[OMPT] Add support for ompt_event_task_dependences and ompt_event_task_dependence_pair

The attached patch adds support for ompt_event_task_dependences and
ompt_event_task_dependence_pair events from the OMPT specification [1]. These
events only apply to OpenMP 4.0 and 4.1 (aka 4.5) because task dependencies
were introduced in 4.0.

With respect to the changes:

ompt_event_task_dependences
According to the specification, this event is raised after the task has been
created, thefore this event needs to be raised after ompt_event_task_begin
(in __kmp_task_start). However, the dependencies are known at
__kmpc_omp_task_with_deps which occurs before __kmp_task_start. My modifications
extend the ompt_task_info_t struct in order to store the dependencies of the
task when _kmpc_omp_task_with_deps occurs and then they are emitted in
__kmp_task_start just after raising the ompt_event_task_begin. The deps field
is allocated and valid until the event is raised and it is freed and set
to null afterwards.

ompt_event_task_dependence_pair
The processing of the dependences (i.e. checking whenever a dependence is
already satisfied) is done within __kmp_process_deps. That function checks
every dependence and calls the __kmp_track_dependence routine which gives some
support for graphical output. I used that routine to emit the dependence pair
but I also needed to know the sink_task. Despite the fact that the code within
KMP_SUPPORT_GRAPH_OUTPUT refers to task_sink it may be null because
sink->dn.task (there's a comment regarding this) and in fact it does not point
to a proper pointer value because the value is set in node->dn.task = task;
after the __kmp_process_deps calls in __kmp_check_deps. I have extended the
__kmp_process_deps and __kmp_track_dependence parameter list to receive the
sink_task.

[1] https://github.com/OpenMPToolsInterface/OMPT-Technical-Report/blob/target/ompt-tr.pdf

Patch by Harald Servat
Differential Revision: http://reviews.llvm.org/D14746

llvm-svn: 259038

6 files changed, 170 insertions, 11 deletions

diff --git a/openmp/runtime/src/include/40/ompt.h.var b/openmp/runtime/src/include/40/ompt.h.var
index a1387fab43b..ff1d86be859 100644
--- a/openmp/runtime/src/include/40/ompt.h.var
+++ b/openmp/runtime/src/include/40/ompt.h.var
@@ -171,7 +171,10 @@
     macro (ompt_event_destroy_lock,             ompt_wait_callback_t,          59) /* lock destruction */       \
     macro (ompt_event_destroy_nest_lock,        ompt_wait_callback_t,          60) /* nest lock destruction */  \
                                                                                                                 \
-    macro (ompt_event_flush,                    ompt_callback_t,               61) /* after executing flush */
+    macro (ompt_event_flush,                    ompt_callback_t,               61) /* after executing flush */  \
+                                                                                                                \
+    macro (ompt_event_task_dependences,         ompt_task_dependences_callback_t, 69) /* report task dependences  */\
+    macro (ompt_event_task_dependence_pair,     ompt_task_pair_callback_t,     70) /* report task dependence pair */
 
 
 
@@ -206,6 +209,23 @@ typedef struct ompt_frame_s {
 } ompt_frame_t;
 
 
+/*---------------------
+ * dependences types
+ *---------------------*/
+
+typedef enum ompt_task_dependence_flag_e {
+    // a two bit field for the dependence type
+    ompt_task_dependence_type_out   = 1,
+    ompt_task_dependence_type_in    = 2,
+    ompt_task_dependence_type_inout = 3,
+} ompt_task_dependence_flag_t;
+
+typedef struct ompt_task_dependence_s {
+    void *variable_addr;
+    uint32_t  dependence_flags;
+} ompt_task_dependence_t;
+
+
 /*****************************************************************************
  * enumerations for thread states and runtime events
  *****************************************************************************/
@@ -325,6 +345,13 @@ typedef void (*ompt_new_task_callback_t) (
     void *task_function               /* pointer to outlined function */
 );
 
+/* task dependences */
+typedef void (*ompt_task_dependences_callback_t) (
+    ompt_task_id_t task_id,            /* ID of task with dependences */
+    const ompt_task_dependence_t *deps,/* vector of task dependences  */
+    int ndeps                          /* number of dependences       */
+);
+
 /* program */
 typedef void (*ompt_control_callback_t) (
     uint64_t command,                 /* command of control call      */
diff --git a/openmp/runtime/src/include/41/ompt.h.var b/openmp/runtime/src/include/41/ompt.h.var
index da7647fc675..ecf313ab890 100644
--- a/openmp/runtime/src/include/41/ompt.h.var
+++ b/openmp/runtime/src/include/41/ompt.h.var
@@ -171,7 +171,10 @@
     macro (ompt_event_destroy_lock,             ompt_wait_callback_t,          59) /* lock destruction */       \
     macro (ompt_event_destroy_nest_lock,        ompt_wait_callback_t,          60) /* nest lock destruction */  \
                                                                                                                 \
-    macro (ompt_event_flush,                    ompt_callback_t,               61) /* after executing flush */
+    macro (ompt_event_flush,                    ompt_callback_t,               61) /* after executing flush */  \
+                                                                                                                \
+    macro (ompt_event_task_dependences,         ompt_task_dependences_callback_t, 69) /* report task dependences  */\
+    macro (ompt_event_task_dependence_pair,     ompt_task_pair_callback_t,     70) /* report task dependence pair */
 
 
 
@@ -206,6 +209,23 @@ typedef struct ompt_frame_s {
 } ompt_frame_t;
 
 
+/*---------------------
+ * dependences types
+ *---------------------*/
+
+typedef enum ompt_task_dependence_flag_e {
+    // a two bit field for the dependence type
+    ompt_task_dependence_type_out   = 1,
+    ompt_task_dependence_type_in    = 2,
+    ompt_task_dependence_type_inout = 3,
+} ompt_task_dependence_flag_t;
+
+typedef struct ompt_task_dependence_s {
+    void *variable_addr;
+    uint32_t  dependence_flags;
+} ompt_task_dependence_t;
+
+
 /*****************************************************************************
  * enumerations for thread states and runtime events
  *****************************************************************************/
@@ -325,6 +345,13 @@ typedef void (*ompt_new_task_callback_t) (
     void *task_function               /* pointer to outlined function */
 );
 
+/* task dependences */
+typedef void (*ompt_task_dependences_callback_t) (
+    ompt_task_id_t task_id,            /* ID of task with dependences */
+    const ompt_task_dependence_t *deps,/* vector of task dependences  */
+    int ndeps                          /* number of dependences       */
+);
+
 /* program */
 typedef void (*ompt_control_callback_t) (
     uint64_t command,                 /* command of control call      */
diff --git a/openmp/runtime/src/kmp_taskdeps.cpp b/openmp/runtime/src/kmp_taskdeps.cpp
index da085ce50cc..c4be40a0f40 100644
--- a/openmp/runtime/src/kmp_taskdeps.cpp
+++ b/openmp/runtime/src/kmp_taskdeps.cpp
@@ -196,7 +196,8 @@ __kmp_depnode_list_free ( kmp_info_t *thread, kmp_depnode_list *list )
 }
 
 static inline void
-__kmp_track_dependence ( kmp_depnode_t *source, kmp_depnode_t *sink )
+__kmp_track_dependence ( kmp_depnode_t *source, kmp_depnode_t *sink,
+                         kmp_task_t *sink_task )
 {
 #ifdef KMP_SUPPORT_GRAPH_OUTPUT
     kmp_taskdata_t * task_source = KMP_TASK_TO_TASKDATA(source->dn.task);
@@ -204,12 +205,27 @@ __kmp_track_dependence ( kmp_depnode_t *source, kmp_depnode_t *sink )
 
     __kmp_printf("%d(%s) -> %d(%s)\n", source->dn.id, task_source->td_ident->psource, sink->dn.id, task_sink->td_ident->psource);
 #endif
+#if OMPT_SUPPORT && OMPT_TRACE
+    /* OMPT tracks dependences between task (a=source, b=sink) in which
+       task a blocks the execution of b through the ompt_new_dependence_callback */
+    if (ompt_enabled &&
+        ompt_callbacks.ompt_callback(ompt_event_task_dependence_pair))
+    {
+        kmp_taskdata_t * task_source = KMP_TASK_TO_TASKDATA(source->dn.task);
+        kmp_taskdata_t * task_sink = KMP_TASK_TO_TASKDATA(sink_task);
+
+        ompt_callbacks.ompt_callback(ompt_event_task_dependence_pair)(
+          task_source->ompt_task_info.task_id,
+          task_sink->ompt_task_info.task_id);
+    }
+#endif /* OMPT_SUPPORT && OMPT_TRACE */
 }
 
 template< bool filter >
 static inline kmp_int32
 __kmp_process_deps ( kmp_int32 gtid, kmp_depnode_t *node, kmp_dephash_t *hash,
-                     bool dep_barrier,kmp_int32 ndeps, kmp_depend_info_t *dep_list)
+                     bool dep_barrier,kmp_int32 ndeps, kmp_depend_info_t *dep_list,
+                     kmp_task_t *task )
 {
     KA_TRACE(30, ("__kmp_process_deps<%d>: T#%d processing %d depencies : dep_barrier = %d\n", filter, gtid, ndeps, dep_barrier ) );
     
@@ -231,7 +247,7 @@ __kmp_process_deps ( kmp_int32 gtid, kmp_depnode_t *node, kmp_dephash_t *hash,
                 if ( indep->dn.task ) {
                     KMP_ACQUIRE_DEPNODE(gtid,indep);
                     if ( indep->dn.task ) {
-                        __kmp_track_dependence(indep,node);
+                        __kmp_track_dependence(indep,node,task);
                         indep->dn.successors = __kmp_add_node(thread, indep->dn.successors, node);
                         KA_TRACE(40,("__kmp_process_deps<%d>: T#%d adding dependence from %p to %p\n",
                                  filter,gtid, KMP_TASK_TO_TASKDATA(indep->dn.task), KMP_TASK_TO_TASKDATA(node->dn.task)));
@@ -247,7 +263,7 @@ __kmp_process_deps ( kmp_int32 gtid, kmp_depnode_t *node, kmp_dephash_t *hash,
         } else if ( last_out && last_out->dn.task ) {
             KMP_ACQUIRE_DEPNODE(gtid,last_out);
             if ( last_out->dn.task ) {
-                __kmp_track_dependence(last_out,node);
+                __kmp_track_dependence(last_out,node,task);
                 last_out->dn.successors = __kmp_add_node(thread, last_out->dn.successors, node);
                 KA_TRACE(40,("__kmp_process_deps<%d>: T#%d adding dependence from %p to %p\n", 
                              filter,gtid, KMP_TASK_TO_TASKDATA(last_out->dn.task), KMP_TASK_TO_TASKDATA(node->dn.task)));
@@ -312,8 +328,10 @@ __kmp_check_deps ( kmp_int32 gtid, kmp_depnode_t *node, kmp_task_t *task, kmp_de
     // used to pack all npredecessors additions into a single atomic operation at the end
     int npredecessors;
 
-    npredecessors = __kmp_process_deps<true>(gtid, node, hash, dep_barrier, ndeps, dep_list);
-    npredecessors += __kmp_process_deps<false>(gtid, node, hash, dep_barrier, ndeps_noalias, noalias_dep_list);
+    npredecessors = __kmp_process_deps<true>(gtid, node, hash, dep_barrier,
+      ndeps, dep_list, task);
+    npredecessors += __kmp_process_deps<false>(gtid, node, hash, dep_barrier,
+      ndeps_noalias, noalias_dep_list, task);
 
     node->dn.task = task;
     KMP_MB();
@@ -404,6 +422,51 @@ __kmpc_omp_task_with_deps( ident_t *loc_ref, kmp_int32 gtid, kmp_task_t * new_ta
     kmp_info_t *thread = __kmp_threads[ gtid ];
     kmp_taskdata_t * current_task = thread->th.th_current_task;
 
+#if OMPT_SUPPORT && OMPT_TRACE
+    /* OMPT grab all dependences if requested by the tool */
+    if (ompt_enabled && ndeps+ndeps_noalias > 0 &&
+        ompt_callbacks.ompt_callback(ompt_event_task_dependences))
+	{
+        kmp_int32 i;
+
+        new_taskdata->ompt_task_info.ndeps = ndeps+ndeps_noalias;
+        new_taskdata->ompt_task_info.deps = (ompt_task_dependence_t *)
+          KMP_OMPT_DEPS_ALLOC(thread,
+             (ndeps+ndeps_noalias)*sizeof(ompt_task_dependence_t));
+
+        KMP_ASSERT(new_taskdata->ompt_task_info.deps != NULL);
+
+        for (i = 0; i < ndeps; i++)
+        {
+            new_taskdata->ompt_task_info.deps[i].variable_addr =
+              (void*) dep_list[i].base_addr;
+            if (dep_list[i].flags.in && dep_list[i].flags.out)
+                new_taskdata->ompt_task_info.deps[i].dependence_flags =
+                  ompt_task_dependence_type_inout;
+            else if (dep_list[i].flags.out)
+                new_taskdata->ompt_task_info.deps[i].dependence_flags =
+                  ompt_task_dependence_type_out;
+            else if (dep_list[i].flags.in)
+                new_taskdata->ompt_task_info.deps[i].dependence_flags =
+                  ompt_task_dependence_type_in;
+        }
+        for (i = 0; i < ndeps_noalias; i++)
+        {
+            new_taskdata->ompt_task_info.deps[ndeps+i].variable_addr =
+              (void*) noalias_dep_list[i].base_addr;
+            if (noalias_dep_list[i].flags.in && noalias_dep_list[i].flags.out)
+                new_taskdata->ompt_task_info.deps[ndeps+i].dependence_flags =
+                  ompt_task_dependence_type_inout;
+            else if (noalias_dep_list[i].flags.out)
+                new_taskdata->ompt_task_info.deps[ndeps+i].dependence_flags =
+                  ompt_task_dependence_type_out;
+            else if (noalias_dep_list[i].flags.in)
+                new_taskdata->ompt_task_info.deps[ndeps+i].dependence_flags =
+                  ompt_task_dependence_type_in;
+        }
+    }
+#endif /* OMPT_SUPPORT && OMPT_TRACE */
+
     bool serial = current_task->td_flags.team_serial || current_task->td_flags.tasking_ser || current_task->td_flags.final;
 #if OMP_41_ENABLED
     serial = serial && !(new_taskdata->td_flags.proxy == TASK_PROXY);
diff --git a/openmp/runtime/src/kmp_tasking.c b/openmp/runtime/src/kmp_tasking.c
index fc9348c94e2..3df0c9e73dd 100644
--- a/openmp/runtime/src/kmp_tasking.c
+++ b/openmp/runtime/src/kmp_tasking.c
@@ -462,6 +462,22 @@ __kmp_task_start( kmp_int32 gtid, kmp_task_t * task, kmp_taskdata_t * current_ta
             taskdata->ompt_task_info.function);
     }
 #endif
+#if OMP_40_ENABLED && OMPT_SUPPORT && OMPT_TRACE
+    /* OMPT emit all dependences if requested by the tool */
+    if (ompt_enabled && taskdata->ompt_task_info.ndeps > 0 &&
+        ompt_callbacks.ompt_callback(ompt_event_task_dependences))
+	{
+        ompt_callbacks.ompt_callback(ompt_event_task_dependences)(
+            taskdata->ompt_task_info.task_id,
+            taskdata->ompt_task_info.deps,
+            taskdata->ompt_task_info.ndeps
+        );
+		/* We can now free the allocated memory for the dependencies */
+		KMP_OMPT_DEPS_FREE (thread, taskdata->ompt_task_info.deps);
+        taskdata->ompt_task_info.deps = NULL;
+        taskdata->ompt_task_info.ndeps = 0;
+    }
+#endif /* OMP_40_ENABLED && OMPT_SUPPORT && OMPT_TRACE */
 
     return;
 }
@@ -760,6 +776,10 @@ __kmp_task_init_ompt( kmp_taskdata_t * task, int tid, void * function )
         task->ompt_task_info.function = function;
         task->ompt_task_info.frame.exit_runtime_frame = NULL;
         task->ompt_task_info.frame.reenter_runtime_frame = NULL;
+#if OMP_40_ENABLED
+        task->ompt_task_info.ndeps = 0;
+        task->ompt_task_info.deps = NULL;
+#endif /* OMP_40_ENABLED */
     }
 }
 #endif
diff --git a/openmp/runtime/src/ompt-event-specific.h b/openmp/runtime/src/ompt-event-specific.h
index 28c1512ac26..fdf121371a0 100644
--- a/openmp/runtime/src/ompt-event-specific.h
+++ b/openmp/runtime/src/ompt-event-specific.h
@@ -141,4 +141,12 @@
 
 #define ompt_event_flush_implemented                    ompt_event_UNIMPLEMENTED
 
+#if OMP_40_ENABLED
+# define ompt_event_task_dependences_implemented         ompt_event_MAY_ALWAYS_TRACE
+# define ompt_event_task_dependence_pair_implemented     ompt_event_MAY_ALWAYS_TRACE
+#else
+# define ompt_event_task_dependences_implemented         ompt_event_UNIMPLEMENTED
+# define ompt_event_task_dependence_pair_implemented     ompt_event_UNIMPLEMENTED
+#endif /* OMP_40_ENABLED */
+
 #endif
diff --git a/openmp/runtime/src/ompt-internal.h b/openmp/runtime/src/ompt-internal.h
index 64e8d2e8fd6..42da9d86793 100644
--- a/openmp/runtime/src/ompt-internal.h
+++ b/openmp/runtime/src/ompt-internal.h
@@ -26,9 +26,13 @@ typedef struct ompt_callbacks_s {
 
 
 typedef struct {
-    ompt_frame_t        frame;
-    void*               function;
-    ompt_task_id_t      task_id;
+    ompt_frame_t            frame;
+    void*                   function;
+    ompt_task_id_t          task_id;
+#if OMP_40_ENABLED
+    int                     ndeps;
+    ompt_task_dependence_t  *deps;
+#endif /* OMP_40_ENABLED */
 } ompt_task_info_t;
 
 
@@ -62,6 +66,16 @@ typedef struct {
 
 extern ompt_callbacks_t ompt_callbacks;
 
+#if OMP_40_ENABLED && OMPT_SUPPORT && OMPT_TRACE
+#if USE_FAST_MEMORY
+#  define KMP_OMPT_DEPS_ALLOC __kmp_fast_allocate
+#  define KMP_OMPT_DEPS_FREE __kmp_fast_free
+# else
+#  define KMP_OMPT_DEPS_ALLOC __kmp_thread_malloc
+#  define KMP_OMPT_DEPS_FREE __kmp_thread_free
+# endif
+#endif /* OMP_40_ENABLED && OMPT_SUPPORT && OMPT_TRACE */
+
 #ifdef __cplusplus
 extern "C" {
 #endif