A task scheduling method, apparatus, and computing system where computing units are divided into multiple partitions, each with a scheduler and multiple task queues of different priorities based on task criticality. The task scheduling method determines partition criticality. If inter-partition scheduling is needed, the second partition scheduler moves a task from a first task queue in the first partition to its own queue. The first partition scheduler moves a task from a second task queue in the second partition to its queue, where the two task queues have different priorities.