Multicore Scheduling & Load Balancing

Multicore Scheduling & Load Balancing

Tasks run on several CPU cores at once. The left part of each lane shows the core's ready queue (head on the left); the right part is the per-core Gantt timeline. Compare one shared ready queue with per-core queues, processor affinity and load balancing.

How to use

• Cores −/+ sets 1–8 cores; Shared / Per-core picks the queue model.
• Balance runs one balancing pass at the start of every tick (per-core only); Rebalance runs one pass now.
• Tasks are typed as work@core (work 1–30, preferred core 0–7), up to 12 tasks. Type, paste a list, or use a preset.
• The side panel shows task progress (Tasks) or queue lengths and imbalance (Load). Compare runs the three queue models to completion; click a row to load it.

Rules

Tick

Each core runs its queue head for 1 tick, then re-queues it (RR, q = 1).

Shared

One global queue; idle cores pull from it.

Per-core

Each core keeps its own persistent queue.

Balance

Busiest → idlest queue while lengths differ by more than 1.

Migration

Task runs on a different core than last time (counted once).

Soft

Prefer the last core; may migrate.

Hard

Only the preferred core; never moved.

Makespan

Ticks until every task has finished.

Config

Changing cores, model, affinity or tasks restarts at tick 0.

Legend

Keys