run_queue Struct Reference

Per-CPU run queue — schedules strands (execution contexts). More...

Collaboration diagram for run_queue:

Public Member Functions
auto	operator= (const run_queue &)=delete
auto	operator= (run_queue &&)=delete
	run_queue (const run_queue &)=delete
	run_queue (run_queue &&)=delete

Data Fields
i64	avg_vruntime_acc {0}
u64	avg_weight_acc {0}
std::atomic< u64 >	balance_attached_asce {0}
std::atomic< u8 >	balance_attached_dgp_state {0}
std::atomic< u32 >	balance_attached_domain_id {0}
std::atomic< i64 >	balance_avg_vruntime_acc {0}
std::atomic< u64 >	balance_avg_weight_acc {0}
std::atomic< u64 >	balance_cfs_load_avg {0}
std::atomic< u64 >	balance_cfs_runnable_avg {0}
std::atomic< u64 >	balance_cfs_util_avg {0}
std::atomic< u32 >	balance_cpu_capacity_curr {1024}
std::atomic< u64 >	balance_min_vruntime {0}
	Lockless advisory view consumed by remote balancers.
std::atomic< u32 >	balance_nr_running {0}
std::atomic< u32 >	balance_online {0}
sched_avg_rq	cfs_avg {}
	Per-CPU PELT averaging accumulator (mirrors Linux's cfs_rq::avg).
sync::qspinlock::qspinlock	context_lock {}
	Serializes ASCE / CR1 loads during context switch.
u32	cpu_capacity_curr {1024}
	Current CPU capacity — dynamically reduced by RT pressure and restored by tick decay (matches Linux scheduler).
u32	cpu_capacity_orig {1024}
	Original CPU capacity (SMT-aware throughput units, 1024=max).
u16	cpu_id {0}
	Logical CPU ID this run_queue belongs to.
dgp::domain::strand *	curr {nullptr}
	Currently running strand.
lib::list_head	deadline_queue {}
	Conservative EDF deadline queue.
lib::rb_root_cached	fair_tree {}
	EEVDF fair scheduling tree (cached leftmost = earliest deadline).
dgp::domain::strand *	idle_strand {nullptr}
	This CPU's idle strand.
sync::qspinlock::qspinlock	lock {}
	Protects run queue data (enqueue/dequeue/select).
u64	min_vruntime {0}
	Monotonically increasing vruntime floor.
u32	nohz_flags {0}
	Per-CPU nohz idle state (tickless flags).
u32	nr_running {0}
	Total runnable strands on this CPU.
std::atomic< u32 >	online {0}
	Set to 1 when this CPU's run queue is fully initialized.
rt::bw::rt_bandwidth	rt_bw {}
	Per-CPU RT bandwidth controller (1 s period, 950 ms quota).
lib::list_head	rt_queue {}
	RT FIFO queue (domains dequeued in FIFO order).
sd::sched_domain_tree	sd_tree {}
	Per-CPU scheduler-domain hierarchy (SMT → MC → DIE → BOOK → NUMA).

Detailed Description

Per-CPU run queue — schedules strands (execution contexts).

---

The documentation for this struct was generated from the following file:

• zxfoundation/sched/rq.cxxm