Fixes and improvements to tasking in barriers

[jlpeyton]

Summary

1. When the number of threads in a team increases, new threads need to have all their barrier struct fields initialized. We were missing the parent_bar and team fields.
2. For non-forkjoin barriers, we now do the __kmp_task_team_setup before the gather. The setup now sets up the task_team that all the threads will switch to after the barrier, but it needs to be done before other threads do the switch.
3. Remove an unneeded assignment of tt_found_tasks in task team free function.

[tlwilmar]

LGTM

openmp/trunk/runtime/src/kmp_barrier.cpp

if (uninitialized || team_sz_changed || tid_changed) {
                     break;
                 }
                 ++d;
             }
         }
         thr_bar->offset = 7-(tid-thr_bar->parent_tid-1);
         thr_bar->old_tid = tid;
         thr_bar->wait_flag = KMP_BARRIER_NOT_WAITING;
+        thr_bar->team = team;
+        thr_bar->parent_bar = &team->t.t_threads[thr_bar->parent_tid]->th.th_bar[bt].bb;
     }
     if (uninitialized || team_changed || tid_changed) {
         thr_bar->team = team;
         thr_bar->parent_bar = &team->t.t_threads[thr_bar->parent_tid]->th.th_bar[bt].bb;
         retval = true;
     }
     if (uninitialized || team_sz_changed || tid_changed) {
         thr_bar->nproc = nproc;
#if USE_DEBUGGER
         } else {
             this_thr->th.th_bar[bt].bb.b_worker_arrived += 1;
         } // if
 #endif /* USE_DEBUGGER */
         if (reduce != NULL) {
             //KMP_DEBUG_ASSERT( is_split == TRUE );  // #C69956
             this_thr->th.th_local.reduce_data = reduce_data;
         }
+
+        if (KMP_MASTER_TID(tid) && __kmp_tasking_mode != tskm_immediate_exec)
+            __kmp_task_team_setup(this_thr, team, 0); // use 0 to only setup the current team if nthreads > 1
+
         switch (__kmp_barrier_gather_pattern[bt]) {
         case bp_hyper_bar: {
             KMP_ASSERT(__kmp_barrier_gather_branch_bits[bt]); // don't set branch bits to 0; use linear
             __kmp_hyper_barrier_gather(bt, this_thr, gtid, tid, reduce
                                        USE_ITT_BUILD_ARG(itt_sync_obj) );
             break;
         }
         case bp_hierarchical_bar: {
#endif /* USE_DEBUGGER */
 
         KMP_MB();
 
         if (KMP_MASTER_TID(tid)) {
             status = 0;
             if (__kmp_tasking_mode != tskm_immediate_exec) {
                 __kmp_task_team_wait(this_thr, team
                                      USE_ITT_BUILD_ARG(itt_sync_obj) );
-                __kmp_task_team_setup(this_thr, team, 0); // use 0 to only setup the current team if nthreads > 1
             }
 #if USE_DEBUGGER
             // Let the debugger know: All threads are arrived and starting leaving the barrier.
             team->t.t_bar[bt].b_team_arrived += 1;
 #endif
 
 #if USE_ITT_BUILD
             /* TODO: In case of split reduction barrier, master thread may send acquired event early,

openmp/trunk/runtime/src/kmp_tasking.c

__kmp_free_task_team( kmp_info_t *thread, kmp_task_team_t *task_team )
     KA_TRACE( 20, ( "__kmp_free_task_team: T#%d task_team = %p\n",
                     thread ? __kmp_gtid_from_thread( thread ) : -1, task_team ) );
 
     // Put task team back on free list
     __kmp_acquire_bootstrap_lock( & __kmp_task_team_lock );
 
     KMP_DEBUG_ASSERT( task_team -> tt.tt_next == NULL );
     task_team -> tt.tt_next = __kmp_free_task_teams;
-    TCW_4(task_team -> tt.tt_found_tasks, FALSE);
     TCW_PTR(__kmp_free_task_teams, task_team);
 
     __kmp_release_bootstrap_lock( & __kmp_task_team_lock );
 }
 
 
 //------------------------------------------------------------------------------
 // __kmp_reap_task_teams: