, int *npset_names)
{
+#if PMIX_NUMERIC_VERSION >= 0x00040000
ompi_instance_refresh_pmix_psets (PMIX_QUERY_NUM_PSETS);
+#endif
*npset_names = ompi_instance_builtin_count + ompi_mpi_instance_num_pmix_psets;
return OMPI_SUCCESS;
@@ -1079,6 +1085,7 @@ int ompi_instance_get_num_psets (ompi_instance_t *instance, int *npset_names)
int ompi_instance_get_nth_pset (ompi_instance_t *instance, int n, int *len, char *pset_name)
{
+#if PMIX_NUMERIC_VERSION >= 0x00040000
if (NULL == ompi_mpi_instance_pmix_psets && n >= ompi_instance_builtin_count) {
ompi_instance_refresh_pmix_psets (PMIX_QUERY_PSET_NAMES);
}
@@ -1101,10 +1108,11 @@ int ompi_instance_get_nth_pset (ompi_instance_t *instance, int n, int *len, char
} else {
strncpy (pset_name, ompi_mpi_instance_pmix_psets[n - ompi_instance_builtin_count], *len);
}
-
+#endif
return OMPI_SUCCESS;
}
+#if PMIX_NUMERIC_VERSION >= 0x00040000
static int ompi_instance_group_world (ompi_instance_t *instance, ompi_group_t **group_out)
{
ompi_group_t *group;
@@ -1318,9 +1326,11 @@ static int ompi_instance_get_pmix_pset_size (ompi_instance_t *instance, const ch
return OMPI_SUCCESS;
}
+#endif
int ompi_group_from_pset (ompi_instance_t *instance, const char *pset_name, ompi_group_t **group_out)
{
+#if PMIX_NUMERIC_VERSION >= 0x00040000
if (NULL == group_out) {
return OMPI_ERR_BAD_PARAM;
}
@@ -1343,10 +1353,14 @@ int ompi_group_from_pset (ompi_instance_t *instance, const char *pset_name, ompi
}
return ompi_instance_group_pmix_pset (instance, pset_name, group_out);
+#else
+ return OMPI_ERR_BAD_PARAM;
+#endif
}
int ompi_instance_get_pset_info (ompi_instance_t *instance, const char *pset_name, opal_info_t **info_used)
{
+#if PMIX_NUMERIC_VERSION >= 0x00040000
ompi_info_t *info = ompi_info_allocate ();
char tmp[16];
size_t size = 0UL;
@@ -1387,4 +1401,7 @@ int ompi_instance_get_pset_info (ompi_instance_t *instance, const char *pset_nam
}
return ret;
+#else
+ return OMPI_ERR_BAD_PARAM;
+#endif
}
diff --git a/ompi/runtime/ompi_rte.c b/ompi/runtime/ompi_rte.c
index 2a2d66bbc3..e491e6f0c7 100644
--- a/ompi/runtime/ompi_rte.c
+++ b/ompi/runtime/ompi_rte.c
@@ -1080,9 +1080,10 @@ void ompi_rte_abort_peers(opal_process_name_t *procs,
return;
}
-static size_t handler = SIZE_MAX;
static volatile bool debugger_event_active = true;
+#if PMIX_NUMERIC_VERSION > 0x00040000
+static size_t handler = SIZE_MAX;
static void _release_fn(size_t refid, pmix_status_t status,
const pmix_proc_t *source,
pmix_info_t info[], size_t ninfo,
@@ -1096,9 +1097,11 @@ static void _release_fn(size_t refid, pmix_status_t status,
}
debugger_event_active = false;
}
+#endif
void ompi_rte_breakpoint(char *name)
{
+#if PMIX_NUMERIC_VERSION > 0x00040000
pmix_info_t directive;
char *evar;
int rc, code = PMIX_DEBUGGER_RELEASE;
@@ -1141,6 +1144,7 @@ void ompi_rte_breakpoint(char *name)
/* deregister the event handler */
PMIx_Deregister_event_handler(handler, NULL, NULL);
+#endif
}
/*
diff --git a/opal/mca/pmix/pmix-internal.h b/opal/mca/pmix/pmix-internal.h
index 4e10393f60..ab5503dfcd 100644
--- a/opal/mca/pmix/pmix-internal.h
+++ b/opal/mca/pmix/pmix-internal.h
@@ -284,6 +284,36 @@ typedef struct {
* is to be returned
* t - the expected data type
*/
+#if PMIX_NUMERIC_VERSION < 0x00040000
+#define OPAL_MODEX_RECV_VALUE_OPTIONAL(r, s, p, d, t) \
+ do { \
+ pmix_proc_t _proc; \
+ pmix_value_t *_kv = NULL; \
+ pmix_info_t _info; \
+ size_t _sz; \
+ pmix_key_t _key; \
+ PMIX_LOAD_KEY(_key, (s)); \
+ OPAL_OUTPUT_VERBOSE((1, opal_pmix_verbose_output, \
+ "%s[%s:%d] MODEX RECV VALUE OPTIONAL FOR PROC %s KEY %s", \
+ OPAL_NAME_PRINT(OPAL_PROC_MY_NAME), __FILE__, __LINE__, \
+ OPAL_NAME_PRINT(*(p)), (s))); \
+ OPAL_PMIX_CONVERT_NAME(&_proc, (p)); \
+ PMIX_INFO_LOAD(&_info, PMIX_OPTIONAL, NULL, PMIX_BOOL); \
+ (r) = PMIx_Get(&(_proc), _key, &(_info), 1, &(_kv)); \
+ PMIX_INFO_DESTRUCT(&_info); \
+ if (NULL == _kv) { \
+ (r) = PMIX_ERR_NOT_FOUND; \
+ } else if (_kv->type != (t)) { \
+ (r) = PMIX_ERR_TYPE_MISMATCH; \
+ } else if (PMIX_SUCCESS == (r) && NULL != (d)) { \
+ PMIX_VALUE_UNLOAD((r), _kv, (void **) (d), &_sz); \
+ } \
+ if (NULL != _kv) { \
+ PMIX_VALUE_RELEASE(_kv); \
+ } \
+ } while (0);
+
+#else
#define OPAL_MODEX_RECV_VALUE_OPTIONAL(r, s, p, d, t) \
do { \
pmix_proc_t _proc; \
@@ -310,6 +340,8 @@ typedef struct {
} \
} while (0);
+#endif
+
/**
* Provide a simplified macro for retrieving modex data
* from another process when we want the PMIx module
@@ -325,6 +357,36 @@ typedef struct {
* is to be returned
* t - the expected data type
*/
+#if PMIX_NUMERIC_VERSION < 0x00040000
+#define OPAL_MODEX_RECV_VALUE_IMMEDIATE(r, s, p, d, t) \
+ do { \
+ pmix_proc_t _proc; \
+ pmix_value_t *_kv = NULL; \
+ pmix_info_t _info; \
+ size_t _sz; \
+ pmix_key_t _key; \
+ PMIX_LOAD_KEY(_key, (s)); \
+ OPAL_OUTPUT_VERBOSE((1, opal_pmix_verbose_output, \
+ "%s[%s:%d] MODEX RECV VALUE IMMEDIATE FOR PROC %s KEY %s", \
+ OPAL_NAME_PRINT(OPAL_PROC_MY_NAME), __FILE__, __LINE__, \
+ OPAL_NAME_PRINT(*(p)), (s))); \
+ OPAL_PMIX_CONVERT_NAME(&_proc, (p)); \
+ PMIX_INFO_LOAD(&_info, PMIX_IMMEDIATE, NULL, PMIX_BOOL); \
+ (r) = PMIx_Get(&(_proc), _key, &(_info), 1, &(_kv)); \
+ PMIX_INFO_DESTRUCT(&_info); \
+ if (NULL == _kv) { \
+ (r) = PMIX_ERR_NOT_FOUND; \
+ } else if (_kv->type != (t)) { \
+ (r) = PMIX_ERR_TYPE_MISMATCH; \
+ } else if (PMIX_SUCCESS == (r)) { \
+ PMIX_VALUE_UNLOAD((r), _kv, (void **) (d), &_sz); \
+ } \
+ if (NULL != _kv) { \
+ PMIX_VALUE_RELEASE(_kv); \
+ } \
+ } while (0);
+
+#else
#define OPAL_MODEX_RECV_VALUE_IMMEDIATE(r, s, p, d, t) \
do { \
pmix_proc_t _proc; \
@@ -351,6 +413,8 @@ typedef struct {
} \
} while (0);
+#endif
+
/**
* Provide a simplified macro for retrieving modex data
* from another process:
@@ -363,14 +427,34 @@ typedef struct {
* is to be returned
* t - the expected data type
*/
+#if PMIX_NUMERIC_VERSION < 0x00040000
+#define OPAL_MODEX_RECV_VALUE(r, s, p, d, t) \
+ do { \
+ pmix_proc_t _proc; \
+ pmix_value_t *_kv = NULL; \
+ size_t _sz; \
+ pmix_key_t _key; \
+ PMIX_LOAD_KEY(_key, (s)); \
+ OPAL_PMIX_CONVERT_NAME(&_proc, (p)); \
+ (r) = PMIx_Get(&(_proc), _key, NULL, 0, &(_kv)); \
+ if (NULL == _kv) { \
+ (r) = PMIX_ERR_NOT_FOUND; \
+ } else if (_kv->type != (t)) { \
+ (r) = PMIX_ERR_TYPE_MISMATCH; \
+ } else if (PMIX_SUCCESS == (r)) { \
+ PMIX_VALUE_UNLOAD((r), _kv, (void **) (d), &_sz); \
+ } \
+ if (NULL != _kv) { \
+ PMIX_VALUE_RELEASE(_kv); \
+ } \
+ } while (0);
+
+#else
#define OPAL_MODEX_RECV_VALUE(r, s, p, d, t) \
do { \
pmix_proc_t _proc; \
pmix_value_t *_kv = NULL; \
size_t _sz; \
- OPAL_OUTPUT_VERBOSE( \
- (1, opal_pmix_verbose_output, "%s[%s:%d] MODEX RECV VALUE FOR PROC %s KEY %s", \
- OPAL_NAME_PRINT(OPAL_PROC_MY_NAME), __FILE__, __LINE__, OPAL_NAME_PRINT(*(p)), (s))); \
OPAL_PMIX_CONVERT_NAME(&_proc, (p)); \
(r) = PMIx_Get(&(_proc), (s), NULL, 0, &(_kv)); \
if (NULL == _kv) { \
@@ -384,6 +468,7 @@ typedef struct {
PMIX_VALUE_RELEASE(_kv); \
} \
} while (0);
+#endif
/**
* Provide a simplified macro for retrieving modex data
@@ -398,6 +483,36 @@ typedef struct {
* sz - pointer to a location wherein the number of bytes
* in the data object can be returned (size_t)
*/
+#if PMIX_NUMERIC_VERSION < 0x00040000
+#define OPAL_MODEX_RECV_STRING_OPTIONAL(r, s, p, d, sz) \
+ do { \
+ pmix_proc_t _proc; \
+ pmix_value_t *_kv = NULL; \
+ pmix_info_t _info; \
+ pmix_key_t _key; \
+ PMIX_LOAD_KEY(_key, (s)); \
+ OPAL_OUTPUT_VERBOSE((1, opal_pmix_verbose_output, \
+ "%s[%s:%d] MODEX RECV STRING OPTIONAL FOR PROC %s KEY %s", \
+ OPAL_NAME_PRINT(OPAL_PROC_MY_NAME), __FILE__, __LINE__, \
+ OPAL_NAME_PRINT(*(p)), (s))); \
+ *(d) = NULL; \
+ *(sz) = 0; \
+ OPAL_PMIX_CONVERT_NAME(&_proc, (p)); \
+ PMIX_INFO_LOAD(&_info, PMIX_OPTIONAL, NULL, PMIX_BOOL); \
+ (r) = PMIx_Get(&(_proc), _key, &(_info), 1, &(_kv)); \
+ if (NULL == _kv) { \
+ (r) = PMIX_ERR_NOT_FOUND; \
+ } else if (PMIX_SUCCESS == (r)) { \
+ *(d) = (uint8_t *) _kv->data.bo.bytes; \
+ *(sz) = _kv->data.bo.size; \
+ _kv->data.bo.bytes = NULL; /* protect the data */ \
+ } \
+ if (NULL != _kv) { \
+ PMIX_VALUE_RELEASE(_kv); \
+ } \
+ } while (0);
+
+#else
#define OPAL_MODEX_RECV_STRING_OPTIONAL(r, s, p, d, sz) \
do { \
pmix_proc_t _proc; \
@@ -424,6 +539,8 @@ typedef struct {
} \
} while (0);
+#endif
+
/**
* Provide a simplified macro for retrieving modex data
* from another process:
@@ -688,6 +805,7 @@ OPAL_DECLSPEC int opal_pmix_register_cleanup(char *path, bool directory, bool ig
/* protect against early versions of PMIx */
#if PMIX_VERSION_MAJOR == 3
+# define PMIX_ERR_PROC_TERM_WO_SYNC -200
# if PMIX_VERSION_MINOR == 0
# if PMIX_VERSION_RELEASE == 0
# define PMIX_NUMERIC_VERSION 0x00030000
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment