diff mbox

[gomp4.1] document more things

Message ID 55DF184C.5070209@redhat.com
State New
Headers show

Commit Message

Aldy Hernandez Aug. 27, 2015, 2:01 p.m. UTC 
While I'm here... I've added more comments throughout.

I'm still not convinced we need anything but GOMP_TASK_WAITING in enum 
gomp_task_kind, but perhaps you needed it for future usage?  If so, 
could you provide a description for GOMP_TASK_IFFALSE?

OK for branch?
commit 45fdbc84123fefde6bf99aaf87099317aaa38f24
Author: Aldy Hernandez <aldyh@redhat.com>
Date:   Wed Aug 26 07:28:39 2015 -0700

    	* libgomp.h (enum gomp_task_kind): Comment fields.
    	(struct gomp_task): Comment some fields.
    	* task.c (GOMP_task): Add comments.
    	(gomp_task_run_pre): Same.

Comments

Jakub Jelinek Aug. 27, 2015, 2:14 p.m. UTC | #1 
On Thu, Aug 27, 2015 at 07:01:48AM -0700, Aldy Hernandez wrote:
> commit 45fdbc84123fefde6bf99aaf87099317aaa38f24
> Author: Aldy Hernandez <aldyh@redhat.com>
> Date:   Wed Aug 26 07:28:39 2015 -0700
> 
>     	* libgomp.h (enum gomp_task_kind): Comment fields.
>     	(struct gomp_task): Comment some fields.
>     	* task.c (GOMP_task): Add comments.
>     	(gomp_task_run_pre): Same.
> 
> diff --git a/libgomp/libgomp.h b/libgomp/libgomp.h
> index 3d705ef..34f064c 100644
> --- a/libgomp/libgomp.h
> +++ b/libgomp/libgomp.h
> @@ -263,9 +263,13 @@ extern char *goacc_device_type;
>  
>  enum gomp_task_kind
>  {
> +  /* Thread not directly created by user.  For instance, when creating
> +     a team of threads or an ICV.  */

That is misleading, especially with the (possibly) upcoming thread task.
"implicit task" is an OpenMP tasking term:
"A task generated by an implicit parallel region or generated when a parallel
construct is encountered during execution."
but there is really no reason to explain what it is, so you should just say
in the comment
  /* Implicit task.  */

>    GOMP_TASK_IMPLICIT,
>    GOMP_TASK_IFFALSE,

This is for if (0) tasks.  The right OpenMP tasking term is "undeferred
task", perhaps we should just replace GOMP_TASK_IFFALSE with
GOMP_TASK_UNDEFERRED everywhere.

> +  /* Task created by GOMP_task and waiting to be run.  */
>    GOMP_TASK_WAITING,
> +  /* Task currently scheduled and about to execute.  */

Well, not just about to execute, that kind is used even for the task that is
already executing.

Otherwise LGTM.

	Jakub
diff mbox

Patch

diff --git a/libgomp/libgomp.h b/libgomp/libgomp.h
index 3d705ef..34f064c 100644
--- a/libgomp/libgomp.h
+++ b/libgomp/libgomp.h
@@ -263,9 +263,13 @@  extern char *goacc_device_type;
 
 enum gomp_task_kind
 {
+  /* Thread not directly created by user.  For instance, when creating
+     a team of threads or an ICV.  */
   GOMP_TASK_IMPLICIT,
   GOMP_TASK_IFFALSE,
+  /* Task created by GOMP_task and waiting to be run.  */
   GOMP_TASK_WAITING,
+  /* Task currently scheduled and about to execute.  */
   GOMP_TASK_TIED
 };
 
@@ -313,8 +317,11 @@  struct gomp_task
   struct gomp_task *prev_child;
   struct gomp_task *next_queue;
   struct gomp_task *prev_queue;
+  /* Next task in the current taskgroup.  */
   struct gomp_task *next_taskgroup;
+  /* Previous task in the current taskgroup.  */
   struct gomp_task *prev_taskgroup;
+  /* Taskgroup this task belongs in.  */
   struct gomp_taskgroup *taskgroup;
   struct gomp_dependers_vec *dependers;
   struct htab *depend_hash;
diff --git a/libgomp/task.c b/libgomp/task.c
index 7c7bae4..c0197cd 100644
--- a/libgomp/task.c
+++ b/libgomp/task.c
@@ -388,6 +388,7 @@  GOMP_task (void (*fn) (void *), void *data, void (*cpyfn) (void *, void *),
       parent->children = task;
       if (taskgroup)
 	{
+	  /* If applicable, place task into its taskgroup.  */
 	  if (taskgroup->children)
 	    {
 	      task->next_taskgroup = taskgroup->children;
@@ -451,7 +452,14 @@  gomp_task_run_pre (struct gomp_task *child_task, struct gomp_task *parent,
 {
   if (parent)
     {
-      /* Remove child_task from parent.  */
+      /* Adjust children such that it will point to a next child,
+	 while the current one is scheduled to be executed.  This way,
+	 GOMP_taskwait (and others) can schedule a next task while
+	 waiting.
+
+	 Do not remove it entirely from the circular list, as it is
+	 still a child, though not one we should consider first (say
+	 by GOMP_taskwait).  */
       if (parent->children == child_task)
 	parent->children = child_task->next_child;
 
@@ -465,10 +473,14 @@  gomp_task_run_pre (struct gomp_task *child_task, struct gomp_task *parent,
 	    parent->taskwait->last_parent_depends_on = NULL;
 	}
     }
-  /* Remove child_task from taskgroup.  */
+
+  /* Adjust taskgroup to point to the next taskgroup.  See note above
+     regarding adjustment of children as to why the child_task is not
+     removed entirely from the circular list.  */
   if (taskgroup && taskgroup->children == child_task)
     taskgroup->children = child_task->next_taskgroup;
 
+  /* Remove child_task from the task_queue.  */
   child_task->prev_queue->next_queue = child_task->next_queue;
   child_task->next_queue->prev_queue = child_task->prev_queue;
   if (team->task_queue == child_task)
@@ -479,6 +491,7 @@  gomp_task_run_pre (struct gomp_task *child_task, struct gomp_task *parent,
 	team->task_queue = NULL;
     }
   child_task->kind = GOMP_TASK_TIED;
+
   if (--team->task_queued_count == 0)
     gomp_team_barrier_clear_task_pending (&team->barrier);
   if ((gomp_team_barrier_cancelled (&team->barrier)