Re: RFC: Ideal Adaptive Spinning Conditions

[Darren Hart <dvhltc@us.ibm.com>]

Steven Rostedt wrote:
> On Wed, 2010-03-31 at 19:13 -0700, Darren Hart wrote:
>> Steven Rostedt wrote:
>>> On Wed, 2010-03-31 at 16:21 -0700, Darren Hart wrote:
>>>
>>>> o What type of lock hold times do we expect to benefit?
>>> 0 (that's a zero) :-p
>>>
>>> I haven't seen your patches but you are not doing a heuristic approach,
>>> are you? That is, do not "spin" hoping the lock will suddenly become
>>> free. I was against that for -rt and I would be against that for futex
>>> too.
>> I'm not sure what you're getting at here. Adaptive spinning is indeed 
>> hoping the lock will become free while you are spinning and checking 
>> it's owner...
> 
> I'm talking about the original idea people had of "lets spin for 50us
> and hope it is unlocked before then", which I thought was not a good
> idea.
> 
> 
>>>> o How much contention is a good match for adaptive spinning?
>>>>    - this is related to the number of threads to run in the test
>>>> o How many spinners should be allowed?
>>>>
>>>> I can share the kernel patches if people are interested, but they are 
>>>> really early, and I'm not sure they are of much value until I better 
>>>> understand the conditions where this is expected to be useful.
>>> Again, I don't know how you implemented your adaptive spinners, but the
>>> trick to it in -rt was that it would only spin while the owner of the
>>> lock was actually running. If it was not running, it would sleep. No
>>> point waiting for a sleeping task to release its lock.
>> It does exactly this.
> 
> OK, that's good.
> 
>>> Is this what you did? Because, IIRC, this only benefited spinlocks
>>> converted to mutexes. It did not help with semaphores, because
>>> semaphores could be held for a long time. Thus, it was good for short
>>> held locks, but hurt performance on long held locks.
>> Trouble is, I'm still seeing performance penalties even on the shortest 
>> critical section possible (lock();unlock();)
> 
> performance penalties compared to what? not having adaptive at all?

Right. See the data in the original mail:


          futex_lock: Result: 635 Kiter/s
futex_lock_adaptive: Result: 542 Kiter/s

So 15% fewer lock/unlock iterations per second with in kernel adaptive 
spinning enabled for a critical section approaching 0 in length. But If 
we agree I'm taking the right approach, then it's time for me to polish 
things up a bit and send them out for review.

-- 
Darren Hart
IBM Linux Technology Center
Real-Time Linux Team