Date: Wed, 1 Jul 2009 15:50:34 +0200
From: Andi Kleen <andi@firstfloor.org>
To: Ingo Molnar <mingo@elte.hu>
Cc: Hitoshi Mitake <mitake@dcl.info.waseda.ac.jp>, andi@firstfloor.org,
       linux-kernel@vger.kernel.org
Subject: Re: [PATCH][RFC] Adding information of counts processes acquired how many spinlocks to schedstat II
Message-ID: <20090701135034.GU6760@one.firstfloor.org>
References: <20090701.152115.706994265076015808.mitake@dcl.info.waseda.ac.jp> <87hbxwj1k3.fsf@basil.nowhere.org> <20090701.174226.419764642024067218.mitake@dcl.info.waseda.ac.jp> <20090701090749.GA13535@elte.hu>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20090701090749.GA13535@elte.hu>
User-Agent: Mutt/1.4.2.2i
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1208
Lines: 26

On Wed, Jul 01, 2009 at 11:07:49AM +0200, Ingo Molnar wrote:
> His arguments are bogus: both lockstat and perfcounters are optional 
> (and default off), and the sw counter can be made near zero cost 
> even if both perfcounters and lockstat is enabled. Also, sw counters 
> are generally per CPU, etc. so not a performance issue.

BTW I looked through the code generated by this again.  As far as I 
can see it still uses a global flag word to enable/disable the counter.

The standard problem -- discussed a lot -- with that set up used to be that 
if the kernel is entered cache cold then you end up with a potential full cache
miss (200-1000 cycles) just to access that global variable. The old LTT
had a neat optimization to patch in immediate values to avoid this problem.

Has this been considered for these counters? With such a change
it might be indeed a "small" cost even in the worst cases.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/