[OpenMP][OMPD]Code movement required for OMPD

These changes don't come under OMPD guard as it is a movement of existing code to capture parallel behavior correctly.
"Runtime Entry Points for OMPD" like "ompd_bp_parallel_begin" and "ompd_bp_parallel_begin" should be placed at the correct execution point for the debugging tool to access proper handles/data.
Without the below changes, in certain cases, debugging tool will pick the wrong parallel and task handle.

Reviewed By: @hbae
Differential Revision: https://reviews.llvm.org/D100366

diff --git a/openmp/runtime/src/kmp_csupport.cpp b/openmp/runtime/src/kmp_csupport.cpp
index 2a7c9a8..0b5f899 100644 (file)
--- a/openmp/runtime/src/kmp_csupport.cpp
+++ b/openmp/runtime/src/kmp_csupport.cpp
@@ -578,9 +578,6 @@ void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32 global_tid) {
     __kmp_free(top);
   }
 
-  // if( serial_team -> t.t_serialized > 1 )
-  serial_team->t.t_level--;
-
   /* pop dispatch buffers stack */
   KMP_DEBUG_ASSERT(serial_team->t.t_dispatch->th_disp_buffer);
   {
@@ -605,6 +602,7 @@ void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32 global_tid) {
     }
 #endif /* KMP_ARCH_X86 || KMP_ARCH_X86_64 */
 
+    __kmp_pop_current_task_from_thread(this_thr);
 #if OMPD_SUPPORT
     if (ompd_state & OMPD_ENABLE_BP)
       ompd_bp_parallel_end();
@@ -623,8 +621,6 @@ void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32 global_tid) {
     this_thr->th.th_dispatch =
         &this_thr->th.th_team->t.t_dispatch[serial_team->t.t_master_tid];
 
-    __kmp_pop_current_task_from_thread(this_thr);
-
     KMP_ASSERT(this_thr->th.th_current_task->td_flags.executing == 0);
     this_thr->th.th_current_task->td_flags.executing = 1;
 
@@ -645,6 +641,7 @@ void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32 global_tid) {
     }
   }
 
+  serial_team->t.t_level--;
   if (__kmp_env_consistency_check)
     __kmp_pop_parallel(global_tid, NULL);
 #if OMPT_SUPPORT

diff --git a/openmp/runtime/src/ompt-specific.cpp b/openmp/runtime/src/ompt-specific.cpp
index 1ad0e17..c90edeb 100644 (file)
--- a/openmp/runtime/src/ompt-specific.cpp
+++ b/openmp/runtime/src/ompt-specific.cpp
@@ -283,10 +283,6 @@ void __ompt_lw_taskteam_link(ompt_lw_taskteam_t *lwt, kmp_info_t *thr,
     link_lwt->ompt_team_info = *OMPT_CUR_TEAM_INFO(thr);
     *OMPT_CUR_TEAM_INFO(thr) = tmp_team;
 
-    ompt_task_info_t tmp_task = lwt->ompt_task_info;
-    link_lwt->ompt_task_info = *OMPT_CUR_TASK_INFO(thr);
-    *OMPT_CUR_TASK_INFO(thr) = tmp_task;
-
     // link the taskteam into the list of taskteams:
     ompt_lw_taskteam_t *my_parent =
         thr->th.th_team->t.ompt_serialized_team_info;
@@ -297,6 +293,10 @@ void __ompt_lw_taskteam_link(ompt_lw_taskteam_t *lwt, kmp_info_t *thr,
       ompd_bp_parallel_begin();
     }
 #endif
+
+    ompt_task_info_t tmp_task = lwt->ompt_task_info;
+    link_lwt->ompt_task_info = *OMPT_CUR_TASK_INFO(thr);
+    *OMPT_CUR_TASK_INFO(thr) = tmp_task;
   } else {
     // this is the first serialized team, so we just store the values in the
     // team and drop the taskteam-object
@@ -313,6 +313,9 @@ void __ompt_lw_taskteam_link(ompt_lw_taskteam_t *lwt, kmp_info_t *thr,
 void __ompt_lw_taskteam_unlink(kmp_info_t *thr) {
   ompt_lw_taskteam_t *lwtask = thr->th.th_team->t.ompt_serialized_team_info;
   if (lwtask) {
+    ompt_task_info_t tmp_task = lwtask->ompt_task_info;
+    lwtask->ompt_task_info = *OMPT_CUR_TASK_INFO(thr);
+    *OMPT_CUR_TASK_INFO(thr) = tmp_task;
 #if OMPD_SUPPORT
     if (ompd_state & OMPD_ENABLE_BP) {
       ompd_bp_parallel_end();
@@ -324,10 +327,6 @@ void __ompt_lw_taskteam_unlink(kmp_info_t *thr) {
     lwtask->ompt_team_info = *OMPT_CUR_TEAM_INFO(thr);
     *OMPT_CUR_TEAM_INFO(thr) = tmp_team;
 
-    ompt_task_info_t tmp_task = lwtask->ompt_task_info;
-    lwtask->ompt_task_info = *OMPT_CUR_TASK_INFO(thr);
-    *OMPT_CUR_TASK_INFO(thr) = tmp_task;
-
     if (lwtask->heap) {
       __kmp_free(lwtask);
       lwtask = NULL;