Re: spin_lock cause ?

[M Kelly <mark.kelly@lexisnexis.com>]

Rick Jones <rick.jones2 <at> hpe.com> writes:

> 
> On 06/01/2016 11:49 AM, M Kelly wrote:
> > So that is helpful.
> > I do not understand the system_call_fastpath->futex->_spin_lock part
> > yet,
> > but that is a separate issue!
> 
> Just a guess, but if your application is using threads, contention on 
> the synchonrization primitives you use to keep things straight can 
> end-up punting to the kernel.
> 
> rick jones
> 

Hi,

I think I recall you from hp, years ago when I was there :-)
Thanks for the info.  I suspect we are too heavy on 
pthread_mutex_lock()/unlock() 
which could probably produce these call graphs:

            --- _spin_lock
               |          
               |--49.11%-- futex_wake
               |          do_futex
               |          sys_futex
               |          system_call_fastpath
               |          __lll_unlock_wake
               |          
               |--44.31%-- futex_wait_setup
               |          futex_wait
               |          do_futex
               |          sys_futex
               |          system_call_fastpath
               |          __lll_lock_wait

thanks again,
mark