Library prosa.analysis.abstract.restricted_supply.bounded_bi.elf

Require Export prosa.analysis.facts.blocking_bound.elf.
Require Export prosa.analysis.abstract.restricted_supply.abstract_rta.
Require Export prosa.analysis.abstract.restricted_supply.iw_instantiation.
Require Export prosa.analysis.abstract.restricted_supply.bounded_bi.aux.
Require Export prosa.analysis.definitions.sbf.busy.
Require Export prosa.analysis.facts.priority.jlfp_with_fp.

Sufficient Condition for Bounded Busy Intervals for RS ELF

In this section, we show that the existence of L such that B + total_hep_rbf L ≤ SBF L, where B is the blocking bound and SBF is a supply-bound function, is a sufficient condition for the existence of bounded busy intervals under ELF scheduling with a restricted-supply processor model.

Section BoundedBusyIntervals.

Consider any type of tasks with relative priority-points ...

  Context {Task : TaskType}.
  Context `{TaskCost Task}.
  Context `{PriorityPoint Task}.

... and any type of jobs associated with these tasks.

  Context {Job : JobType}.
  Context `{JobTask Job Task}.
  Context `{JobArrival Job}.
  Context `{JobCost Job}.

Consider any kind of fully supply-consuming unit-supply uniprocessor model.

  Context `{PState : ProcessorState Job}.
  Hypothesis H_uniprocessor_proc_model : uniprocessor_model PState.
  Hypothesis H_unit_supply_proc_model : unit_supply_proc_model PState.
  Hypothesis H_consumed_supply_proc_model : fully_consuming_proc_model PState.

Consider an FP policy that indicates a higher-or-equal priority relation, and assume that the relation is reflexive, transitive. and total.

  Context (FP : FP_policy Task).
  Hypothesis H_priority_is_reflexive : reflexive_task_priorities FP.
  Hypothesis H_priority_is_transitive : transitive_task_priorities FP.
  Hypothesis H_total_priorities : total_task_priorities FP.

Consider any valid arrival sequence.

Variable arr_seq : arrival_sequence Job.
Hypothesis H_valid_arrival_sequence : valid_arrival_sequence arr_seq.

Next, consider a schedule of this arrival sequence, ...

Variable sched : schedule PState.

... allow for any work-bearing notion of job readiness, ...

Context `{!JobReady Job PState}.
Hypothesis H_job_ready : work_bearing_readiness arr_seq sched.

... and assume that the schedule is valid.

Hypothesis H_sched_valid : valid_schedule sched arr_seq.

Assume that jobs have bounded non-preemptive segments.

  Context `{JobPreemptable Job}.
  Context `{TaskMaxNonpreemptiveSegment Task}.
  Hypothesis H_valid_preemption_model : valid_preemption_model arr_seq sched.
  Hypothesis H_valid_model_with_bounded_nonpreemptive_segments :
    valid_model_with_bounded_nonpreemptive_segments arr_seq sched.

Further assume that the schedule follows the ELF scheduling policy.

Hypothesis H_respects_policy :
respects_JLFP_policy_at_preemption_point arr_seq sched (ELF FP).

Recall that busy_intervals_are_bounded_by is an abstract notion. Hence, we need to introduce interference and interfering workload. We will use the restricted-supply instantiations.

We say that job j incurs interference at time t iff it cannot execute due to (1) the lack of supply at time t, (2) service inversion (i.e., a lower-priority job receiving service at t), or a higher-or-equal-priority job receiving service.

#[local] Instance rs_jlfp_interference : Interference Job :=
rs_jlfp_interference arr_seq sched.

The interfering workload, in turn, is defined as the sum of the blackout predicate, service inversion predicate, and the interfering workload of jobs with higher or equal priority.

#[local] Instance rs_jlfp_interfering_workload : InterferingWorkload Job :=
rs_jlfp_interfering_workload arr_seq sched.

Assume that the schedule is work-conserving in the abstract sense.

Hypothesis H_work_conserving : definitions.work_conserving arr_seq sched.

Consider an arbitrary task set ts, ...

Variable ts : list Task.

... assume that all jobs come from the task set, ...

Hypothesis H_all_jobs_from_taskset : all_jobs_from_taskset arr_seq ts.

... and that the cost of a job does not exceed its task's WCET.

Hypothesis H_valid_job_cost : arrivals_have_valid_job_costs arr_seq.

Let max_arrivals be a family of valid arrival curves, i.e., for any task tsk in ts, max_arrival tsk is (1) an arrival bound of tsk, and (2) it is a monotonic function that equals 0 for the empty interval = 0.

  Context `{MaxArrivals Task}.
  Hypothesis H_valid_arrival_curve : valid_taskset_arrival_curve ts max_arrivals.
  Hypothesis H_is_arrival_curve : taskset_respects_max_arrivals arr_seq ts.

Let tsk be any task in ts that is to be analyzed.

Variable tsk : Task.
Hypothesis H_tsk_in_ts : tsk \in ts.

Consider a unit SBF valid in busy intervals (w.r.t. task tsk). That is, (1) SBF 0 = 0, (2) for any duration Δ, the supply produced during a busy-interval prefix of length Δ is at least SBF Δ, and (3) SBF makes steps of at most one.

  Context {SBF : SupplyBoundFunction}.
  Hypothesis H_valid_SBF : valid_busy_sbf arr_seq sched tsk SBF.
  Hypothesis H_unit_SBF : unit_supply_bound_function SBF.

The next step is to establish a bound on the maximum busy-window length, which aRSA requires to be given. To this end, let L be any positive fixed point of the busy-interval recurrence. As the lemma busy_intervals_are_bounded_rs_elf shows, under ELF scheduling, this is sufficient to guarantee that all busy intervals are bounded by L.

  Variable L : duration.
  Hypothesis H_L_positive : 0 < L.

  Hypothesis H_fixed_point:
    ∀ (A : duration),
      blocking_bound ts tsk A + total_hep_request_bound_function_FP ts tsk L ≤ SBF L.

Next, we provide a step-by-step proof of busy-interval boundedness.

Section StepByStepProof.

Consider any job j of task tsk that has a positive job cost and is in the arrival sequence.

    Variable j : Job.
    Hypothesis H_j_arrives : arrives_in arr_seq j.
    Hypothesis H_job_of_tsk : job_of_task tsk j.
    Hypothesis H_job_cost_positive : job_cost_positive j.

Consider to be the start of the busy-interval.

Variable : instant.

Now we have two cases: (1) when the busy-interval prefix continues until time instant + L and (2) when the busy interval prefix terminates earlier. In either case, we can show that the busy-interval prefix is bounded.

We start with the first case, where the busy-interval prefix continues until time instant + L.

Section Case1.

Consider that [t1, job_arrival j] and [t1, t1 + L) are both busy-interval prefixes of job j. Note that at this point we do not necessarily know that job_arrival j ≤ L; hence, we assume the existence of both prefixes.

Hypothesis H_busy_prefix_arr : busy_interval_prefix arr_seq sched j (job_arrival j ).+1.
Hypothesis H_busy_prefix_L : busy_interval_prefix arr_seq sched j ( + L).

The crucial point to note is that the sum of the job's cost (represented as workload_of_job) and the interfering workload in the interval [t1, t1 + L) is bounded by L due to the fixed point H_fixed_point.

      Local Lemma workload_is_bounded :
        workload_of_job arr_seq j ( + L) + cumulative_interfering_workload j ( + L) ≤ L.
      Proof.
        rewrite (cumulative_interfering_workload_split _ _ _).
        rewrite (leqRW (blackout_during_bound _ _ _ _ _ _ _ _ ( + L) _ _ _)); (try apply H_valid_SBF) ⇒ //.
        rewrite // addnC -!addnA.
        have E: ∀ a b c, a ≤ c → b ≤ c - a → a + b ≤ c by move ⇒ ? ? ? ? ?; lia.
        apply: E; first by lia.
        rewrite subKn; last by apply: sbf_bounded_by_duration ⇒ //.
        specialize (H_fixed_point (job_arrival j - )).
        rewrite -(leqRW H_fixed_point); apply leq_add.
        - apply: leq_trans; first apply: service_inversion_is_bounded ⇒ //.
          + instantiate (1 := fun (A : duration) ⇒ blocking_bound ts tsk A) ⇒ //=.
            by move⇒ jo t t' ARRo TSKo PREFo; apply: nonpreemptive_segments_bounded_by_blocking ⇒ //.
          + by done.
        - rewrite addnC cumulative_iw_hep_eq_workload_of_ohep workload_job_and_ahep_eq_workload_hep //.
          apply workload_of_jobs_bounded ⇒ //.
          move⇒ j'.
          move: H_job_of_tsk ⇒ /eqP .
          by apply hep_job_implies_hep_task.
        Qed.

It follows that + L is a quiet time, which means that the busy prefix ends (i.e., it is bounded).

      Local Lemma busy_prefix_is_bounded_case1 :
        ∃ t2 ,
          job_arrival j <
          ∧ ≤ + L
          ∧ busy_interval arr_seq sched j .
      Proof.
        have PEND : pending sched j (job_arrival j) by apply job_pending_at_arrival ⇒ //.
        enough(∃ t2 , job_arrival j < ∧ ≤ + L ∧ definitions.busy_interval sched j ) as BUSY.
        { have [ [ [ ]]] := BUSY.
          eexists; split; first by exact: .
          split; first by done.
          by apply instantiated_busy_interval_equivalent_busy_interval.
        }
        eapply busy_interval.busy_interval_is_bounded; eauto 2 ⇒ //.
        - by eapply instantiated_i_and_w_no_speculative_execution; eauto 2 ⇒ //.
        - by apply instantiated_busy_interval_prefix_equivalent_busy_interval_prefix ⇒ //.
        - by apply workload_is_bounded ⇒ //.
      Qed.

    End Case1.

Next, we consider the case when the interval [t1, t1 + L) is not a busy-interval prefix.

Section Case2.

Consider that [t1, job_arrival j] is a busy-interval prefix of j and [t1, t1 + L) is not.

      Hypothesis H_arrives : ≤ job_arrival j.
      Hypothesis H_busy_prefix_arr : busy_interval_prefix arr_seq sched j (job_arrival j ).+1.
      Hypothesis H_no_busy_prefix_L : ¬ busy_interval_prefix arr_seq sched j ( + L).

From the properties of busy intervals, one can conclude that j's arrival time is less than + L.

      Local Lemma job_arrival_is_bounded :
        job_arrival j < + L.
      Proof.
        move_neq_up FF.
        move: (H_busy_prefix_arr) (H_busy_prefix_arr) ⇒ PREFIX PREFIXA.
        apply instantiated_busy_interval_prefix_equivalent_busy_interval_prefix in PREFIXA ⇒ //.
        move: (PREFIXA) ⇒ GTC.
        eapply workload_exceeds_interval with (Δ := L) in PREFIX ⇒ //.
        { move_neq_down PREFIX.
          rewrite (cumulative_interfering_workload_split _ _ _).
          rewrite (leqRW (blackout_during_bound _ _ _ _ _ _ _ _ (job_arrival j ).+1 _ _ _)); (try apply H_valid_SBF) ⇒ //.
          rewrite addnC -!addnA.
          have E: ∀ a b c, a ≤ c → b ≤ c - a → a + b ≤ c by move ⇒ ? ? ? ? ?; lia.
          apply: E; first by lia.
          rewrite subKn; last by apply: sbf_bounded_by_duration ⇒ //.
          specialize (H_fixed_point (job_arrival j - )).
          rewrite -(leqRW H_fixed_point); apply leq_add.
          { rewrite (leqRW (service_inversion_widen _ _ _ _ _ (job_arrival j ).+1 _ _ )).
            - apply: leq_trans.
              + apply: service_inversion_is_bounded ⇒ //.
                move ⇒ *; instantiate (1 := fun (A : duration) ⇒ blocking_bound ts tsk A) ⇒ //.
                by apply: nonpreemptive_segments_bounded_by_blocking ⇒ //.
              + by done.
            - by done.
            - lia.
          }
          { rewrite addnC cumulative_iw_hep_eq_workload_of_ohep workload_job_and_ahep_eq_workload_hep //.
            apply workload_of_jobs_bounded ⇒ //.
            move⇒ j'.
            move: H_job_of_tsk ⇒ /eqP .
            by apply hep_job_implies_hep_task. }
        }
      Qed.

Lemma job_arrival_is_bounded implies that the busy-interval prefix starts at time , continues until job_arrival j + 1, and then terminates before + L. Or, in other words, there is point in time such that (1) j's arrival is bounded by , (2) is bounded by + L, and (3) [t1, t2) is the busy interval of job j.

      Local Lemma busy_prefix_is_bounded_case2:
        ∃ t2 , job_arrival j < ∧ ≤ + L ∧ busy_interval arr_seq sched j .
      Proof.
        have LE: job_arrival j < + L
          by apply job_arrival_is_bounded ⇒ //; try apply instantiated_busy_interval_prefix_equivalent_busy_interval_prefix.
        move: (H_busy_prefix_arr) ⇒ PREFIX.
        move: (H_no_busy_prefix_L) ⇒ NOPREF.
        apply instantiated_busy_interval_prefix_equivalent_busy_interval_prefix in PREFIX ⇒ //.
        have BUSY := terminating_busy_prefix_is_busy_interval _ _ _ _ _ _ _ PREFIX.
        edestruct BUSY as [ BUS]; clear BUSY; try apply TSK; eauto 2 ⇒ //.
        { move ⇒ T; apply: NOPREF.
          by apply instantiated_busy_interval_prefix_equivalent_busy_interval_prefix in T ⇒ //.
        }
        ∃ ; split; last split.
        { move: BUS ⇒ [[A _] _]; lia. }
        { move_neq_up FA.
          apply: NOPREF; split; [lia | split; first by apply H_busy_prefix_arr].
          split.
          - move⇒ t NEQ.
            apply abstract_busy_interval_classic_busy_interval_prefix in BUS ⇒ //.
            by apply BUS; lia.
          - lia.
        }
        { by apply instantiated_busy_interval_equivalent_busy_interval ⇒ //. }
      Qed.

    End Case2.

  End StepByStepProof.

Combining the cases analyzed above, we conclude that busy intervals of jobs released by task tsk are bounded by L.

  Lemma busy_intervals_are_bounded_rs_elf :
    busy_intervals_are_bounded_by arr_seq sched tsk L.
  Proof.
    move ⇒ j ARR TSK POS.
    have PEND : pending sched j (job_arrival j) by apply job_pending_at_arrival ⇒ //.
    have [ [GE PREFIX]]:
      ∃ t1 , ≤ job_arrival j
            ∧ busy_interval_prefix arr_seq sched j (job_arrival j).+1
        by apply: busy_interval_prefix_exists.
    ∃ .
    enough(∃ t2 , job_arrival j < ∧ ≤ + L ∧ busy_interval arr_seq sched j ) as BUSY.
    { move: BUSY ⇒ [ [LT [LE BUSY]]]; eexists; split; last first.
      { split; first by exact: LE.
        by apply instantiated_busy_interval_equivalent_busy_interval. }
      { by apply/andP; split. }
    }
    { have [LPREF|NOPREF] := busy_interval_prefix_case ltac:(eauto) j ( + L).
      { apply busy_prefix_is_bounded_case1 ⇒ //.
        by apply instantiated_busy_interval_prefix_equivalent_busy_interval_prefix ⇒ //. }
      { apply busy_prefix_is_bounded_case2⇒ //.
        move⇒ NP; apply: NOPREF.
        by apply instantiated_busy_interval_prefix_equivalent_busy_interval_prefix ⇒ //.
      }
    }
  Qed.

End BoundedBusyIntervals.