That is, if the main processor most of the time expects data from RAM or L3 cache with cache skipping, but the system is in real time (priority of the stream in real time) and the stream is connected (affinity) to the kernel and works without switching the stream / context, what load (use) of the CPU-Core should be displayed on modern x86_64 ?
That is, processor usage is displayed as a decrease only when registering in standby mode?
And if anyone knows if in this case the behavior is different from other processors: ARM, Power [PC], Sparc?
Explanation : shows CPU usage in the standard task manager on Windows
The hardware thread (the logical core) that stopped by skipping the cache cannot do anything else, so it is still considered busy for task managers / CPU time accounting / process time scheduler / like this.
This is true for all architectures.
Without a hyper-thread, the “hardware thread” / “logical core” coincides with the “physical core”.
Morphcore / other on-the-fly, changing between a hyper-thread and a more powerful single core, can make a difference between a thread that makes many executable blocks busy, and a thread that is locked in the cache skips a lot of time.
. , , ., .
: /. GUI . , /, (- , ) . .
, ISR, . .
, , 2 20., 20 , (1000-20)/1000 = 98%.
, - . , , .
sysadmin, , , , . Sysadmins , , ( preempting) , . .