Re: [PATCH v8 01/11] cpuidle/poll_state: poll via smp_cond_load_relaxed()

[Ankur Arora <ankur.a.arora@oracle.com>]

Christoph Lameter (Ampere) <cl@gentwo.org> writes:

> On Wed, 16 Oct 2024, Ankur Arora wrote:
>
>> > The other core will wake our core up by sending an IPI. The IPI will
>> > invoke a scheduler function on our core and the WFE will continue.
>>
>> Why? The target core is not sleeping. It is *polling* on a memory
>> address (on arm64, via LDXR; WFE). Ergo an IPI is not needed to tell
>> it that a need-resched bit is set.
>
> The IPI is sent to interrupt the process that is not sleeping. This is
> done so the busy processor can reschedule the currently running process
> and respond to the event.

The scheduler treats idle specially (if the architecture defines
TIF_POLLING_NRFLAG). There's also the sched_wake_idle_without_ipi
tracepoint for this path.

$ sudo perf stat -e sched:sched_wake_idle_without_ipi perf bench sched pipe
# Running 'sched/pipe' benchmark:
# Executed 1000000 pipe operations between two processes

     Total time: 5.173 [sec]

       5.173613 usecs/op
         193288 ops/sec

 Performance counter stats for 'perf bench sched pipe':

         1,992,368      sched:sched_wake_idle_without_ipi

       5.178976487 seconds time elapsed

       0.396076000 seconds user
       6.999566000 seconds sys

--
ankur