Ensure that ORTE processes such as mpirun and orted never inadvertently bind themselves to cores. Change the mca param name used by the rank_file mapper to get user directives on slot lists to be different from that used by MPI procs to discover their binding. Add a cmd line option to orterun to make it easier for a user to specify the slot list (basically, hide the mca param name).

Discussed and reviewed with Lenny and Jeff.

This commit was SVN r19062.

orte/mca/rmaps/base/base.h

@@ -71,6 +71,8 @@ typedef struct {
     bool display_map;
     /* balance load across nodes */
     bool loadbalance;
+    /* slot list, if provided by user */
+    char *slot_list;
 } orte_rmaps_base_t;
 
 /**

orte/mca/rmaps/base/rmaps_base_open.c

@@ -127,6 +127,11 @@ int orte_rmaps_base_open(void)
         }
     }
     
+    /* did the user provide a slot list? */
+    param = mca_base_param_reg_string_name("rmaps", "base_slot_list",
+                                           "List of processor IDs to bind MPI processes to (e.g., used in conjunction with rank files) [default=NULL]",
+                                           false, false, NULL, &orte_rmaps_base.slot_list);
+
     /* Should we schedule on the local node or not? */
 
     mca_base_param_reg_int_name("rmaps", "base_no_schedule_local",

orte/mca/rmaps/rank_file/rmaps_rank_file.c

@@ -190,11 +190,8 @@ static int map_app_by_node(
         }
         /* Allocate a slot on this node */
         node = (orte_node_t*) cur_node_item;
-        if ( NULL != orte_mca_rmaps_rank_file_slot_list){
-            node->slot_list = (char*) malloc(64*sizeof(char));
-            if ( NULL != node->slot_list ) {
-                strcpy(node->slot_list, orte_mca_rmaps_rank_file_slot_list);
-            }
+        if ( NULL != orte_rmaps_base.slot_list ) {
+            node->slot_list = strdup(orte_rmaps_base.slot_list);
         }
         if (ORTE_SUCCESS != (rc = orte_rmaps_base_claim_slot(jdata, node, vpid_start + num_alloc, app->idx,
                                              nodes, jdata->map->oversubscribe, true))) {
@@ -300,11 +297,8 @@ static int map_app_by_slot(
                 ++num_alloc;
                 continue;
             }
-            if ( NULL != orte_mca_rmaps_rank_file_slot_list){
-                node->slot_list = (char*) malloc(64*sizeof(char));
-                if ( NULL != node->slot_list ) {
-                    strcpy(node->slot_list, orte_mca_rmaps_rank_file_slot_list);
-                }
+            if ( NULL != orte_rmaps_base.slot_list ) {
+                node->slot_list = strdup(orte_rmaps_base.slot_list);
             }
             if (ORTE_SUCCESS != (rc = orte_rmaps_base_claim_slot(jdata, node, vpid_start + num_alloc, app->idx,
                                                  nodes, jdata->map->oversubscribe, true))) {

orte/mca/rmaps/rank_file/rmaps_rank_file_component.c

@@ -23,11 +23,12 @@
 
 #include "orte/runtime/orte_globals.h"
 #include "orte/mca/ras/ras_types.h"
+#include "orte/util/show_help.h"
 
 #include "opal/mca/base/base.h"
 #include "opal/mca/base/mca_base_param.h"
-#include "orte/util/show_help.h"
 
+#include "orte/mca/rmaps/base/base.h"
 #include "orte/mca/rmaps/base/rmaps_private.h"
 #include "orte/mca/rmaps/rank_file/rmaps_rank_file.h"
 #include "orte/mca/rmaps/rank_file/rmaps_rank_file_lex.h"
@@ -74,8 +75,6 @@ orte_rmaps_rank_file_component_t mca_rmaps_rank_file_component = {
   */
 static int orte_rmaps_rank_file_open(void)
 {
-    int index = 0;
-    
     mca_rmaps_rank_file_component.priority = 0;
     
     mca_base_param_reg_string(&mca_rmaps_rank_file_component.super.base_version,
@@ -86,15 +85,10 @@ static int orte_rmaps_rank_file_open(void)
         mca_rmaps_rank_file_component.priority = 100;
     }
     
-    index = mca_base_param_find("opal", NULL, "paffinity_base_slot_list"); 
-    if (index >= 0) { 
-        if (OPAL_SUCCESS == mca_base_param_lookup_string(index, &orte_mca_rmaps_rank_file_slot_list)) { 
-            if (NULL != orte_mca_rmaps_rank_file_slot_list) { 
-                mca_rmaps_rank_file_component.priority = 100; 
-            } 
-        } 
+    if (NULL != orte_rmaps_base.slot_list) { 
+        mca_rmaps_rank_file_component.priority = 100; 
     } 
-
+    
     return ORTE_SUCCESS;
 }
 

orte/tools/orterun/orterun.c

@@ -222,6 +222,9 @@ static opal_cmd_line_init_t cmd_line_init[] = {
     { "rmaps", "base", "n_pernode", '\0', "npernode", "npernode", 1,
         NULL, OPAL_CMD_LINE_TYPE_INT,
         "Launch n processes per node on all allocated nodes" },
+    { "rmaps", "base", "slot_list", '\0', "slot-list", "slot-list", 1,
+        NULL, OPAL_CMD_LINE_TYPE_STRING,
+        "List of processor IDs to bind MPI processes to (e.g., used in conjunction with rank files)" },
     { "rmaps", "base", "no_oversubscribe", '\0', "nooversubscribe", "nooversubscribe", 0,
       NULL, OPAL_CMD_LINE_TYPE_BOOL,
       "Nodes are not to be oversubscribed, even if the system supports such operation"},