Merging r285283:

------------------------------------------------------------------------
r285283 | achurbanov | 2016-10-27 06:43:07 -0500 (Thu, 27 Oct 2016) | 6 lines

Fixed a memory leak related to task dependencies.

Differential Revision: http://reviews.llvm.org/D25504

Patch by Alex Duran.

------------------------------------------------------------------------

llvm-svn: 287559

4 files changed, 65 insertions, 13 deletions

diff --git a/openmp/runtime/src/kmp.h b/openmp/runtime/src/kmp.h
index 6fc378b..dd73992 100644
--- a/openmp/runtime/src/kmp.h
+++ b/openmp/runtime/src/kmp.h
@@ -3327,6 +3327,8 @@ extern kmp_task_t* __kmp_task_alloc( ident_t *loc_ref, kmp_int32 gtid,
   kmp_routine_entry_t task_entry );
 extern void __kmp_init_implicit_task( ident_t *loc_ref, kmp_info_t *this_thr,
                   kmp_team_t *team, int tid, int set_curr_task );
+extern void __kmp_finish_implicit_task(kmp_info_t *this_thr);
+extern void __kmp_free_implicit_task(kmp_info_t *this_thr);
 
 int __kmp_execute_tasks_32(kmp_info_t *thread, kmp_int32 gtid, kmp_flag_32 *flag, int final_spin,
                            int *thread_finished,
@@ -3496,12 +3498,18 @@ void __kmpc_omp_task_complete( ident_t *loc_ref, kmp_int32 gtid, kmp_task_t *tas
 KMP_EXPORT void __kmpc_taskgroup( ident_t * loc, int gtid );
 KMP_EXPORT void __kmpc_end_taskgroup( ident_t * loc, int gtid );
 
-KMP_EXPORT kmp_int32 __kmpc_omp_task_with_deps ( ident_t *loc_ref, kmp_int32 gtid, kmp_task_t * new_task,
-                                                 kmp_int32 ndeps, kmp_depend_info_t *dep_list,
-                                                 kmp_int32 ndeps_noalias, kmp_depend_info_t *noalias_dep_list );
-KMP_EXPORT void __kmpc_omp_wait_deps ( ident_t *loc_ref, kmp_int32 gtid, kmp_int32 ndeps, kmp_depend_info_t *dep_list,
-                                          kmp_int32 ndeps_noalias, kmp_depend_info_t *noalias_dep_list );
-extern void __kmp_release_deps ( kmp_int32 gtid, kmp_taskdata_t *task );
+KMP_EXPORT kmp_int32 __kmpc_omp_task_with_deps(
+    ident_t *loc_ref, kmp_int32 gtid, kmp_task_t *new_task, kmp_int32 ndeps,
+    kmp_depend_info_t *dep_list, kmp_int32 ndeps_noalias,
+    kmp_depend_info_t *noalias_dep_list);
+KMP_EXPORT void __kmpc_omp_wait_deps(ident_t *loc_ref, kmp_int32 gtid,
+                                     kmp_int32 ndeps,
+                                     kmp_depend_info_t *dep_list,
+                                     kmp_int32 ndeps_noalias,
+                                     kmp_depend_info_t *noalias_dep_list);
+extern void __kmp_release_deps(kmp_int32 gtid, kmp_taskdata_t *task);
+extern void __kmp_dephash_free_entries(kmp_info_t *thread, kmp_dephash_t *h);
+extern void __kmp_dephash_free(kmp_info_t *thread, kmp_dephash_t *h);
 
 extern kmp_int32 __kmp_omp_task( kmp_int32 gtid, kmp_task_t * new_task, bool serialize_immediate );
 
diff --git a/openmp/runtime/src/kmp_runtime.c b/openmp/runtime/src/kmp_runtime.c
index 7b2ef06..77b8c07 100644
--- a/openmp/runtime/src/kmp_runtime.c
+++ b/openmp/runtime/src/kmp_runtime.c
@@ -5743,6 +5743,7 @@ __kmp_reap_thread(
     }; // if
 #endif /* KMP_AFFINITY_SUPPORTED */
 
+    __kmp_free_implicit_task(thread);
     __kmp_reap_team( thread->th.th_serial_team );
     thread->th.th_serial_team = NULL;
     __kmp_free( thread );
@@ -6784,6 +6785,8 @@ __kmp_run_after_invoked_task( int gtid, int tid, kmp_info_t *this_thr,
 {
     if( __kmp_env_consistency_check )
         __kmp_pop_parallel( gtid, team->t.t_ident );
+
+    __kmp_finish_implicit_task(this_thr);
 }
 
 int
diff --git a/openmp/runtime/src/kmp_taskdeps.cpp b/openmp/runtime/src/kmp_taskdeps.cpp
index 46ba5df..47d0cd2 100644
--- a/openmp/runtime/src/kmp_taskdeps.cpp
+++ b/openmp/runtime/src/kmp_taskdeps.cpp
@@ -97,7 +97,8 @@ __kmp_dephash_create ( kmp_info_t *thread, kmp_taskdata_t *current_task )
     else
         h_size = KMP_DEPHASH_OTHER_SIZE;
 
-    kmp_int32 size = h_size * sizeof(kmp_dephash_entry_t) + sizeof(kmp_dephash_t);
+    kmp_int32 size =
+        h_size * sizeof(kmp_dephash_entry_t *) + sizeof(kmp_dephash_t);
 
 #if USE_FAST_MEMORY
     h = (kmp_dephash_t *) __kmp_fast_allocate( thread, size );
@@ -118,13 +119,13 @@ __kmp_dephash_create ( kmp_info_t *thread, kmp_taskdata_t *current_task )
     return h;
 }
 
-static void
-__kmp_dephash_free ( kmp_info_t *thread, kmp_dephash_t *h )
+void
+__kmp_dephash_free_entries(kmp_info_t *thread, kmp_dephash_t *h)
 {
-    for ( size_t i=0; i < h->size; i++ ) {
-        if ( h->buckets[i] ) {
+    for (size_t i = 0; i < h->size; i++) {
+        if (h->buckets[i]) {
             kmp_dephash_entry_t *next;
-            for ( kmp_dephash_entry_t *entry = h->buckets[i]; entry; entry = next ) {
+            for (kmp_dephash_entry_t *entry = h->buckets[i]; entry; entry = next) {
                 next = entry->next_in_bucket;
                 __kmp_depnode_list_free(thread,entry->last_ins);
                 __kmp_node_deref(thread,entry->last_out);
@@ -134,8 +135,15 @@ __kmp_dephash_free ( kmp_info_t *thread, kmp_dephash_t *h )
                 __kmp_thread_free(thread,entry);
 #endif
             }
+            h->buckets[i] = 0;
         }
     }
+}
+
+void
+__kmp_dephash_free(kmp_info_t *thread, kmp_dephash_t *h)
+{
+    __kmp_dephash_free_entries(thread, h);
 #if USE_FAST_MEMORY
     __kmp_fast_free(thread,h);
 #else
diff --git a/openmp/runtime/src/kmp_tasking.c b/openmp/runtime/src/kmp_tasking.c
index ce1b63d..9a64045 100644
--- a/openmp/runtime/src/kmp_tasking.c
+++ b/openmp/runtime/src/kmp_tasking.c
@@ -852,7 +852,6 @@ __kmp_init_implicit_task( ident_t *loc_ref, kmp_info_t *this_thr, kmp_team_t *te
     task->td_flags.freed       = 0;
 
 #if OMP_40_ENABLED
-    task->td_dephash = NULL;
     task->td_depnode = NULL;
 #endif
 
@@ -861,6 +860,7 @@ __kmp_init_implicit_task( ident_t *loc_ref, kmp_info_t *this_thr, kmp_team_t *te
         task->td_allocated_child_tasks  = 0; // Not used because do not need to deallocate implicit task
 #if OMP_40_ENABLED
         task->td_taskgroup = NULL;           // An implicit task does not have taskgroup
+        task->td_dephash = NULL;
 #endif
         __kmp_push_current_task_to_thread( this_thr, team, tid );
     } else {
@@ -876,6 +876,39 @@ __kmp_init_implicit_task( ident_t *loc_ref, kmp_info_t *this_thr, kmp_team_t *te
                   tid, team, task ) );
 }
 
+
+//-----------------------------------------------------------------------------
+//// __kmp_finish_implicit_task: Release resources associated to implicit tasks
+//// at the end of parallel regions. Some resources are kept for reuse in the
+//// next parallel region.
+////
+//// thread:  thread data structure corresponding to implicit task
+//
+void
+__kmp_finish_implicit_task(kmp_info_t *thread)
+{
+    kmp_taskdata_t *task = thread->th.th_current_task;
+    if (task->td_dephash)
+        __kmp_dephash_free_entries(thread, task->td_dephash);
+}
+
+
+//-----------------------------------------------------------------------------
+//// __kmp_free_implicit_task: Release resources associated to implicit tasks
+//// when these are destroyed regions
+////
+//// thread:  thread data structure corresponding to implicit task
+//
+void
+__kmp_free_implicit_task(kmp_info_t *thread)
+{
+    kmp_taskdata_t *task = thread->th.th_current_task;
+    if (task->td_dephash)
+        __kmp_dephash_free(thread, task->td_dephash);
+    task->td_dephash = NULL;
+}
+
+
 // Round up a size to a power of two specified by val
 // Used to insert padding between structures co-allocated using a single malloc() call
 static size_t