Index: runtime/src/kmp_runtime.cpp
===================================================================
--- runtime/src/kmp_runtime.cpp
+++ runtime/src/kmp_runtime.cpp
@@ -3890,11 +3890,11 @@
           ompt_thread_initial, __ompt_get_thread_data_internal());
     }
     ompt_data_t *task_data;
-    __ompt_get_task_info_internal(0, NULL, &task_data, NULL, NULL, NULL);
-    if (ompt_enabled.ompt_callback_task_create) {
-      ompt_callbacks.ompt_callback(ompt_callback_task_create)(
-          NULL, NULL, task_data, ompt_task_initial, 0, NULL);
-      // initial task has nothing to return to
+    ompt_data_t *parallel_data;
+    __ompt_get_task_info_internal(0, NULL, &task_data, NULL, &parallel_data, NULL);
+    if (ompt_enabled.ompt_callback_implicit_task) {
+      ompt_callbacks.ompt_callback(ompt_callback_implicit_task)(
+          ompt_scope_begin, parallel_data, task_data, 1, 1, ompt_task_initial);
     }
 
     ompt_set_thread_state(root_thread, ompt_state_work_serial);
@@ -3984,6 +3984,13 @@
 #endif /* KMP_OS_WINDOWS */
 
 #if OMPT_SUPPORT
+  ompt_data_t *task_data;
+  ompt_data_t *parallel_data;
+  __ompt_get_task_info_internal(0, NULL, &task_data, NULL, &parallel_data, NULL);
+  if (ompt_enabled.ompt_callback_implicit_task) {
+    ompt_callbacks.ompt_callback(ompt_callback_implicit_task)(
+        ompt_scope_end, parallel_data, task_data, 1, 1, ompt_task_initial);
+  }
   if (ompt_enabled.ompt_callback_thread_end) {
     ompt_callbacks.ompt_callback(ompt_callback_thread_end)(
         &(root->r.r_uber_thread->th.ompt_thread_info.thread_data));
Index: runtime/src/ompt-general.cpp
===================================================================
--- runtime/src/ompt-general.cpp
+++ runtime/src/ompt-general.cpp
@@ -361,10 +361,11 @@
           ompt_thread_initial, __ompt_get_thread_data_internal());
     }
     ompt_data_t *task_data;
-    __ompt_get_task_info_internal(0, NULL, &task_data, NULL, NULL, NULL);
-    if (ompt_enabled.ompt_callback_task_create) {
-      ompt_callbacks.ompt_callback(ompt_callback_task_create)(
-          NULL, NULL, task_data, ompt_task_initial, 0, NULL);
+    ompt_data_t *parallel_data;
+    __ompt_get_task_info_internal(0, NULL, &task_data, NULL, &parallel_data, NULL);
+    if (ompt_enabled.ompt_callback_implicit_task) {
+      ompt_callbacks.ompt_callback(ompt_callback_implicit_task)(
+          ompt_scope_begin, parallel_data, task_data, 1, 1, ompt_task_initial);
     }
 
     ompt_set_thread_state(root_thread, ompt_state_work_serial);
Index: runtime/test/ompt/callback.h
===================================================================
--- runtime/test/ompt/callback.h
+++ runtime/test/ompt/callback.h
@@ -451,10 +451,29 @@
       if(task_data->ptr)
         printf("%s\n", "0: task_data initially not null");
       task_data->value = ompt_get_unique_id();
-      printf("%" PRIu64 ": ompt_event_implicit_task_begin: parallel_id=%" PRIu64 ", task_id=%" PRIu64 ", team_size=%" PRIu32 ", thread_num=%" PRIu32 "\n", ompt_get_thread_data()->value, parallel_data->value, task_data->value, team_size, thread_num);
+
+      //there is no parallel_begin callback for implicit parallel region
+      //thus it is initialized in initial task
+      if(flags & ompt_task_initial)
+      {
+        char buffer[2048];
+
+        format_task_type(flags, buffer);
+        if(parallel_data->ptr)
+          printf("%s\n", "0: parallel_data initially not null");
+        parallel_data->value = ompt_get_unique_id();
+        printf("%" PRIu64 ": ompt_event_task_create: parent_task_id=%" PRIu64 ", parent_task_frame.exit=%p, parent_task_frame.reenter=%p, new_task_id=%" PRIu64 ", codeptr_ra=%p, task_type=%s=%d, has_dependences=%s\n", ompt_get_thread_data()->value, 0 , NULL, NULL, task_data->value, NULL, buffer, flags, "no");
+      } else {
+        printf("%" PRIu64 ": ompt_event_implicit_task_begin: parallel_id=%" PRIu64 ", task_id=%" PRIu64 ", team_size=%" PRIu32 ", thread_num=%" PRIu32 "\n", ompt_get_thread_data()->value, parallel_data->value, task_data->value, team_size, thread_num);
+      }
+
       break;
     case ompt_scope_end:
-      printf("%" PRIu64 ": ompt_event_implicit_task_end: parallel_id=%" PRIu64 ", task_id=%" PRIu64 ", team_size=%" PRIu32 ", thread_num=%" PRIu32 "\n", ompt_get_thread_data()->value, (parallel_data)?parallel_data->value:0, task_data->value, team_size, thread_num);
+      if(flags & ompt_task_initial){
+        printf("%" PRIu64 ": ompt_event_initial_task_end: parallel_id=%" PRIu64 ", task_id=%" PRIu64 ", team_size=%" PRIu32 ", thread_num=%" PRIu32 "\n", ompt_get_thread_data()->value, (parallel_data)?parallel_data->value:0, task_data->value, team_size, thread_num);
+      } else {
+        printf("%" PRIu64 ": ompt_event_implicit_task_end: parallel_id=%" PRIu64 ", task_id=%" PRIu64 ", team_size=%" PRIu32 ", thread_num=%" PRIu32 "\n", ompt_get_thread_data()->value, (parallel_data)?parallel_data->value:0, task_data->value, team_size, thread_num);
+      }
       break;
   }
 }
@@ -627,17 +646,6 @@
 
   format_task_type(type, buffer);
 
-  //there is no parallel_begin callback for implicit parallel region
-  //thus it is initialized in initial task
-  if(type & ompt_task_initial)
-  {
-    ompt_data_t *parallel_data;
-    ompt_get_parallel_info(0, &parallel_data, NULL);
-    if(parallel_data->ptr)
-      printf("%s\n", "0: parallel_data initially not null");
-    parallel_data->value = ompt_get_unique_id();
-  }
-
   printf("%" PRIu64 ": ompt_event_task_create: parent_task_id=%" PRIu64 ", parent_task_frame.exit=%p, parent_task_frame.reenter=%p, new_task_id=%" PRIu64 ", codeptr_ra=%p, task_type=%s=%d, has_dependences=%s\n", ompt_get_thread_data()->value, encountering_task_data ? encountering_task_data->value : 0, encountering_task_frame ? encountering_task_frame->exit_frame.ptr : NULL, encountering_task_frame ? encountering_task_frame->enter_frame.ptr : NULL, new_task_data->value, codeptr_ra, buffer, type, has_dependences ? "yes" : "no");
 }
 
Index: runtime/test/ompt/misc/interoperability.cpp
===================================================================
--- runtime/test/ompt/misc/interoperability.cpp
+++ runtime/test/ompt/misc/interoperability.cpp
@@ -75,6 +75,10 @@
 // CHECK-SAME: parallel_id=[[PARALLEL_ID_1]], task_id=[[PARENT_TASK_ID_1]]
 // CHECK-SAME: invoker={{[0-9]+}}
 
+// CHECK: {{^}}[[MASTER_ID_1]]: ompt_event_initial_task_end:
+// CHECK-SAME: parallel_id={{[0-9]+}}, task_id=[[PARENT_TASK_ID_1]],
+// CHECK-SAME: team_size=1, thread_num=1
+
 // CHECK: {{^}}[[MASTER_ID_1]]: ompt_event_thread_end:
 // CHECK-SAME: thread_id=[[MASTER_ID_1]]
 
@@ -101,6 +105,10 @@
 // CHECK-SAME: parallel_id=[[PARALLEL_ID_2]], task_id=[[PARENT_TASK_ID_2]]
 // CHECK-SAME: invoker={{[0-9]+}}
 
+// CHECK: {{^}}[[MASTER_ID_2]]: ompt_event_initial_task_end:
+// CHECK-SAME: parallel_id={{[0-9]+}}, task_id=[[PARENT_TASK_ID_2]],
+// CHECK-SAME: team_size=1, thread_num=1
+
 // CHECK: {{^}}[[MASTER_ID_2]]: ompt_event_thread_end:
 // CHECK-SAME: thread_id=[[MASTER_ID_2]]