Library prosa.implementation.refinements.task

Task Refinement

In this module, we define a generic version of concrete task and prove that it is isomorphic to the standard, natural-number-based definition.
We begin by defining the generic task type.
Section Definitions.

In this file, we refine the Prosa standard implementation of jobs and tasks.
  Definition Task := [eqType of concrete_task].
  Definition Job := [eqType of concrete_job].

Consider a generic type T, for which all the basic arithmetical operations, ordering, and equality are defined.
  Context `{T : Type}.
  Context `{!zero_of T, !one_of T, !sub_of T, !add_of T, !mul_of T,
            !div_of T, !mod_of T, !eq_of T, !leq_of T, !lt_of T}.
  Context `{!eq_of (seq T)}.
  Context `{!eq_of (@task_arrivals_bound_T T)}.

We define a generic task type that uses T for each numerical value.
  Structure task_T :=
    {
    task_id_T: T;
    task_cost_T: T;
    task_arrival_T: @task_arrivals_bound_T T;
    task_deadline_T: T;
    task_priority_T: T
    }.

Further, we define the equality for two generic tasks as the equality of each attribute.
  Definition task_eqdef_T (t1 t2: task_T) :=
    (task_id_T t1 == task_id_T t2)%C &&
    (task_cost_T t1 == task_cost_T t2)%C &&
    (task_arrival_T t1 == task_arrival_T t2)%C &&
    (task_deadline_T t1 == task_deadline_T t2)%C &&
    (task_priority_T t1 == task_priority_T t2)%C.

Next, we define a helper function that convert a minimum inter-arrival time to an arrival curve prefix.
Further, we define the arrival bound of a task as, possibly, (1) periodic, (2) sporadic, or (3) an arrival curve. Note that, internally, the workload bound is always represented by an arrival curve.
  Definition get_extrapolated_arrival_curve_T (tsk : task_T) : T × seq (T × T) :=
    match task_arrival_T tsk with
    | Periodic_T pinter_arrival_to_extrapolated_arrival_curve_T p
    | Sporadic_T minter_arrival_to_extrapolated_arrival_curve_T m
    | ArrivalPrefix_T stepssteps
    end.

With the previous definition in place, we define the workload bound...
... and the task request-bound function.
  Definition task_rbf_T (tsk : task_T) (Δ : T) :=
    (task_cost_T tsk × ConcreteMaxArrivals_T tsk Δ)%C.

Further, we define a valid arrival bound as (1) a positive minimum inter-arrival time/period, or (2) a valid arrival curve prefix.
  Definition valid_arrivals_T (tsk : task_T) : bool :=
    match task_arrival_T tsk with
    | Periodic_T p ⇒ (1 p)%C
    | Sporadic_T m ⇒ (1 m)%C
    | ArrivalPrefix_T ac_prefix_vecvalid_extrapolated_arrival_curve_T ac_prefix_vec
    end.

Next, we define a helper function that yields the horizon of the arrival curve prefix of a task ...
... another one that yields the time steps of the arrival curve, ...
... a function that yields the same time steps, offset by δ, ...
... and a generalization of the previous function that repeats the time steps for each given offset.
In this section, we define two functions used to convert from a generic task to the standard task definition, and vice-versa.
Section Translation.

First, we define the function that maps a generic task to a natural-number task...
  Definition taskT_to_task (tsk: @task_T N) : Task :=
    match tsk with
    | {| task_id_T := id;
         task_cost_T := cost;
         task_arrival_T := arrival_bound;
         task_deadline_T := deadline;
         task_priority_T := priority |}
      ⇒
      {| task.task_id := nat_of_bin id;
         task.task_cost := nat_of_bin cost;
         task.task_arrival := task_abT_to_task_ab arrival_bound;
         task.task_deadline := nat_of_bin deadline;
         task.task_priority := nat_of_bin priority |}
    end.

... and its function relationship.
  Definition Rtask := fun_hrel taskT_to_task.

Finally, we define the converse function, mapping a natural-number task to a generic one.
  Definition task_to_taskT (tsk: Task) : @task_T N :=
    match tsk with
    | {| task.task_id := id;
         task.task_cost := cost;
         task.task_arrival := arrival_bound;
         task.task_deadline := deadline;
         task.task_priority := priority |}
      ⇒
      {| task_id_T := bin_of_nat id;
         task_cost_T := bin_of_nat cost;
         task_arrival_T := task_ab_to_task_abT arrival_bound;
         task_deadline_T := bin_of_nat deadline;
         task_priority_T := bin_of_nat priority |}
    end.

End Translation.

In this fairly technical section, we prove a series of refinements aimed to be able to convert between a standard natural-number task and a generic task.
Section Theory.

We prove the refinement of the constructor functions.
  Global Instance refine_task :
    refines (Rnat ==> Rnat ==> Rtask_ab ==> Rnat ==> Rnat ==> Rtask)%rel
            Build_concrete_task Build_task_T.
  Proof.
    by rewrite refinesE_ i <- _ c <- _ p <- _ d <- _ pr <-.
  Qed.

Next, we prove a refinement for the task id.
  Global Instance refine_task_id :
    refines (Rtask ==> Rnat)%rel task_id task_id_T.
  Proof.
    rewrite refinesE_ tsk <-.
    by destruct tsk.
  Qed.

Next, we prove a refinement for the task cost.
  Global Instance refine_task_cost :
    refines (Rtask ==> Rnat)%rel task_cost task_cost_T.
  Proof.
    rewrite refinesE_ tsk <-.
    by destruct tsk.
  Qed.

Next, we prove a refinement for the task arrival.
  Global Instance refine_task_arrival :
    refines (Rtask ==> Rtask_ab)%rel task_arrival task_arrival_T.
  Proof.
    rewrite refinesE_ tsk <-.
    by destruct tsk.
  Qed.

Next, we prove a refinement for the task deadline.
  Global Instance refine_task_deadline :
    refines (Rtask ==> Rnat)%rel task_deadline task_deadline_T.
  Proof.
    rewrite refinesE_ tsk <-.
    by destruct tsk.
  Qed.

Next, we prove a refinement for the task priority.
  Global Instance refine_task_priority :
    refines (Rtask ==> Rnat)%rel task_priority task_priority_T.
  Proof.
    rewrite refinesE_ tsk <-.
    by destruct tsk.
  Qed.

Next, we prove a refinement for the task period.
  Global Instance refine_Periodic :
    refines (Rnat ==> Rtask_ab)%rel Periodic Periodic_T.
  Proof.
    rewrite refinesEt t' Rt.
    compute in Rt; subst.
    by compute.
  Qed.

Next, we prove a refinement for the task minimum inter-arrival time.
  Global Instance refine_Sporadic :
    refines (Rnat ==> Rtask_ab)%rel Sporadic Sporadic_T.
  Proof.
    rewrite refinesEt t' Rt.
    compute in Rt; subst.
    by compute.
  Qed.

Next, we prove a refinement for the task id conversion.
  Local Instance refine_task' :
    refines (unify (A:= task_T) ==> Rtask)%rel taskT_to_task id.
  Proof.
    rewrite refinesEtsk tsk' UNI.
    by have UNIT := unify_rel; last specialize (UNIT _ _ _ UNI); subst tsk'; clear UNI.
  Qed.

Finally, we prove a refinement for the id conversion applied to a list of tasks.
  Local Instance refine_task_set' :
    refines (unify (A:= seq task_T) ==> list_R Rtask)%rel (map taskT_to_task) id.
  Proof.
    rewrite refinesEts ts' UNI.
    have UNIT := unify_rel; last specialize (UNIT _ _ _ UNI); subst ts'; clear UNI.
    induction ts as [ | tsk ts].
    - by simpl; apply list_R_nil_R.
    - simpl; apply list_R_cons_R; last by done.
      by unfold Rtask, fun_hrel.
  Qed.

End Theory.

For convenience, we define a simple notation to declare concrete tasks using numbers represented in binary, adopted in POET's certificates.
Require Export NArith.

Notation "{| 'id:' c1 'cost:' c2 'deadline:' c3 'arrival:' c4 'priority:' c5 }" := {|
  task_id_T := c1;
  task_cost_T := c2;
  task_deadline_T := c3;
  task_arrival_T := c4;
  task_priority_T := c5 |} (at level 6,
                            right associativity,
                            only parsing ).

Notation "{| 'id:' c1 'cost:' c2 'deadline:' c3 'arrival:' c4 }" := {|
  task_id_T := c1;
  task_cost_T := c2;
  task_deadline_T := c3;
  task_arrival_T := c4;
  task_priority_T := 0%N |} (at level 6,
                             right associativity,
                             only parsing ).