How does the kernel scheduler know a timeslice has passed?

Question

In my previous question How does the kernel scheduler know how to pre-empt a process? I was given an answer to how pre-emption occurs.

Now I am wondering, how does the kernel scheduler know that a timeslice has passed? I read up on the hardware timer solution which makes sense to me, but then I read that most current operating systems (e.g. Windows, Linux, etc.) do not use hardware timers, but rather software timers.

How can software timers be used to pre-empt a process once it has taken up its timeslice (e.g. it did not pre-empt itself.) It seems like some hardware timer would be necessary?

Answer 1

It seems like some hardware timer would be necessary?

Yes, the kernel relies on hardware to generate an interrupt at regular intervals. On PCs, this was historically the 8253/8254 programmable interval timer, or an emulation thereof, then the local APIC timer, then the HPET.

Current Linux kernels can be built to run “tickless” when possible: the kernel will program timers to only fire when necessary, and if a given CPU is running a single process, that may well be “never”. In most cases, dynamic ticks are used, so the kernel sets timers up to fire at varying intervals depending on its requirements — fewer interrupts means fewer wake-ups, which means idle CPUs can be kept in low-power modes for longer periods, which saves energy.