UNIX for Advanced & Expert Users

Hi Experts,

If a thread running on CPU needs some data from RAM (from near or far RAM) it requests the data (this moment of time is t0) and gets it in moment of time in t1, and continues his work.

Between t0 and t1, while data is transferring from RAM, the thread is goes out from CPU? Is the time between t0 and t1 included or excluded from CPU time?

Thanks

This seems a bit vague, but in general cpu time should be based on the threads state, and only accummlates when it is in a compute/runnable state. This is reflected in the r column on vmstat. I am not clear on what you mean by near of far ram, perhaps you mean cache versus a dma operation. Still, compute time should be based on state and it might vary across different architectures.

blackrageous, you answers in correct direction, thanks.
by near and far memory I mean numa. We can skip this, let's say just RAM.

cpu time is based on state, yes, and this is the question - what is exactly the state a thread is having in the moment while date is transferring from memory to cache. Is the thread goes from cpu and rescheduled into run queue or still be on cpu and running?
Let's imagine that a thread worked 1 hour without any blocked operations like I/O, just calculating with RAM, the thread is only working thread in idle multi-core system, and a half of time was consumed by transferring data from memory to cache. In such case cpu time of the thread is 30 minutes or 60 ?