Complete no click mode for Linux 3.10

In a typical Linux system, running CPUs are usually interrupted 100 to 1000 times per second by the clock. The clock interrupts to reconsider which process should run, respond to read-write update (RCU) calls, and handle other necessary transactions. This cyclical clock interrupt or click is much more reasonable than the notorious global kernel lock (BKL), but its disadvantage is that it affects performance and is not energy efficient.

A few years ago, the Linux community developed an optional no-tick (tickless) mode, and when the CPU was idle, the default shutdown clock was interrupted. It allows idle CPUs to be in deep sleep for a long time, reducing power consumption. It is important for a battery-powered system such as a laptop phone. However, this no-tick pattern also has drawbacks that are not conducive to high latency sensitive environments, because the CPU in wake sleep has a time overhead.

The Linux 3.10 Kernel introduces a nearly total no click (full tickless) mode, which is closed by default. According to the kernel developer's test, it can save 1% of CPU time. When full no-tick mode is enabled, the running CPU will turn off the clock if it has only one running process. If a second process occurs, the tick appears again for the scheduler to allocate resources based on the time period.

Technically speaking, this is not a total click, can only say approximation. High-performance and real-time computing will benefit from this new model, because it is a common strategy for a CPU to perform a task in both environments.