InitRech 2015/2016, sujet 28 - Historique des versions

Nrichez : /* Summary */

2016-06-19T19:11:55Z

‎Summary

2016-06-19T18:59:59Z

‎Summary

2016-06-19T18:58:02Z

‎Applications

2016-06-19T17:55:28Z

‎Main contribution

2016-06-19T17:26:04Z

‎Summary

2016-06-19T13:37:46Z

Page créée avec « =Summary= =Main contribution= =Applications= »

Nouvelle page

=Summary=

=Main contribution=

=Applications=

@@ Ligne 35 : / Ligne 35 : @@
-Finally, researchers take into account the problem with extended precedences.
+Finally, researchers took into account the problem with extended precedences.
-They have extended the previous results to periodic extended precedences. But they considered in detail how instances of the tasks are related. For each precedences, only the initial release times must be adjusted in order to respects the previous property for all task instances. Moreover, deadlines relative are modified in the same way as simple precedences.
+They extended the previous results to periodic extended precedences. But they considered in detail how instances of the tasks were related. For each precedences, only the initial release times must be adjusted in order to respect the previous property for all task instances. Moreover, deadlines relative are modified in the same way as simple precedences.
 =Main contribution=

@@ Ligne 1 : / Ligne 1 : @@
 =Summary=
-This article speaks about different scheduling techniques of dependent periodic tasks, without synchronization mechanisms.
+This article is about different scheduling techniques of dependent periodic tasks, without synchronization mechanisms.
 The main goal of these techniques is to design algorithms for the scheduling of critical embedded systems by fixing priority scheduling policies for different classes of dependent tasks.
@@ Ligne 20 : / Ligne 20 : @@
-Researchers were initially inspired of a schedulability test for periodic tasks with simple precedences in general.
+Researchers were initially inspired by a schedulability test for periodic tasks with simple precedences in general.
-Tasks can then be scheduled using the deadline-monotonic (DM) policy but the respect of precedence constraints is ensured by semaphores. Then, they use a different approach by encoding precedences in task real-time attributes, adjusting the release times and the deadlines of the tasks in addition to earliest-deadline-first policy (EDF).
+Then, tasks can be scheduled using the deadline-monotonic (DM) policy but the respect of precedence constraints is ensured by semaphores. Then, they use a different approach by encoding precedences in task real-time attributes, adjusting the release times and the deadlines in addition to earliest-deadline-first policy (EDF).

← Version précédente		Version du 19 juin 2016 à 18:58
Ligne 56 :		Ligne 56 :

	=Applications=		=Applications=
		+
		+	This research can improve stability and compactness in embedded systems.
		+
		+	For spacecraft and aircraft flight control systems for instance, we need to supervise the speed, attitude and position of the vehicle in real-time.
		+
		+	For a critical embedded system, we cannot afford an unexpected failure from a bad scheduling.
		+
		+
		+	This kind of scheduling is very interesting for all systems considered as a real-time system, having lot of different input and output to manage.
		+
		+	We can use this to improve the security of self-driving cars for instance.
		+
		+	There are many parameters to manage, like distance, braking, cornering, engine, tracks, environment...

← Version précédente		Version du 19 juin 2016 à 17:55
Ligne 40 :		Ligne 40 :

	=Main contribution=		=Main contribution=
		+
		+	This article studies the scheduling of critical embedded systems, without synchronization mechanisms to prevent deadlocks or scheduling anomalies.
		+
		+
		+	If we do not allow semaphore synchronizations, the system is not schedulable with a static priority policy.
		+
		+	Indeed, the task graph must be unfolded for the execution and this can lead to important computation overhead and high memory consumption as the scheduler needs to make its decisions according to a large task set.
		+
		+
		+	With semaphore, during the execution, the system becomes unschedulable due to a task that does not take its complete wcet to execute.
		+
		+	Moreover, known schedulability tests for scheduling priorities with synchronization mechanisms only provide sufficient schedulability conditions and not necessary schedulability conditions.
		+
		+	At last, highly critical systems must go though certification processes and a system with policies without semaphore synchronizations are easier to certify, because of a more compact Operating System.

	=Applications=		=Applications=