[OpenMP] Refined the logic to give a regular task from a hidden helper task

In current implementation, if a regular task depends on a hidden helper task,
and when the hidden helper task is releasing its dependences, it directly calls
`__kmp_omp_task`. This could cause a problem that if `__kmp_push_task` returns
`TASK_NOT_PUSHED`, the task will be executed immediately. However, the hidden
helper threads are assumed to only execute hidden helper tasks. This could cause
problems because when calling `__kmp_omp_task`, the encountering gtid, which is
not the real one of the thread, is passed.

This patch uses `__kmp_give_task`, but because it is a static function, a new
wrapper `__kmpc_give_task` is added.

Reviewed By: AndreyChurbanov

Differential Revision: https://reviews.llvm.org/D106572

Author: shiltian

openmp/runtime/src/kmp_taskdeps.h

@@ -85,6 +85,8 @@ static inline void __kmp_dephash_free(kmp_info_t *thread, kmp_dephash_t *h) {
 #endif
 }
 
+extern void __kmpc_give_task(kmp_task_t *ptask, kmp_int32 start);
+
 static inline void __kmp_release_deps(kmp_int32 gtid, kmp_taskdata_t *task) {
   kmp_info_t *thread = __kmp_threads[gtid];
   kmp_depnode_t *node = task->td_depnode;
@@ -143,7 +145,9 @@ static inline void __kmp_release_deps(kmp_int32 gtid, kmp_taskdata_t *task) {
           // encountering thread's queue; otherwise, it can be pushed to its own
           // queue.
           if (!next_taskdata->td_flags.hidden_helper) {
-            __kmp_omp_task(task->encountering_gtid, successor->dn.task, false);
+            __kmpc_give_task(
+                successor->dn.task,
+                __kmp_tid_from_gtid(next_taskdata->encountering_gtid));
           } else {
             __kmp_omp_task(gtid, successor->dn.task, false);
           }

openmp/runtime/src/kmp_tasking.cpp

@@ -3920,26 +3920,10 @@ void __kmpc_proxy_task_completed(kmp_int32 gtid, kmp_task_t *ptask) {
             gtid, taskdata));
 }
 
-/*!
-@ingroup TASKING
-@param ptask Task which execution is completed
-
-Execute the completion of a proxy task from a thread that could not belong to
-the team.
-*/
-void __kmpc_proxy_task_completed_ooo(kmp_task_t *ptask) {
+void __kmpc_give_task(kmp_task_t *ptask, kmp_int32 start = 0) {
   KMP_DEBUG_ASSERT(ptask != NULL);
   kmp_taskdata_t *taskdata = KMP_TASK_TO_TASKDATA(ptask);
 
-  KA_TRACE(
-      10,
-      ("__kmp_proxy_task_completed_ooo(enter): proxy task completing ooo %p\n",
-       taskdata));
-
-  KMP_DEBUG_ASSERT(taskdata->td_flags.proxy == TASK_PROXY);
-
-  __kmp_first_top_half_finish_proxy(taskdata);
-
   // Enqueue task to complete bottom half completion from a thread within the
   // corresponding team
   kmp_team_t *team = taskdata->td_team;
@@ -3948,7 +3932,7 @@ void __kmpc_proxy_task_completed_ooo(kmp_task_t *ptask) {
 
   // This should be similar to start_k = __kmp_get_random( thread ) % nthreads
   // but we cannot use __kmp_get_random here
-  kmp_int32 start_k = 0;
+  kmp_int32 start_k = start;
   kmp_int32 pass = 1;
   kmp_int32 k = start_k;
 
@@ -3962,6 +3946,29 @@ void __kmpc_proxy_task_completed_ooo(kmp_task_t *ptask) {
       pass = pass << 1;
 
   } while (!__kmp_give_task(thread, k, ptask, pass));
+}
+
+/*!
+@ingroup TASKING
+@param ptask Task which execution is completed
+
+Execute the completion of a proxy task from a thread that could not belong to
+the team.
+*/
+void __kmpc_proxy_task_completed_ooo(kmp_task_t *ptask) {
+  KMP_DEBUG_ASSERT(ptask != NULL);
+  kmp_taskdata_t *taskdata = KMP_TASK_TO_TASKDATA(ptask);
+
+  KA_TRACE(
+      10,
+      ("__kmp_proxy_task_completed_ooo(enter): proxy task completing ooo %p\n",
+       taskdata));
+
+  KMP_DEBUG_ASSERT(taskdata->td_flags.proxy == TASK_PROXY);
+
+  __kmp_first_top_half_finish_proxy(taskdata);
+
+  __kmpc_give_task(ptask);
 
   __kmp_second_top_half_finish_proxy(taskdata);