Problem

Thread A must wait for a thread on another processor to change a flag bit in memory. The scheduler can either allow A to spin, repeatedly retesting the flag, or it can deschedule A, allowing some other thread to run. Suppose it takes a total of 10 milliseconds for the operating system to switch a processor from one thread to another. If the operating system deschedules thread A and immediately reschedules it, then it wastes 20 milliseconds. If, instead, A starts spinning at time t₀, and the flag changes at t₁, then the operating system will have wasted t₁ t₀ time doing unproductive work.

A prescient scheduler is one that can predict the future. If it foresees that the flag will change in less than 20 milliseconds, it makes sense to have A spin, wasting less than 20 milliseconds, because descheduling and rescheduling A wastes 20 milliseconds. If, on the other hand, it takes more than 20 milliseconds for the flag to change, it makes sense to replace A with another thread, wasting no more than 20 milliseconds.

Your assignment is to implement a scheduler that never wastes more than twice the time a prescient scheduler would have wasted under the same circumstances.