Merge pull request #8758 from rhc54/topic/bkpt

Add a soft "breakpoint" capability

Author: rhc54

3rd-party/openpmix

@@ -1,35 +1,21 @@
 enable_orterun_prefix_by_default=yes
-enable_mpi_thread_multiple=no
-enable_mem_debug=no
+enable_oshmem=no
+enable_sensors=no
 enable_mem_profile=no
 enable_debug_symbols=yes
 enable_binaries=yes
-enable_heterogeneous=no
 enable_picky=yes
 enable_debug=yes
 enable_shared=yes
 enable_static=no
-enable_memchecker=no
+enable_io_romio=no
 enable_ipv6=no
+enable_man_pages=no
 enable_mpi_fortran=no
-enable_mpi_cxx=no
-enable_mpi_cxx_seek=no
-enable_cxx_exceptions=no
-enable_mpi_java=no
-enable_io_romio=no
-enable_contrib_no_build=libnbc
-enable_install_libpmix=yes
+enable_memchecker=no
+enable_mca_no_build=memchecker,coll-adapt,coll-cuda,coll-demo,coll-ftagree,coll-han,coll-hcoll,coll-inter,coll-libnbc,coll-monitoring,coll-portals4,coll-tuned,common-monitoring,common-ompio,fbtl,fcoll,fs,io,mtl,osc,pml-cm,pml-monitoring,pml-ucx,pml-v,sharedfp,topo,vprotocol,btl-ofi,btl-portals4,btl-smcuda,btl-uct,btl-ugni,btl-usnic,common-cuda,common-ofi,common-ucx
+enable_contrib_no_build=libompitrace
 with_memory_manager=no
-with_tm=no
-with_psm=no
-with_psm2=no
 with_devel_headers=yes
-with_libfabric=no
-with_portals=no
-with_valgrind=no
-if [ -n "$SLURMHOME" ] ; then
-    with_slurm=$SLURMHOME
-    with_pmi=$SLURMHOME
-else
-    with_slurm=no
-fi
+with_xgrid=no
+with_slurm=no

3rd-party/prrte

@@ -1,24 +1,19 @@
-enable_orcm=no
 enable_oshmem=no
 enable_sensors=no
-enable_mpi_thread_multiple=yes
-enable_mem_debug=yes
 enable_mem_profile=no
 enable_debug_symbols=yes
 enable_binaries=yes
-enable_heterogeneous=yes
 enable_picky=yes
 enable_debug=yes
 enable_shared=yes
 enable_static=no
 enable_io_romio=no
 enable_ipv6=no
+enable_man_pages=no
 enable_mpi_fortran=no
-enable_mpi_cxx=no
-enable_mpi_cxx_seek=no
 enable_memchecker=no
-enable_mca_no_build=memchecker
-enable_contrib_no_build=libnbc
+enable_mca_no_build=memchecker,coll-adapt,coll-cuda,coll-demo,coll-ftagree,coll-han,coll-hcoll,coll-inter,coll-libnbc,coll-monitoring,coll-portals4,coll-tuned,common-monitoring,common-ompio,fbtl,fcoll,fs,io,mtl,osc,pml-cm,pml-monitoring,pml-ucx,pml-v,sharedfp,topo,vprotocol,btl-ofi,btl-portals4,btl-smcuda,btl-uct,btl-ugni,btl-usnic,common-cuda,common-ofi,common-ucx
+enable_contrib_no_build=libompitrace
 with_memory_manager=no
 with_devel_headers=yes
 with_xgrid=no

contrib/platform/intel/bend/linux

@@ -1051,32 +1051,35 @@ static void _release_fn(size_t refid, pmix_status_t status,
     debugger_event_active = false;
 }
 
-/*
- * Wait for a debugger if asked.  We support two ways of waiting for
- * attaching debuggers
- */
-void ompi_rte_wait_for_debugger(void)
+void ompi_rte_breakpoint(char *name)
 {
     pmix_info_t directive;
     char *evar;
-    int time, code = PMIX_ERR_DEBUGGER_RELEASE;
-    pmix_info_t info;
+    int rc, code = PMIX_DEBUGGER_RELEASE;
+    pmix_info_t info[2];
+    uint32_t u32, *u32ptr;
+    opal_process_name_t pname;
+
+    if (NULL != name
+        && NULL != (evar = getenv("OMPI_BREAKPOINT"))
+        && 0 != strcasecmp(evar, name)) {
+        /* they don't want to stop here */
+        return;
+    }
 
     /* check PMIx to see if we are under a debugger */
-    if (NULL == getenv("PMIX_DEBUG_WAIT_FOR_NOTIFY") &&
-        NULL == getenv("PMIX_TEST_DEBUGGER_ATTACH")) {
+    u32ptr = &u32;
+    pname.jobid = opal_process_info.my_name.jobid;
+    pname.vpid = OPAL_VPID_WILDCARD;
+    OPAL_MODEX_RECV_VALUE_OPTIONAL(rc, "PMIX_DEBUG_STOP_IN_APP",
+                                   &pname, &u32ptr, PMIX_PROC_RANK);
+    if (PMIX_SUCCESS != rc) {
         /* if not, just return */
         return;
     }
-
-    /* if we are being debugged, then we need to find
-     * the correct plug-ins
-     */
-    ompi_debugger_setup_dlls();
-
-    if (NULL != (evar = getenv("PMIX_TEST_DEBUGGER_SLEEP"))) {
-        time = strtol(evar, NULL, 10);
-        sleep(time);
+    /* are we included? */
+    if (!PMIX_CHECK_RANK(u32, opal_process_info.myprocid.rank)) {
+        /* no - ignore it */
         return;
     }
 
@@ -1086,11 +1089,13 @@ void ompi_rte_wait_for_debugger(void)
     PMIX_INFO_DESTRUCT(&directive);
 
     /* notify the host that we are waiting */
-    PMIX_INFO_LOAD(&info, PMIX_EVENT_NON_DEFAULT, NULL, PMIX_BOOL);
-    PMIx_Notify_event(PMIX_DEBUG_WAITING_FOR_NOTIFY,
+    PMIX_INFO_LOAD(&info[0], PMIX_EVENT_NON_DEFAULT, NULL, PMIX_BOOL);
+    PMIX_INFO_LOAD(&info[1], PMIX_BREAKPOINT, "mpi-init", PMIX_STRING);
+    PMIx_Notify_event(PMIX_READY_FOR_DEBUG,
                       &opal_process_info.myprocid,
-                      PMIX_RANGE_RM, &info, 1, NULL, NULL);
-    PMIX_INFO_DESTRUCT(&info);
+                      PMIX_RANGE_RM, info, 2, NULL, NULL);
+    PMIX_INFO_DESTRUCT(&info[0]);
+    PMIX_INFO_DESTRUCT(&info[1]);
 
     /* let the MPI progress engine run while we wait for debugger release */
     OMPI_WAIT_FOR_COMPLETION(debugger_event_active);
@@ -1099,6 +1104,22 @@ void ompi_rte_wait_for_debugger(void)
     PMIx_Deregister_event_handler(handler, NULL, NULL);
 }
 
+/*
+ * Wait for a debugger if asked.  We support two ways of waiting for
+ * attaching debuggers
+ */
+void ompi_rte_wait_for_debugger(void)
+{
+    if (NULL != getenv("PMIX_TEST_DEBUGGER_ATTACH")
+        || NULL == getenv("OMPI_BREAKPOINT")) {
+        ompi_rte_breakpoint(NULL);
+        return;
+    }
+
+    /* check for the "mpi-init" breakpoint */
+    ompi_rte_breakpoint("mpi-init");
+}
+
 static int _setup_top_session_dir(char **sdir)
 {
     char *tmpdir;

contrib/platform/intel/bend/mac

@@ -216,6 +216,8 @@ OMPI_DECLSPEC int ompi_rte_convert_string_to_process_name(opal_process_name_t *n
 OMPI_DECLSPEC int ompi_rte_convert_process_name_to_string(char** name_string,
                                                           const opal_process_name_t *name);
 
+OMPI_DECLSPEC void ompi_rte_breakpoint(char *name);
+
 #define OMPI_LOCAL_JOBID(n) \
     ( (n) & 0x0000ffff)
 #define OMPI_JOB_FAMILY(n)  \