diff --git a/openmp/runtime/CMakeLists.txt b/openmp/runtime/CMakeLists.txt
index 93eb14f10a50a..76f9f1d5840c8 100644
--- a/openmp/runtime/CMakeLists.txt
+++ b/openmp/runtime/CMakeLists.txt
@@ -391,9 +391,9 @@ if(LIBOMP_OMPD_SUPPORT AND ((NOT LIBOMP_OMPT_SUPPORT) OR (NOT "${CMAKE_SYSTEM_NA
   set(LIBOMP_OMPD_SUPPORT FALSE)
 endif()
 
-# OMPX Taskgraph support
-# Whether to build with OMPX Taskgraph (e.g. task record & replay)
-set(LIBOMP_OMPX_TASKGRAPH FALSE CACHE BOOL "OMPX-taskgraph (task record & replay)?")
+# OMP Taskgraph support
+# Whether to build with OMP Taskgraph (e.g. task record & replay)
+set(LIBOMP_TASKGRAPH_EXPERIMENTAL FALSE CACHE BOOL "Experimental OMP taskgraph (task record & replay)")
 
 # Error check hwloc support after config-ix has run
 if(LIBOMP_USE_HWLOC AND (NOT LIBOMP_HAVE_HWLOC))
@@ -464,7 +464,7 @@ if(${OPENMP_STANDALONE_BUILD})
   libomp_say("Use Adaptive locks   -- ${LIBOMP_USE_ADAPTIVE_LOCKS}")
   libomp_say("Use quad precision   -- ${LIBOMP_USE_QUAD_PRECISION}")
   libomp_say("Use Hwloc library    -- ${LIBOMP_USE_HWLOC}")
-  libomp_say("Use OMPX-taskgraph   -- ${LIBOMP_OMPX_TASKGRAPH}")
+  libomp_say("Use OMP taskgraph    -- ${LIBOMP_TASKGRAPH_EXPERIMENTAL}")
 endif()
 
 add_subdirectory(src)
diff --git a/openmp/runtime/src/kmp.h b/openmp/runtime/src/kmp.h
index 83afc0e83f231..2c89bee9f8c2b 100644
--- a/openmp/runtime/src/kmp.h
+++ b/openmp/runtime/src/kmp.h
@@ -2627,7 +2627,7 @@ typedef struct {
   } ed;
 } kmp_event_t;
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
 // Initial number of allocated nodes while recording
 #define INIT_MAPSIZE 50
 
@@ -2678,7 +2678,7 @@ typedef struct kmp_tdg_info {
 extern int __kmp_tdg_dot;
 extern kmp_int32 __kmp_max_tdgs;
 extern kmp_tdg_info_t **__kmp_global_tdgs;
-extern kmp_int32 __kmp_curr_tdg_idx;
+extern kmp_int32 __kmp_curr_tdg_id;
 extern kmp_int32 __kmp_successors_size;
 extern std::atomic<kmp_int32> __kmp_tdg_task_id;
 extern kmp_int32 __kmp_num_tdg;
@@ -2687,7 +2687,7 @@ extern kmp_int32 __kmp_num_tdg;
 typedef struct kmp_tasking_flags { /* Total struct must be exactly 32 bits */
 #if defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __ORDER_BIG_ENDIAN__)
   /* Same fields as in the #else branch, but in reverse order */
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   unsigned reserved31 : 4;
   unsigned onced : 1;
 #else
@@ -2746,7 +2746,7 @@ typedef struct kmp_tasking_flags { /* Total struct must be exactly 32 bits */
   unsigned native : 1; /* 1==gcc-compiled task, 0==intel */
   unsigned target : 1;
   unsigned hidden_helper : 1; /* 1 == hidden helper task */
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   unsigned onced : 1; /* 1==ran once already, 0==never ran, record & replay purposes */
   unsigned reserved31 : 4; /* reserved for library use */
 #else
@@ -2801,7 +2801,7 @@ struct kmp_taskdata { /* aligned during dynamic allocation       */
 #if OMPT_SUPPORT
   ompt_task_info_t ompt_task_info;
 #endif
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   bool is_taskgraph = 0; // whether the task is within a TDG
   kmp_tdg_info_t *tdg; // used to associate task with a TDG
   kmp_int32 td_tdg_task_id; // local task id in its TDG
@@ -4379,7 +4379,7 @@ KMP_EXPORT void __kmpc_init_nest_lock_with_hint(ident_t *loc, kmp_int32 gtid,
                                                 void **user_lock,
                                                 uintptr_t hint);
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
 // Taskgraph's Record & Replay mechanism
 // __kmp_tdg_is_recording: check whether a given TDG is recording
 // status: the tdg's current status
@@ -4392,6 +4392,9 @@ KMP_EXPORT kmp_int32 __kmpc_start_record_task(ident_t *loc, kmp_int32 gtid,
                                               kmp_int32 tdg_id);
 KMP_EXPORT void __kmpc_end_record_task(ident_t *loc, kmp_int32 gtid,
                                        kmp_int32 input_flags, kmp_int32 tdg_id);
+KMP_EXPORT void __kmpc_taskgraph(ident_t *loc_ref, kmp_int32 gtid,
+                                 kmp_int32 input_flags, kmp_uint32 tdg_id,
+                                 void (*entry)(void *), void *args);
 #endif
 /* Interface to fast scalable reduce methods routines */
 
diff --git a/openmp/runtime/src/kmp_config.h.cmake b/openmp/runtime/src/kmp_config.h.cmake
index 40f1087fd7f27..1f966008c60a5 100644
--- a/openmp/runtime/src/kmp_config.h.cmake
+++ b/openmp/runtime/src/kmp_config.h.cmake
@@ -46,8 +46,8 @@
 #define OMPT_SUPPORT LIBOMP_OMPT_SUPPORT
 #cmakedefine01 LIBOMP_OMPD_SUPPORT
 #define OMPD_SUPPORT LIBOMP_OMPD_SUPPORT
-#cmakedefine01 LIBOMP_OMPX_TASKGRAPH
-#define OMPX_TASKGRAPH LIBOMP_OMPX_TASKGRAPH
+#cmakedefine01 LIBOMP_TASKGRAPH_EXPERIMENTAL
+#define OMP_TASKGRAPH_EXPERIMENTAL LIBOMP_TASKGRAPH_EXPERIMENTAL
 #cmakedefine01 LIBOMP_PROFILING_SUPPORT
 #define OMP_PROFILING_SUPPORT LIBOMP_PROFILING_SUPPORT
 #cmakedefine01 LIBOMP_OMPT_OPTIONAL
diff --git a/openmp/runtime/src/kmp_global.cpp b/openmp/runtime/src/kmp_global.cpp
index 323d13e948b42..bcc318a180995 100644
--- a/openmp/runtime/src/kmp_global.cpp
+++ b/openmp/runtime/src/kmp_global.cpp
@@ -551,12 +551,12 @@ int __kmp_nesting_mode = 0;
 int __kmp_nesting_mode_nlevels = 1;
 int *__kmp_nesting_nth_level;
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
 // TDG record & replay
 int __kmp_tdg_dot = 0;
 kmp_int32 __kmp_max_tdgs = 100;
 kmp_tdg_info_t **__kmp_global_tdgs = NULL;
-kmp_int32 __kmp_curr_tdg_idx =
+kmp_int32 __kmp_curr_tdg_id =
     0; // Id of the current TDG being recorded or executed
 kmp_int32 __kmp_num_tdg = 0;
 kmp_int32 __kmp_successors_size = 10; // Initial succesor size list for
diff --git a/openmp/runtime/src/kmp_settings.cpp b/openmp/runtime/src/kmp_settings.cpp
index b9d615f43b570..36d8d0ffa8bb3 100644
--- a/openmp/runtime/src/kmp_settings.cpp
+++ b/openmp/runtime/src/kmp_settings.cpp
@@ -1266,7 +1266,7 @@ static void __kmp_stg_parse_num_threads(char const *name, char const *value,
   K_DIAG(1, ("__kmp_dflt_team_nth == %d\n", __kmp_dflt_team_nth));
 } // __kmp_stg_parse_num_threads
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
 static void __kmp_stg_parse_max_tdgs(char const *name, char const *value,
                                      void *data) {
   __kmp_stg_parse_int(name, value, 0, INT_MAX, &__kmp_max_tdgs);
@@ -5742,7 +5742,7 @@ static kmp_setting_t __kmp_stg_table[] = {
     {"LIBOMP_NUM_HIDDEN_HELPER_THREADS",
      __kmp_stg_parse_num_hidden_helper_threads,
      __kmp_stg_print_num_hidden_helper_threads, NULL, 0, 0},
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
     {"KMP_MAX_TDGS", __kmp_stg_parse_max_tdgs, __kmp_std_print_max_tdgs, NULL,
      0, 0},
     {"KMP_TDG_DOT", __kmp_stg_parse_tdg_dot, __kmp_stg_print_tdg_dot, NULL, 0,
diff --git a/openmp/runtime/src/kmp_taskdeps.cpp b/openmp/runtime/src/kmp_taskdeps.cpp
index abbca752f0587..b1a0848fc722f 100644
--- a/openmp/runtime/src/kmp_taskdeps.cpp
+++ b/openmp/runtime/src/kmp_taskdeps.cpp
@@ -222,7 +222,7 @@ static kmp_depnode_list_t *__kmp_add_node(kmp_info_t *thread,
 static inline void __kmp_track_dependence(kmp_int32 gtid, kmp_depnode_t *source,
                                           kmp_depnode_t *sink,
                                           kmp_task_t *sink_task) {
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   kmp_taskdata_t *task_source = KMP_TASK_TO_TASKDATA(source->dn.task);
   kmp_taskdata_t *task_sink = KMP_TASK_TO_TASKDATA(sink_task);
   if (source->dn.task && sink_task) {
@@ -311,7 +311,7 @@ __kmp_depnode_link_successor(kmp_int32 gtid, kmp_info_t *thread,
   // link node as successor of list elements
   for (kmp_depnode_list_t *p = plist; p; p = p->next) {
     kmp_depnode_t *dep = p->node;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
     kmp_tdg_status tdg_status = KMP_TDG_NONE;
     if (task) {
       kmp_taskdata_t *td = KMP_TASK_TO_TASKDATA(task);
@@ -325,7 +325,7 @@ __kmp_depnode_link_successor(kmp_int32 gtid, kmp_info_t *thread,
       KMP_ACQUIRE_DEPNODE(gtid, dep);
       if (dep->dn.task) {
         if (!dep->dn.successors || dep->dn.successors->node != node) {
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
           if (!(__kmp_tdg_is_recording(tdg_status)) && task)
 #endif
             __kmp_track_dependence(gtid, dep, node, task);
@@ -352,7 +352,7 @@ static inline kmp_int32 __kmp_depnode_link_successor(kmp_int32 gtid,
   if (!sink)
     return 0;
   kmp_int32 npredecessors = 0;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   kmp_tdg_status tdg_status = KMP_TDG_NONE;
   kmp_taskdata_t *td = KMP_TASK_TO_TASKDATA(task);
   if (task) {
@@ -367,7 +367,7 @@ static inline kmp_int32 __kmp_depnode_link_successor(kmp_int32 gtid,
     KMP_ACQUIRE_DEPNODE(gtid, sink);
     if (sink->dn.task) {
       if (!sink->dn.successors || sink->dn.successors->node != source) {
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
         if (!(__kmp_tdg_is_recording(tdg_status)) && task)
 #endif
           __kmp_track_dependence(gtid, sink, source, task);
@@ -376,7 +376,7 @@ static inline kmp_int32 __kmp_depnode_link_successor(kmp_int32 gtid,
                     "%p\n",
                     gtid, KMP_TASK_TO_TASKDATA(sink->dn.task),
                     KMP_TASK_TO_TASKDATA(task)));
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
         if (__kmp_tdg_is_recording(tdg_status)) {
           kmp_taskdata_t *tdd = KMP_TASK_TO_TASKDATA(sink->dn.task);
           if (tdd->is_taskgraph) {
@@ -694,7 +694,7 @@ kmp_int32 __kmpc_omp_task_with_deps(ident_t *loc_ref, kmp_int32 gtid,
   kmp_info_t *thread = __kmp_threads[gtid];
   kmp_taskdata_t *current_task = thread->th.th_current_task;
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   // record TDG with deps
   if (new_taskdata->is_taskgraph &&
       __kmp_tdg_is_recording(new_taskdata->tdg->tdg_status)) {
@@ -714,7 +714,7 @@ kmp_int32 __kmpc_omp_task_with_deps(ident_t *loc_ref, kmp_int32 gtid,
 
         __kmp_free(old_record);
 
-        for (kmp_int i = old_size; i < new_size; i++) {
+        for (kmp_uint i = old_size; i < new_size; i++) {
           kmp_int32 *successorsList = (kmp_int32 *)__kmp_allocate(
               __kmp_successors_size * sizeof(kmp_int32));
           new_record[i].task = nullptr;
diff --git a/openmp/runtime/src/kmp_taskdeps.h b/openmp/runtime/src/kmp_taskdeps.h
index f6bfb39218a21..0792baf67f162 100644
--- a/openmp/runtime/src/kmp_taskdeps.h
+++ b/openmp/runtime/src/kmp_taskdeps.h
@@ -96,7 +96,7 @@ 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) {
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if (task->is_taskgraph && !(__kmp_tdg_is_recording(task->tdg->tdg_status))) {
     kmp_node_info_t *TaskInfo = &(task->tdg->record_map[task->td_tdg_task_id]);
 
@@ -140,7 +140,7 @@ static inline void __kmp_release_deps(kmp_int32 gtid, kmp_taskdata_t *task) {
                 gtid, task));
 
   KMP_ACQUIRE_DEPNODE(gtid, node);
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if (!task->is_taskgraph ||
       (task->is_taskgraph && !__kmp_tdg_is_recording(task->tdg->tdg_status)))
 #endif
diff --git a/openmp/runtime/src/kmp_tasking.cpp b/openmp/runtime/src/kmp_tasking.cpp
index 37836fb457537..fd39a4b2712fc 100644
--- a/openmp/runtime/src/kmp_tasking.cpp
+++ b/openmp/runtime/src/kmp_tasking.cpp
@@ -37,7 +37,7 @@ static void __kmp_alloc_task_deque(kmp_info_t *thread,
 static int __kmp_realloc_task_threads_data(kmp_info_t *thread,
                                            kmp_task_team_t *task_team);
 static void __kmp_bottom_half_finish_proxy(kmp_int32 gtid, kmp_task_t *ptask);
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
 static kmp_tdg_info_t *__kmp_find_tdg(kmp_int32 tdg_id);
 int __kmp_taskloop_task(int gtid, void *ptask);
 #endif
@@ -70,7 +70,7 @@ static bool __kmp_task_is_allowed(int gtid, const kmp_int32 is_constrained,
   }
   // Check mutexinoutset dependencies, acquire locks
   kmp_depnode_t *node = tasknew->td_depnode;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if (!tasknew->is_taskgraph && UNLIKELY(node && (node->dn.mtx_num_locks > 0))) {
 #else
   if (UNLIKELY(node && (node->dn.mtx_num_locks > 0))) {
@@ -665,7 +665,7 @@ static void __kmp_free_task(kmp_int32 gtid, kmp_taskdata_t *taskdata,
   task->data2.priority = 0;
 
   taskdata->td_flags.freed = 1;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   // do not free tasks in taskgraph
   if (!taskdata->is_taskgraph) {
 #endif
@@ -675,7 +675,7 @@ static void __kmp_free_task(kmp_int32 gtid, kmp_taskdata_t *taskdata,
 #else /* ! USE_FAST_MEMORY */
   __kmp_thread_free(thread, taskdata);
 #endif
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   } else {
     taskdata->td_flags.complete = 0;
     taskdata->td_flags.started = 0;
@@ -779,7 +779,7 @@ static bool __kmp_track_children_task(kmp_taskdata_t *taskdata) {
         flags.detachable == TASK_DETACHABLE || flags.hidden_helper;
   ret = ret ||
         KMP_ATOMIC_LD_ACQ(&taskdata->td_parent->td_incomplete_child_tasks) > 0;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if (taskdata->td_taskgroup && taskdata->is_taskgraph)
     ret = ret || KMP_ATOMIC_LD_ACQ(&taskdata->td_taskgroup->count) > 0;
 #endif
@@ -802,7 +802,7 @@ static void __kmp_task_finish(kmp_int32 gtid, kmp_task_t *task,
   kmp_info_t *thread = __kmp_threads[gtid];
   kmp_task_team_t *task_team =
       thread->th.th_task_team; // might be NULL for serial teams...
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   // to avoid seg fault when we need to access taskdata->td_flags after free when using vanilla taskloop
   bool is_taskgraph;
 #endif
@@ -815,7 +815,7 @@ static void __kmp_task_finish(kmp_int32 gtid, kmp_task_t *task,
 
   KMP_DEBUG_ASSERT(taskdata->td_flags.tasktype == TASK_EXPLICIT);
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   is_taskgraph = taskdata->is_taskgraph;
 #endif
 
@@ -923,7 +923,7 @@ static void __kmp_task_finish(kmp_int32 gtid, kmp_task_t *task,
 
   if (completed) {
     taskdata->td_flags.complete = 1; // mark the task as completed
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
     taskdata->td_flags.onced = 1; // mark the task as ran once already
 #endif
 
@@ -942,7 +942,7 @@ static void __kmp_task_finish(kmp_int32 gtid, kmp_task_t *task,
 #endif
           KMP_ATOMIC_DEC(&taskdata->td_parent->td_incomplete_child_tasks);
       KMP_DEBUG_ASSERT(children >= 0);
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
       if (taskdata->td_taskgroup && !taskdata->is_taskgraph)
 #else
       if (taskdata->td_taskgroup)
@@ -985,7 +985,7 @@ static void __kmp_task_finish(kmp_int32 gtid, kmp_task_t *task,
   // KMP_DEBUG_ASSERT( resumed_task->td_flags.executing == 0 );
   resumed_task->td_flags.executing = 1; // resume previous task
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if (is_taskgraph && __kmp_track_children_task(taskdata) &&
       taskdata->td_taskgroup) {
     // TDG: we only release taskgroup barrier here because
@@ -1113,7 +1113,7 @@ void __kmp_init_implicit_task(ident_t *loc_ref, kmp_info_t *this_thr,
   task->td_flags.executing = 1;
   task->td_flags.complete = 0;
   task->td_flags.freed = 0;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   task->td_flags.onced = 0;
 #endif
 
@@ -1159,7 +1159,7 @@ void __kmp_finish_implicit_task(kmp_info_t *thread) {
   if (task->td_dephash) {
     int children;
     task->td_flags.complete = 1;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
     task->td_flags.onced = 1;
 #endif
     children = KMP_ATOMIC_LD_ACQ(&task->td_incomplete_child_tasks);
@@ -1390,7 +1390,7 @@ kmp_task_t *__kmp_task_alloc(ident_t *loc_ref, kmp_int32 gtid,
   taskdata->td_flags.executing = 0;
   taskdata->td_flags.complete = 0;
   taskdata->td_flags.freed = 0;
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   taskdata->td_flags.onced = 0;
   taskdata->is_taskgraph = 0;
   taskdata->tdg = nullptr;
@@ -1430,12 +1430,12 @@ kmp_task_t *__kmp_task_alloc(ident_t *loc_ref, kmp_int32 gtid,
     }
   }
 
-#if OMPX_TASKGRAPH
-  kmp_tdg_info_t *tdg = __kmp_find_tdg(__kmp_curr_tdg_idx);
+#if OMP_TASKGRAPH_EXPERIMENTAL
+  kmp_tdg_info_t *tdg = __kmp_find_tdg(__kmp_curr_tdg_id);
   if (tdg && __kmp_tdg_is_recording(tdg->tdg_status) &&
       (task_entry != (kmp_routine_entry_t)__kmp_taskloop_task)) {
     taskdata->is_taskgraph = 1;
-    taskdata->tdg = __kmp_global_tdgs[__kmp_curr_tdg_idx];
+    taskdata->tdg = tdg;
     taskdata->td_task_id = KMP_GEN_TASK_ID();
     taskdata->td_tdg_task_id = KMP_ATOMIC_INC(&__kmp_tdg_task_id);
   }
@@ -1795,7 +1795,7 @@ kmp_int32 __kmp_omp_task(kmp_int32 gtid, kmp_task_t *new_task,
                          bool serialize_immediate) {
   kmp_taskdata_t *new_taskdata = KMP_TASK_TO_TASKDATA(new_task);
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if (new_taskdata->is_taskgraph &&
       __kmp_tdg_is_recording(new_taskdata->tdg->tdg_status)) {
     kmp_tdg_info_t *tdg = new_taskdata->tdg;
@@ -1816,7 +1816,7 @@ kmp_int32 __kmp_omp_task(kmp_int32 gtid, kmp_task_t *new_task,
 
         __kmp_free(old_record);
 
-        for (kmp_int i = old_size; i < new_size; i++) {
+        for (kmp_uint i = old_size; i < new_size; i++) {
           kmp_int32 *successorsList = (kmp_int32 *)__kmp_allocate(
               __kmp_successors_size * sizeof(kmp_int32));
           new_record[i].task = nullptr;
@@ -2364,10 +2364,10 @@ the reduction either does not use omp_orig object, or the omp_orig is accessible
 without help of the runtime library.
 */
 void *__kmpc_task_reduction_init(int gtid, int num, void *data) {
-#if OMPX_TASKGRAPH
-  kmp_tdg_info_t *tdg = __kmp_find_tdg(__kmp_curr_tdg_idx);
+#if OMP_TASKGRAPH_EXPERIMENTAL
+  kmp_tdg_info_t *tdg = __kmp_find_tdg(__kmp_curr_tdg_id);
   if (tdg && __kmp_tdg_is_recording(tdg->tdg_status)) {
-    kmp_tdg_info_t *this_tdg = __kmp_global_tdgs[__kmp_curr_tdg_idx];
+    kmp_tdg_info_t *this_tdg = __kmp_find_tdg(__kmp_curr_tdg_id);
     this_tdg->rec_taskred_data =
         __kmp_allocate(sizeof(kmp_task_red_input_t) * num);
     this_tdg->rec_num_taskred = num;
@@ -2391,15 +2391,12 @@ Note: this entry supposes the optional compiler-generated initializer routine
 has two parameters, pointer to object to be initialized and pointer to omp_orig
 */
 void *__kmpc_taskred_init(int gtid, int num, void *data) {
-#if OMPX_TASKGRAPH
-  kmp_tdg_info_t *tdg = __kmp_find_tdg(__kmp_curr_tdg_idx);
+#if OMP_TASKGRAPH_EXPERIMENTAL
+  kmp_tdg_info_t *tdg = __kmp_find_tdg(__kmp_curr_tdg_id);
   if (tdg && __kmp_tdg_is_recording(tdg->tdg_status)) {
-    kmp_tdg_info_t *this_tdg = __kmp_global_tdgs[__kmp_curr_tdg_idx];
-    this_tdg->rec_taskred_data =
-        __kmp_allocate(sizeof(kmp_task_red_input_t) * num);
-    this_tdg->rec_num_taskred = num;
-    KMP_MEMCPY(this_tdg->rec_taskred_data, data,
-               sizeof(kmp_task_red_input_t) * num);
+    tdg->rec_taskred_data = __kmp_allocate(sizeof(kmp_task_red_input_t) * num);
+    tdg->rec_num_taskred = num;
+    KMP_MEMCPY(tdg->rec_taskred_data, data, sizeof(kmp_task_red_input_t) * num);
   }
 #endif
   return __kmp_task_reduction_init(gtid, num, (kmp_taskred_input_t *)data);
@@ -2448,10 +2445,10 @@ void *__kmpc_task_reduction_get_th_data(int gtid, void *tskgrp, void *data) {
   kmp_int32 num;
   kmp_int32 tid = thread->th.th_info.ds.ds_tid;
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if ((thread->th.th_current_task->is_taskgraph) &&
       (!__kmp_tdg_is_recording(
-          __kmp_global_tdgs[__kmp_curr_tdg_idx]->tdg_status))) {
+          __kmp_find_tdg(__kmp_curr_tdg_id)->tdg_status))) {
     tg = thread->th.th_current_task->td_taskgroup;
     KMP_ASSERT(tg != NULL);
     KMP_ASSERT(tg->reduce_data != NULL);
@@ -4226,7 +4223,7 @@ static void __kmp_first_top_half_finish_proxy(kmp_taskdata_t *taskdata) {
   KMP_DEBUG_ASSERT(taskdata->td_flags.freed == 0);
 
   taskdata->td_flags.complete = 1; // mark the task as completed
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   taskdata->td_flags.onced = 1;
 #endif
 
@@ -4431,8 +4428,9 @@ void __kmp_fulfill_event(kmp_event_t *event) {
 //      indicating whether we need to update task->td_task_id
 // returns:  a pointer to the allocated kmp_task_t structure (task).
 kmp_task_t *__kmp_task_dup_alloc(kmp_info_t *thread, kmp_task_t *task_src
-#if OMPX_TASKGRAPH
-                                 , int taskloop_recur
+#if OMP_TASKGRAPH_EXPERIMENTAL
+                                 ,
+                                 int taskloop_recur
 #endif
 ) {
   kmp_task_t *task;
@@ -4462,7 +4460,7 @@ kmp_task_t *__kmp_task_dup_alloc(kmp_info_t *thread, kmp_task_t *task_src
   task = KMP_TASKDATA_TO_TASK(taskdata);
 
   // Initialize new task (only specific fields not affected by memcpy)
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   if (taskdata->is_taskgraph && !taskloop_recur &&
       __kmp_tdg_is_recording(taskdata_src->tdg->tdg_status))
     taskdata->td_tdg_task_id = KMP_ATOMIC_INC(&__kmp_tdg_task_id);
@@ -4695,7 +4693,7 @@ void __kmp_taskloop_linear(ident_t *loc, int gtid, kmp_task_t *task,
       }
     }
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
     next_task = __kmp_task_dup_alloc(thread, task, /* taskloop_recur */ 0);
 #else
     next_task = __kmp_task_dup_alloc(thread, task); // allocate new task
@@ -4897,7 +4895,7 @@ void __kmp_taskloop_recur(ident_t *loc, int gtid, kmp_task_t *task,
   lb1 = ub0 + st;
 
   // create pattern task for 2nd half of the loop
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   next_task = __kmp_task_dup_alloc(thread, task,
                                    /* taskloop_recur */ 1);
 #else
@@ -4935,7 +4933,7 @@ void __kmp_taskloop_recur(ident_t *loc, int gtid, kmp_task_t *task,
   p->codeptr_ra = codeptr_ra;
 #endif
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   kmp_taskdata_t *new_task_data = KMP_TASK_TO_TASKDATA(new_task);
   new_task_data->tdg = taskdata->tdg;
   new_task_data->is_taskgraph = 0;
@@ -4980,7 +4978,7 @@ static void __kmp_taskloop(ident_t *loc, int gtid, kmp_task_t *task, int if_val,
     __kmpc_taskgroup(loc, gtid);
   }
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
   KMP_ATOMIC_DEC(&__kmp_tdg_task_id);
 #endif
   // =========================================================================
@@ -5231,7 +5229,25 @@ bool __kmpc_omp_has_task_team(kmp_int32 gtid) {
   return taskdata->td_task_team != NULL;
 }
 
-#if OMPX_TASKGRAPH
+#if OMP_TASKGRAPH_EXPERIMENTAL
+// __kmpc_taskgraph: record or replay taskgraph
+// loc_ref:     Location of TDG, not used yet
+// gtid:        Global Thread ID of the encountering thread
+// input_flags: Flags associated with the TDG
+// tdg_id:      ID of the TDG to record, for now, incremental integer
+// entry:       Pointer to the entry function
+// args:        Pointer to the function arguments
+void __kmpc_taskgraph(ident_t *loc_ref, kmp_int32 gtid, kmp_int32 input_flags,
+                      kmp_uint32 tdg_id, void (*entry)(void *), void *args) {
+  kmp_int32 res = __kmpc_start_record_task(loc_ref, gtid, input_flags, tdg_id);
+  // When res = 1, we either start recording or only execute tasks
+  // without recording. Need to execute entry function in both cases.
+  if (res)
+    entry(args);
+
+  __kmpc_end_record_task(loc_ref, gtid, input_flags, tdg_id);
+}
+
 // __kmp_find_tdg: identify a TDG through its ID
 // tdg_id: ID of the TDG
 // returns: If a TDG corresponding to this ID is found and not
@@ -5245,9 +5261,14 @@ static kmp_tdg_info_t *__kmp_find_tdg(kmp_int32 tdg_id) {
     __kmp_global_tdgs = (kmp_tdg_info_t **)__kmp_allocate(
         sizeof(kmp_tdg_info_t *) * __kmp_max_tdgs);
 
-  if ((__kmp_global_tdgs[tdg_id]) &&
-      (__kmp_global_tdgs[tdg_id]->tdg_status != KMP_TDG_NONE))
-    res = __kmp_global_tdgs[tdg_id];
+  for (kmp_int32 i = 0; i < __kmp_num_tdg; ++i) {
+    if ((__kmp_global_tdgs[i]) && (__kmp_global_tdgs[i]->tdg_id == tdg_id) &&
+        (__kmp_global_tdgs[i]->tdg_status != KMP_TDG_NONE)) {
+      res = __kmp_global_tdgs[i];
+      __kmp_curr_tdg_id = tdg_id;
+      break;
+    }
+  }
   return res;
 }
 
@@ -5256,7 +5277,8 @@ static kmp_tdg_info_t *__kmp_find_tdg(kmp_int32 tdg_id) {
 // gtid:   Global Thread ID
 void __kmp_print_tdg_dot(kmp_tdg_info_t *tdg, kmp_int32 gtid) {
   kmp_int32 tdg_id = tdg->tdg_id;
-  KA_TRACE(10, ("__kmp_print_tdg_dot(enter): T#%d tdg_id=%d \n", gtid, tdg_id));
+  KA_TRACE(10, ("__kmp_print_tdg_dot(enter): T#%d tdg_id=%d \n",
+                __kmp_get_gtid(), tdg_id));
 
   char file_name[20];
   sprintf(file_name, "tdg_%d.dot", tdg_id);
@@ -5282,7 +5304,8 @@ void __kmp_print_tdg_dot(kmp_tdg_info_t *tdg, kmp_int32 gtid) {
     }
   }
   fprintf(tdg_file, "}");
-  KA_TRACE(10, ("__kmp_print_tdg_dot(exit): T#%d tdg_id=%d \n", gtid, tdg_id));
+  KA_TRACE(10, ("__kmp_print_tdg_dot(exit): T#%d tdg_id=%d \n",
+                __kmp_get_gtid(), tdg_id));
 }
 
 // __kmp_exec_tdg: launch the execution of a previous
@@ -5347,7 +5370,7 @@ static inline void __kmp_start_record(kmp_int32 gtid,
                                       kmp_int32 tdg_id) {
   kmp_tdg_info_t *tdg =
       (kmp_tdg_info_t *)__kmp_allocate(sizeof(kmp_tdg_info_t));
-  __kmp_global_tdgs[__kmp_curr_tdg_idx] = tdg;
+  __kmp_global_tdgs[__kmp_num_tdg - 1] = tdg;
   // Initializing the TDG structure
   tdg->tdg_id = tdg_id;
   tdg->map_size = INIT_MAPSIZE;
@@ -5372,7 +5395,7 @@ static inline void __kmp_start_record(kmp_int32 gtid,
     KMP_ATOMIC_ST_RLX(&this_record_map[i].npredecessors_counter, 0);
   }
 
-  __kmp_global_tdgs[__kmp_curr_tdg_idx]->record_map = this_record_map;
+  tdg->record_map = this_record_map;
 }
 
 // __kmpc_start_record_task: Wrapper around __kmp_start_record to mark
@@ -5406,10 +5429,14 @@ kmp_int32 __kmpc_start_record_task(ident_t *loc_ref, kmp_int32 gtid,
     __kmp_exec_tdg(gtid, tdg);
     res = 0;
   } else {
-    __kmp_curr_tdg_idx = tdg_id;
-    KMP_DEBUG_ASSERT(__kmp_curr_tdg_idx < __kmp_max_tdgs);
-    __kmp_start_record(gtid, flags, tdg_id);
-    __kmp_num_tdg++;
+    if (__kmp_num_tdg < __kmp_max_tdgs) {
+      __kmp_curr_tdg_id = tdg_id;
+      __kmp_num_tdg++;
+      KMP_DEBUG_ASSERT(__kmp_num_tdg <= __kmp_max_tdgs);
+      __kmp_start_record(gtid, flags, tdg_id);
+    }
+    // if no TDG found, need to execute the task
+    // even not recording
     res = 1;
   }
   KA_TRACE(10, ("__kmpc_start_record_task(exit): T#%d TDG %d starts to %s\n",
diff --git a/openmp/runtime/test/CMakeLists.txt b/openmp/runtime/test/CMakeLists.txt
index 9ee3be6939811..061413fb65ea5 100644
--- a/openmp/runtime/test/CMakeLists.txt
+++ b/openmp/runtime/test/CMakeLists.txt
@@ -30,7 +30,7 @@ update_test_compiler_features()
 pythonize_bool(LIBOMP_USE_HWLOC)
 pythonize_bool(LIBOMP_OMPT_SUPPORT)
 pythonize_bool(LIBOMP_OMPT_OPTIONAL)
-pythonize_bool(LIBOMP_OMPX_TASKGRAPH)
+pythonize_bool(LIBOMP_TASKGRAPH_EXPERIMENTAL)
 pythonize_bool(LIBOMP_HAVE_LIBM)
 pythonize_bool(LIBOMP_HAVE_LIBATOMIC)
 pythonize_bool(OPENMP_STANDALONE_BUILD)
diff --git a/openmp/runtime/test/lit.cfg b/openmp/runtime/test/lit.cfg
index 72da1ba1411f8..483ffc97c8a46 100644
--- a/openmp/runtime/test/lit.cfg
+++ b/openmp/runtime/test/lit.cfg
@@ -123,8 +123,8 @@ if config.has_ompt:
     # for callback.h
     config.test_flags += " -I " + config.test_source_root + "/ompt"
 
-if config.has_ompx_taskgraph:
-    config.available_features.add("ompx_taskgraph")
+if config.has_omp_taskgraph_experimental:
+    config.available_features.add("omp_taskgraph_experimental")
 
 if config.operating_system == 'AIX':
     config.available_features.add("aix")
diff --git a/openmp/runtime/test/lit.site.cfg.in b/openmp/runtime/test/lit.site.cfg.in
index cc8b3b252d7d1..b054651f3e977 100644
--- a/openmp/runtime/test/lit.site.cfg.in
+++ b/openmp/runtime/test/lit.site.cfg.in
@@ -17,7 +17,7 @@ config.target_triple = "@LLVM_TARGET_TRIPLE@"
 config.hwloc_library_dir = "@LIBOMP_HWLOC_LIBRARY_DIR@"
 config.using_hwloc = @LIBOMP_USE_HWLOC@
 config.has_ompt = @LIBOMP_OMPT_SUPPORT@ and @LIBOMP_OMPT_OPTIONAL@
-config.has_ompx_taskgraph = @LIBOMP_OMPX_TASKGRAPH@
+config.has_omp_taskgraph_experimental = @LIBOMP_TASKGRAPH_EXPERIMENTAL@
 config.has_libm = @LIBOMP_HAVE_LIBM@
 config.has_libatomic = @LIBOMP_HAVE_LIBATOMIC@
 config.is_standalone_build = @OPENMP_STANDALONE_BUILD@
diff --git a/openmp/runtime/test/tasking/omp_record_replay.cpp b/openmp/runtime/test/tasking/omp_record_replay.cpp
index 69ad98003a0d6..4fea22e081da9 100644
--- a/openmp/runtime/test/tasking/omp_record_replay.cpp
+++ b/openmp/runtime/test/tasking/omp_record_replay.cpp
@@ -1,4 +1,4 @@
-// REQUIRES: ompx_taskgraph
+// REQUIRES: omp_taskgraph_experimental
 // RUN: %libomp-cxx-compile-and-run
 #include <iostream>
 #include <cassert>
diff --git a/openmp/runtime/test/tasking/omp_record_replay_deps.cpp b/openmp/runtime/test/tasking/omp_record_replay_deps.cpp
index 9b6b370b30efc..4c06ae3f7b273 100644
--- a/openmp/runtime/test/tasking/omp_record_replay_deps.cpp
+++ b/openmp/runtime/test/tasking/omp_record_replay_deps.cpp
@@ -1,4 +1,4 @@
-// REQUIRES: ompx_taskgraph
+// REQUIRES: omp_taskgraph_experimental
 // RUN: %libomp-cxx-compile-and-run
 #include <iostream>
 #include <cassert>
diff --git a/openmp/runtime/test/tasking/omp_record_replay_deps_multi_succ.cpp b/openmp/runtime/test/tasking/omp_record_replay_deps_multi_succ.cpp
index 906fab335f510..6bcd3dee56030 100644
--- a/openmp/runtime/test/tasking/omp_record_replay_deps_multi_succ.cpp
+++ b/openmp/runtime/test/tasking/omp_record_replay_deps_multi_succ.cpp
@@ -1,4 +1,4 @@
-// REQUIRES: ompx_taskgraph
+// REQUIRES: omp_taskgraph_experimental
 // RUN: %libomp-cxx-compile-and-run
 #include <omp.h>
 #include <cassert>
diff --git a/openmp/runtime/test/tasking/omp_record_replay_multiTDGs.cpp b/openmp/runtime/test/tasking/omp_record_replay_multiTDGs.cpp
index 03252843689c4..1864d5d89cc70 100644
--- a/openmp/runtime/test/tasking/omp_record_replay_multiTDGs.cpp
+++ b/openmp/runtime/test/tasking/omp_record_replay_multiTDGs.cpp
@@ -1,4 +1,4 @@
-// REQUIRES: ompx_taskgraph
+// REQUIRES: omp_taskgraph_experimental
 // RUN: %libomp-cxx-compile-and-run
 #include <iostream>
 #include <cassert>
diff --git a/openmp/runtime/test/tasking/omp_record_replay_print_dot.cpp b/openmp/runtime/test/tasking/omp_record_replay_print_dot.cpp
index 2fe55f0815429..7f1f5ccd77d37 100644
--- a/openmp/runtime/test/tasking/omp_record_replay_print_dot.cpp
+++ b/openmp/runtime/test/tasking/omp_record_replay_print_dot.cpp
@@ -1,4 +1,4 @@
-// REQUIRES: ompx_taskgraph
+// REQUIRES: omp_taskgraph_experimental
 // RUN: %libomp-cxx-compile-and-run
 #include <iostream>
 #include <fstream>
diff --git a/openmp/runtime/test/tasking/omp_record_replay_taskloop.cpp b/openmp/runtime/test/tasking/omp_record_replay_taskloop.cpp
index 3d88faeeb28ee..163a1b4192d85 100644
--- a/openmp/runtime/test/tasking/omp_record_replay_taskloop.cpp
+++ b/openmp/runtime/test/tasking/omp_record_replay_taskloop.cpp
@@ -1,4 +1,4 @@
-// REQUIRES: ompx_taskgraph
+// REQUIRES: omp_taskgraph_experimental
 // RUN: %libomp-cxx-compile-and-run
 #include <iostream>
 #include <cassert>