Library rt.model.jitter.platform
Require Import rt.util.all.
Require Import rt.model.jitter.task rt.model.jitter.job rt.model.jitter.schedule
rt.model.jitter.priority rt.model.jitter.task_arrival rt.model.jitter.interference.
From mathcomp Require Import ssreflect ssrbool ssrfun eqtype ssrnat seq fintype bigop.
Module Platform.
Import Job SporadicTaskset ScheduleOfSporadicTaskWithJitter SporadicTaskset SporadicTaskArrival Interference Priority.
Section Properties.
Context {sporadic_task: eqType}.
Variable task_cost: sporadic_task → time.
Variable task_period: sporadic_task → time.
Variable task_deadline: sporadic_task → time.
Context {Job: eqType}.
Variable job_cost: Job → time.
Variable job_deadline: Job → time.
Variable job_task: Job → sporadic_task.
Variable job_jitter: Job → time.
(* Assume any job arrival sequence... *)
Context {arr_seq: arrival_sequence Job}.
(* Consider any schedule. *)
Context {num_cpus: nat}.
Variable sched: schedule num_cpus arr_seq.
Section Execution.
(* A scheduler is work-conserving iff when a job j is backlogged,
all processors are busy with other jobs.
Note: backlogged means that jitter has already passed. *)
Definition work_conserving :=
∀ j t,
backlogged job_cost job_jitter sched j t →
∀ cpu, ∃ j_other,
scheduled_on sched j_other cpu t.
(* We also provide an alternative, equivalent definition of work-conserving
based on counting the number of scheduled jobs. *)
Definition work_conserving_count :=
∀ j t,
backlogged job_cost job_jitter sched j t →
size (jobs_scheduled_at sched t) = num_cpus.
End Execution.
Section JLDP.
(* A JLFP/JLDP policy ...*)
Variable higher_eq_priority: JLDP_policy arr_seq.
(* ... is enforced by the scheduler iff at any time t,
a scheduled job has higher (or same) priority than (as)
a backlogged job. *)
Definition enforces_JLDP_policy :=
∀ (j j_hp: JobIn arr_seq) t,
backlogged job_cost job_jitter sched j t →
scheduled sched j_hp t →
higher_eq_priority t j_hp j.
End JLDP.
Section FP.
(* Given an FP policy, ...*)
Variable higher_eq_priority: FP_policy sporadic_task.
(* ... this policy is enforced by the scheduler iff
a corresponding JLDP policy is enforced by the scheduler. *)
Definition enforces_FP_policy :=
enforces_JLDP_policy (FP_to_JLDP job_task higher_eq_priority).
End FP.
Section Lemmas.
(* Assume all jobs have valid parameters, ...*)
Hypothesis H_valid_job_parameters:
∀ (j: JobIn arr_seq),
valid_sporadic_job task_cost task_deadline job_cost job_deadline job_task j.
(* In this section, we prove that the two definitions of work-conserving are equivalent. *)
Section EquivalentDefinitions.
Lemma work_conserving_eq_work_conserving_count :
work_conserving ↔ work_conserving_count.
Proof.
unfold work_conserving, work_conserving_count; split.
{
intros EX j t BACK.
specialize (EX j t BACK).
apply eq_trans with (y := size (enum (processor num_cpus)));
last by rewrite size_enum_ord.
unfold jobs_scheduled_at.
apply eq_trans with (y := size ((\cat_(cpu < num_cpus) map (fun x ⇒ Some x)
(make_sequence (sched cpu t)))));
first by rewrite -map_bigcat_ord size_map.
apply eq_trans with (y := size (\cat_(cpu < num_cpus) [:: sched cpu t])).
{
f_equal; apply eq_bigr; intros cpu _.
destruct (sched cpu t) eqn:SCHED; first by done.
by specialize (EX cpu); des; move: EX ⇒ /eqP EX; rewrite EX in SCHED.
}
rewrite size_bigcat_ord.
apply eq_trans with (y := \sum_(i < num_cpus) 1);
last by rewrite big_const_ord iter_addn mul1n addn0 size_enum_ord.
by apply eq_bigr.
}
{
intros SIZE j t BACK cpu.
specialize (SIZE j t BACK).
destruct ([∃ cpu, sched cpu t = None]) eqn:EX; last first.
{
apply negbT in EX; rewrite negb_exists in EX.
move: EX ⇒ /forallP ALL; specialize (ALL cpu).
destruct (sched cpu t) eqn:SOME; last by done.
by ∃ j0; apply/eqP.
}
{
move: EX ⇒ /existsP [cpu' /eqP EX].
unfold jobs_scheduled_at in SIZE.
move: SIZE ⇒ /eqP SIZE; rewrite -[size _ = _]negbK in SIZE.
move: SIZE ⇒ /negP SIZE; exfalso; apply SIZE; clear SIZE.
rewrite neq_ltn; apply/orP; left.
rewrite size_bigcat_ord.
rewrite → big_mkord_ord with (x0 := 0).
have MKORD := big_mkord (fun x ⇒ true); rewrite -MKORD.
have CAT := big_cat_nat _ (fun x ⇒ true).
rewrite → CAT with (n := cpu'); [simpl | by done | by apply ltnW, ltn_ord].
assert (DIFF: ∃ k, num_cpus = (cpu' + k).+1).
{
∃ (num_cpus - cpu').-1.
rewrite -addnS prednK; last by rewrite ltn_subRL addn0 ltn_ord.
rewrite addnBA; last by apply ltnW, ltn_ord.
by rewrite addnC -addnBA // subnn addn0.
}
des; rewrite {5}DIFF.
rewrite big_nat_recl; last by apply leq_addr.
apply leq_trans with (n := (\sum_(0 ≤ i < cpu') 1) + 1 + (\sum_(cpu' ≤ i < cpu' + k) 1));
last first.
{
rewrite 2!big_const_nat 2!iter_addn 2!mul1n addn0 subn0.
rewrite [cpu' + k]addnC -addnBA // subnn 2!addn0.
by rewrite -addnA [1 + k]addnC addnA addn1 -DIFF.
}
{
rewrite -addn1 addnC [_ + 1]addnC -addnA.
apply leq_add; first by done.
rewrite eq_fun_ord_to_nat; unfold make_sequence at 2; rewrite EX /= add0n.
apply leq_add; apply leq_sum; ins; unfold fun_ord_to_nat; des_eqrefl; try done;
by unfold make_sequence; desf.
}
}
}
Qed.
End EquivalentDefinitions.
End Lemmas.
End Properties.
End Platform.
Require Import rt.model.jitter.task rt.model.jitter.job rt.model.jitter.schedule
rt.model.jitter.priority rt.model.jitter.task_arrival rt.model.jitter.interference.
From mathcomp Require Import ssreflect ssrbool ssrfun eqtype ssrnat seq fintype bigop.
Module Platform.
Import Job SporadicTaskset ScheduleOfSporadicTaskWithJitter SporadicTaskset SporadicTaskArrival Interference Priority.
Section Properties.
Context {sporadic_task: eqType}.
Variable task_cost: sporadic_task → time.
Variable task_period: sporadic_task → time.
Variable task_deadline: sporadic_task → time.
Context {Job: eqType}.
Variable job_cost: Job → time.
Variable job_deadline: Job → time.
Variable job_task: Job → sporadic_task.
Variable job_jitter: Job → time.
(* Assume any job arrival sequence... *)
Context {arr_seq: arrival_sequence Job}.
(* Consider any schedule. *)
Context {num_cpus: nat}.
Variable sched: schedule num_cpus arr_seq.
Section Execution.
(* A scheduler is work-conserving iff when a job j is backlogged,
all processors are busy with other jobs.
Note: backlogged means that jitter has already passed. *)
Definition work_conserving :=
∀ j t,
backlogged job_cost job_jitter sched j t →
∀ cpu, ∃ j_other,
scheduled_on sched j_other cpu t.
(* We also provide an alternative, equivalent definition of work-conserving
based on counting the number of scheduled jobs. *)
Definition work_conserving_count :=
∀ j t,
backlogged job_cost job_jitter sched j t →
size (jobs_scheduled_at sched t) = num_cpus.
End Execution.
Section JLDP.
(* A JLFP/JLDP policy ...*)
Variable higher_eq_priority: JLDP_policy arr_seq.
(* ... is enforced by the scheduler iff at any time t,
a scheduled job has higher (or same) priority than (as)
a backlogged job. *)
Definition enforces_JLDP_policy :=
∀ (j j_hp: JobIn arr_seq) t,
backlogged job_cost job_jitter sched j t →
scheduled sched j_hp t →
higher_eq_priority t j_hp j.
End JLDP.
Section FP.
(* Given an FP policy, ...*)
Variable higher_eq_priority: FP_policy sporadic_task.
(* ... this policy is enforced by the scheduler iff
a corresponding JLDP policy is enforced by the scheduler. *)
Definition enforces_FP_policy :=
enforces_JLDP_policy (FP_to_JLDP job_task higher_eq_priority).
End FP.
Section Lemmas.
(* Assume all jobs have valid parameters, ...*)
Hypothesis H_valid_job_parameters:
∀ (j: JobIn arr_seq),
valid_sporadic_job task_cost task_deadline job_cost job_deadline job_task j.
(* In this section, we prove that the two definitions of work-conserving are equivalent. *)
Section EquivalentDefinitions.
Lemma work_conserving_eq_work_conserving_count :
work_conserving ↔ work_conserving_count.
Proof.
unfold work_conserving, work_conserving_count; split.
{
intros EX j t BACK.
specialize (EX j t BACK).
apply eq_trans with (y := size (enum (processor num_cpus)));
last by rewrite size_enum_ord.
unfold jobs_scheduled_at.
apply eq_trans with (y := size ((\cat_(cpu < num_cpus) map (fun x ⇒ Some x)
(make_sequence (sched cpu t)))));
first by rewrite -map_bigcat_ord size_map.
apply eq_trans with (y := size (\cat_(cpu < num_cpus) [:: sched cpu t])).
{
f_equal; apply eq_bigr; intros cpu _.
destruct (sched cpu t) eqn:SCHED; first by done.
by specialize (EX cpu); des; move: EX ⇒ /eqP EX; rewrite EX in SCHED.
}
rewrite size_bigcat_ord.
apply eq_trans with (y := \sum_(i < num_cpus) 1);
last by rewrite big_const_ord iter_addn mul1n addn0 size_enum_ord.
by apply eq_bigr.
}
{
intros SIZE j t BACK cpu.
specialize (SIZE j t BACK).
destruct ([∃ cpu, sched cpu t = None]) eqn:EX; last first.
{
apply negbT in EX; rewrite negb_exists in EX.
move: EX ⇒ /forallP ALL; specialize (ALL cpu).
destruct (sched cpu t) eqn:SOME; last by done.
by ∃ j0; apply/eqP.
}
{
move: EX ⇒ /existsP [cpu' /eqP EX].
unfold jobs_scheduled_at in SIZE.
move: SIZE ⇒ /eqP SIZE; rewrite -[size _ = _]negbK in SIZE.
move: SIZE ⇒ /negP SIZE; exfalso; apply SIZE; clear SIZE.
rewrite neq_ltn; apply/orP; left.
rewrite size_bigcat_ord.
rewrite → big_mkord_ord with (x0 := 0).
have MKORD := big_mkord (fun x ⇒ true); rewrite -MKORD.
have CAT := big_cat_nat _ (fun x ⇒ true).
rewrite → CAT with (n := cpu'); [simpl | by done | by apply ltnW, ltn_ord].
assert (DIFF: ∃ k, num_cpus = (cpu' + k).+1).
{
∃ (num_cpus - cpu').-1.
rewrite -addnS prednK; last by rewrite ltn_subRL addn0 ltn_ord.
rewrite addnBA; last by apply ltnW, ltn_ord.
by rewrite addnC -addnBA // subnn addn0.
}
des; rewrite {5}DIFF.
rewrite big_nat_recl; last by apply leq_addr.
apply leq_trans with (n := (\sum_(0 ≤ i < cpu') 1) + 1 + (\sum_(cpu' ≤ i < cpu' + k) 1));
last first.
{
rewrite 2!big_const_nat 2!iter_addn 2!mul1n addn0 subn0.
rewrite [cpu' + k]addnC -addnBA // subnn 2!addn0.
by rewrite -addnA [1 + k]addnC addnA addn1 -DIFF.
}
{
rewrite -addn1 addnC [_ + 1]addnC -addnA.
apply leq_add; first by done.
rewrite eq_fun_ord_to_nat; unfold make_sequence at 2; rewrite EX /= add0n.
apply leq_add; apply leq_sum; ins; unfold fun_ord_to_nat; des_eqrefl; try done;
by unfold make_sequence; desf.
}
}
}
Qed.
End EquivalentDefinitions.
End Lemmas.
End Properties.
End Platform.