Re: [RFC] perf_events: support for uncore a.k.a. nest units

[Corey Ashford <cjashfor@linux.vnet.ibm.com>]

On 03/31/2010 07:01 AM, Peter Zijlstra wrote:
> On Tue, 2010-03-30 at 15:12 -0700, Corey Ashford wrote:
>>
>>> Initially I'd not allow per-pmu-per-task contexts
>>> because then things like perf_event_task_sched_out() would get rather
>>> complex.
>>
>> Definitely.  I don't think it makes sense to have per-task context on
>> nest/uncore PMUs.  At least we haven't found any justification for it.
>
> For uncore no, but there is also the hw-breakpoint stuff that is being
> presented as a pmu, for those it would make sense to have a separate
> per-task context.
>
> But doing multiple per-task contexts is something for a next step
> indeed.
>
>>> For RR we can move away from perf_event_task_tick and let the pmu
>>> install a (hr)timer for this on their own.
>>
>> This is necessary I think, because of the access time for some of the PMU's.  I
>> wonder though if it should, perhaps optionally, be off-loaded to a high-priority
>> task to do the switching so that access latency to the PMU can be controlled.
>>
>> As I mentioned when we met, some of the Wire-Speed processor nest PMU control
>> registers are accessed via SCOM, which is an internal, 200 MHz serial bus.  We
>> are being quoted ~525 SCOM bus ticks to do a PMU control register access, which
>> comes out to about 2.5 microseconds.  If you figure 5 accesses to rotate the
>> events on a PMU, that's a minimum of 12.5 microseconds.
>
> Yeah, you mentioned that.. for those things we need some changes anyway,
> since currently we install per-cpu counters using IPIs and expect the
> pmu::enable() method to be synchronous (it has a return value).

That's a good point.  We hadn't considered this issue.

> It would
> be totally unacceptable to do 2.5ms pokes with IRQs disabled.

Just to be clear, it's 2.5us, not 2.5ms, but I think it's still bad... 
in our case, it's about 6600 processor clocks per access.

> The RR thing would be the easiest to solve, just let the timer wake up a
> thread instead of doing the work itself, that's fully isolated to how
> the pmu chooses to implement that. The above mentioned issue however
> would be much more challenging to fix nicely.

It seems like it might need to be done in two phases.  IPI request is 
sent, and then a thread is woken up on the other CPU, it does some work, 
and then sets a status variable and somehow notifies the caller that the 
operation has completed.  I don't know the kernel's communication 
mechanisms well enough to know which one is most appropriate - maybe rwsem?

- Corey