Handle the case where a user's rankfile specifies only cpus, and not socket:cpu pairs.

This commit was SVN r25803.

opal/mca/hwloc/base/hwloc_base_util.c

@@ -847,7 +847,7 @@ static int socket_core_to_cpu_set(char *socket_core_list,
                 opal_argv_free(range);
             }
         } else {
-            /* we don't support other levels yet */
+            /* unrecognized option */
             rc = OPAL_ERR_NOT_SUPPORTED;
             break;
         }
@@ -863,7 +863,12 @@ int opal_hwloc_base_slot_list_parse(const char *slot_str,
 {
     char **item;
     int rc, i;
-    
+    hwloc_obj_t pu;
+    hwloc_cpuset_t pucpus;
+    char **range;
+    size_t range_cnt;
+    int core_id, lower_range, upper_range;
+
     /* bozo checks */
     if (NULL == opal_hwloc_topology) {
         return OPAL_ERR_NOT_SUPPORTED;
@@ -904,9 +909,49 @@ int opal_hwloc_base_slot_list_parse(const char *slot_str,
                 }
             }
         } else {
-            /* we don't support other things yet */
-            opal_argv_free(item);
-            return OPAL_ERR_NOT_SUPPORTED;
+            /* just a core specification - see if one or a range was given */
+            range = opal_argv_split(item[i], '-');
+            range_cnt = opal_argv_count(range);
+            hwloc_bitmap_zero(cpumask);
+            /* see if a range was set or not */
+            switch (range_cnt) {
+            case 1:  /* only one core specified */
+                core_id = atoi(range[0]);
+                /* find the specified logical available cpu */
+                if (NULL == (pu = get_pu(topo, core_id))) {
+                    opal_argv_free(range);
+                    opal_argv_free(item);
+                    return OPAL_ERROR;
+                }
+                /* get the available cpus for that object */
+                pucpus = opal_hwloc_base_get_available_cpus(topo, pu);
+                 /* set that in the mask */
+                hwloc_bitmap_copy(cpumask, pucpus);
+                break;
+                    
+            case 2:  /* range of core id's was given */
+                lower_range = atoi(range[0]);
+                upper_range = atoi(range[1]);
+                hwloc_bitmap_zero(cpumask);
+                for (core_id=lower_range; core_id <= upper_range; core_id++) {
+                    /* find the specified logical available cpu */
+                    if (NULL == (pu = get_pu(topo, core_id))) {
+                        opal_argv_free(range);
+                        opal_argv_free(item);
+                        return OPAL_ERROR;
+                    }
+                    /* get the available cpus for that object */
+                    pucpus = opal_hwloc_base_get_available_cpus(topo, pu);
+                    /* set that in the mask */
+                    hwloc_bitmap_or(cpumask, cpumask, pucpus);
+                }
+                break;
+                
+            default:
+                opal_argv_free(range);
+                opal_argv_free(item);
+                return OPAL_ERROR;
+            }
         }
     }
     opal_argv_free(item);

orte/mca/plm/base/plm_base_launch_support.c

@@ -79,7 +79,11 @@ int orte_plm_base_setup_job(orte_job_t *jdata)
     int rc;
     int32_t ljob;
     int i;
-    
+    orte_node_t *node;
+#if OPAL_HAVE_HWLOC
+    hwloc_topology_t t0;
+#endif
+
     OPAL_OUTPUT_VERBOSE((5, orte_plm_globals.output,
                          "%s plm:base:setup_job for job %s",
                          ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
@@ -124,6 +128,25 @@ int orte_plm_base_setup_job(orte_job_t *jdata)
         return rc;
     }
 
+#if OPAL_HAVE_HWLOC
+    /* if we are not going to launch, then we need to set any
+     * undefined topologies to match our own so the mapper
+     * can operate
+     */
+    if (orte_do_not_launch) {
+        node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, 0);
+        t0 = node->topology;
+        for (i=1; i < orte_node_pool->size; i++) {
+            if (NULL == (node = (orte_node_t*)opal_pointer_array_get_item(orte_node_pool, i))) {
+                continue;
+            }
+            if (NULL == node->topology) {
+                node->topology = t0;
+            }
+        }
+    }
+#endif
+
     /* map the job */
     if (ORTE_SUCCESS != (rc = orte_rmaps.map_job(jdata))) {
         ORTE_ERROR_LOG(rc);

orte/mca/rmaps/base/help-orte-rmaps-base.txt

@@ -131,15 +131,14 @@ You can override this protection by adding the "overload-allowed"
 option to your binding directive.
 #
 [rmaps:no-topology]
-A request was made for nperxxx that requires knowledge of
+A mapping directive was given that requires knowledge of
 a remote node's topology. However, no topology info is
 available for the following node:
 
   Node: %s
 
 The job cannot be executed under this condition. Please either
-remove the nperxxx directive and specify the number of processes
-to use, or investigate the lack of topology info.
+remove the directive or investigate the lack of topology info.
 #
 [rmaps:no-available-cpus]
 While computing bindings, we found no available cpus on

orte/mca/rmaps/base/rmaps_base_binding.c

@@ -512,6 +512,11 @@ static int bind_in_place(orte_job_t *jdata,
 
 int orte_rmaps_base_compute_bindings(orte_job_t *jdata)
 {
+    if (OPAL_BIND_TO_CPUSET == OPAL_GET_BINDING_POLICY(jdata->map->binding)) {
+        /* user specified binding by rankfile - nothing for us to do */
+        return ORTE_SUCCESS;
+    }
+
     if (!OPAL_BINDING_POLICY_IS_SET(jdata->map->binding) ||
         OPAL_BIND_TO_NONE == OPAL_GET_BINDING_POLICY(jdata->map->binding)) {
         /* no binding requested */

orte/mca/rmaps/rank_file/rmaps_rank_file_component.c

@@ -107,6 +107,8 @@ static int orte_rmaps_rank_file_open(void)
         }
         ORTE_SET_MAPPING_POLICY(orte_rmaps_base.mapping, ORTE_MAPPING_BYUSER);
         ORTE_SET_MAPPING_DIRECTIVE(orte_rmaps_base.mapping, ORTE_MAPPING_GIVEN);
+        /* we are going to bind to cpuset since the user is specifying the cpus */
+        OPAL_SET_BINDING_POLICY(opal_hwloc_binding_policy, OPAL_BIND_TO_CPUSET);
         /* make us first */
         my_priority = 10000;
     }