[PATCH 10/12] scheduler: Simplify simple scheduler

[Gedare Bloom]

On Thu, Jun 13, 2013 at 2:54 PM, Sebastian Huber
<sebastian.huber at embedded-brains.de> wrote:
> On 13/06/13 20:45, Gedare Bloom wrote:
>>
>> On Thu, Jun 13, 2013 at 2:35 PM, Sebastian Huber
>> <sebastian.huber at embedded-brains.de>  wrote:
>>>
>>> >On 13/06/13 20:21, Sebastian Huber wrote:
>>>>
>>>> >>
>>>> >>On 13/06/13 18:06, Gedare Bloom wrote:
>>>>>
>>>>> >>>
>>>>> >>>On Thu, Jun 13, 2013 at 12:01 PM, Gedare Bloom<gedare at rtems.org>
>>>>> >>> wrote:
>>>>>>
>>>>>> >>>>
>>>>>> >>>>On Wed, Jun 12, 2013 at 11:29 AM, Sebastian Huber
>>>>>> >>>><sebastian.huber at embedded-brains.de>  wrote:
>>>>>>>
>>>>>>> >>>>>
>>>>>>> >>>>>Add and use _Scheduler_simple_Insert_as_first_order(),
>>>>>>> >>>>>_Scheduler_simple_Insert_as_last_order(),
>>>>>>> >>>>>_Scheduler_simple_Insert_as_first() and
>>>>>>> >>>>>_Scheduler_simple_Insert_as_last().
>>>>>>> >>>>>---
>>>>>>> >>>>>   .../score/inline/rtems/score/schedulersimple.inl   |   46
>>>>>>> >>>>>++++++++++++++++++++
>>>>>>> >>>>>   .../score/src/schedulersimplereadyqueueenqueue.c   |   21
>>>>>>> >>>>> +--------
>>>>>>> >>>>>   .../src/schedulersimplereadyqueueenqueuefirst.c    |   26
>>>>>>> >>>>> +----------
>>>>>>> >>>>>   3 files changed, 50 insertions(+), 43 deletions(-)
>>>>>>> >>>>>
>>>>>>> >>>>>diff --git a/cpukit/score/inline/rtems/score/schedulersimple.inl
>>>>>>> >>>>>b/cpukit/score/inline/rtems/score/schedulersimple.inl
>>>>>>> >>>>>index e67fc3c..1b58c85 100644
>>>>>>> >>>>>--- a/cpukit/score/inline/rtems/score/schedulersimple.inl
>>>>>>> >>>>>+++ b/cpukit/score/inline/rtems/score/schedulersimple.inl
>>>>>>> >>>>>@@ -48,6 +48,52 @@ RTEMS_INLINE_ROUTINE void
>>>>>>> >>>>>_Scheduler_simple_Ready_queue_requeue(
>>>>>>> >>>>>     _Scheduler_simple_Ready_queue_enqueue( the_thread );
>>>>>>> >>>>>   }
>>>>>>> >>>>>
>>>>>>> >>>>>+RTEMS_INLINE_ROUTINE bool
>>>>>>> >>>>> _Scheduler_simple_Insert_as_first_order(
>>>>>>> >>>>>+  const Chain_Node *to_insert,
>>>>>>> >>>>>+  const Chain_Node *next
>>>>>>> >>>>>+)
>>>>>>> >>>>>+{
>>>>>>> >>>>>+  const Thread_Control *thread_to_insert = (const
>>>>>>> >>>>> Thread_Control *)
>>>>>>> >>>>>to_insert;
>>>>>>> >>>>>+  const Thread_Control *thread_next = (const Thread_Control *)
>>>>>>> >>>>> next;
>>>>>>> >>>>>+
>>>>>>> >>>>>+  return thread_to_insert->current_priority <=
>>>>>>> >>>>>thread_next->current_priority;
>>>>>>> >>>>>+}
>>>>>>> >>>>>+
>>>>>>> >>>>>+RTEMS_INLINE_ROUTINE bool
>>>>>>> >>>>> _Scheduler_simple_Insert_as_last_order(
>>>>>>> >>>>>+  const Chain_Node *to_insert,
>>>>>>> >>>>>+  const Chain_Node *next
>>>>>>> >>>>>+)
>>>>>>> >>>>>+{
>>>>>>> >>>>>+  const Thread_Control *thread_to_insert = (const
>>>>>>> >>>>> Thread_Control *)
>>>>>>> >>>>>to_insert;
>>>>>>> >>>>>+  const Thread_Control *thread_next = (const Thread_Control *)
>>>>>>> >>>>> next;
>>>>>>> >>>>>+
>>>>>>> >>>>>+  return thread_to_insert->current_priority <
>>>>>>> >>>>>thread_next->current_priority;
>>>>>>> >>>>>+}
>>>>>>> >>>>>+
>>>>>>> >>>>>+RTEMS_INLINE_ROUTINE void _Scheduler_simple_Insert_as_first(
>>>>>>> >>>>>+  Chain_Control *chain,
>>>>>>> >>>>>+  Thread_Control *to_insert
>>>>>>> >>>>>+)
>>>>>>> >>>>>+{
>>>>>>> >>>>>+  _Chain_Insert_ordered_unprotected(
>>>>>>> >>>>>+    chain,
>>>>>>> >>>>>+    &to_insert->Object.Node,
>>>>>>> >>>>>+    _Scheduler_simple_Insert_as_first_order
>>>>>>> >>>>>+  );
>>>>>>> >>>>>+}
>>>>>>> >>>>>+
>>>>>>> >>>>>+RTEMS_INLINE_ROUTINE void _Scheduler_simple_Insert_as_last(
>>>>>>> >>>>>+  Chain_Control *chain,
>>>>>>> >>>>>+  Thread_Control *to_insert
>>>>>>> >>>>>+)
>>>>>>> >>>>>+{
>>>>>>> >>>>>+  _Chain_Insert_ordered_unprotected(
>>>>>>> >>>>>+    chain,
>>>>>>> >>>>>+    &to_insert->Object.Node,
>>>>>>> >>>>>+    _Scheduler_simple_Insert_as_last_order
>>>>>>> >>>>>+  );
>>>>>>> >>>>>+}
>>>>>>
>>>>>> >>>>
>>>>>> >>>>For these functions _Scheduler_simple_Insert_as_first() and
>>>>>> >>>>_Scheduler_simple_Insert_as_last() the name means they break
>>>>>> >>>> priority
>>>>>> >>>>ties by FIFO and LIFO semantics? Before I read the code, I thought
>>>>>> >>>> it
>>>>>> >>>>meant break ties by inserting to the first or last position of
>>>>>> >>>> ties,
>>>>>> >>>>which is the opposite. I would prefer to have
>>>>>> >>>>_Scheduler_simple_Insert_as_fifo()/lifo() instead for added
>>>>>> >>>> clarity.
>>>>>> >>>>
>>>>>
>>>>> >>>I guess saying fifo/lifo could be confusing also, since the insert
>>>>> >>> is
>>>>> >>>not fifo/lifo, just how priority ties are broken. So maybe
>>>>> >>>_Scheduler_simple_Insert_priority_fifo()/lifo()?
>>>>
>>>> >>
>>>> >>
>>>> >>Suppose we have 0 < 5_0 <= 5_1 < 10, then insert 5_2 as first will
>>>> >> yield
>>>> >>
>>>> >>0 < 5_2 <= 5_0 <= 5_1 < 10
>>>> >>
>>>> >>and insert 5_2 as last will yield
>>>> >>
>>>> >>0 < 5_0 <= 5_1 <= 5_2 < 10
>>>> >>
>>>> >>I will use your suggestions.
>>>> >>
>>>
>>> >
>>> >Now I think that the LIFO/FIFO talks to much about a particular use
>>> > case.
>>> >What about
>>> >
>>> >_Scheduler_simple_Insert_as_first_of_priority_group()
>>> >
>>
>> That would be acceptable if you prefer. It is usual to specify whether
>> ties are broken by lifo/fifo. A priority queue that breaks ties with
>> FIFO is called "stable".
>>
>
> Ok, the name is also a bit long.  Is it useful to change the scheduler
> operation methods as well, e.g. _Scheduler_Enqueue() ->
> _Scheduler_Enqueue_fifo()?
>
It is up to the specific scheduler whether it will implement FIFO to
break ties in priority. Maybe they all do, in which case the rename
would be OK, but I'm not sure. We could, however, rename
_Scheduler_Enqueue_first() to be _Scheduler_Enqueue_lifo().

>
> --
> Sebastian Huber, embedded brains GmbH
>
> Address : Dornierstr. 4, D-82178 Puchheim, Germany
> Phone   : +49 89 189 47 41-16
> Fax     : +49 89 189 47 41-09
> E-Mail  : sebastian.huber at embedded-brains.de
> PGP     : Public key available on request.
>
> Diese Nachricht ist keine geschäftliche Mitteilung im Sinne des EHUG.
>