Subject: Re: [Patch 00/15] Reduce tracing payload size.
From: Steven Rostedt <rostedt@goodmis.org>
To: Andi Kleen <andi@firstfloor.org>
Cc: "Ted Ts'o" <tytso@mit.edu>, David Sharp <dhsharp@google.com>,
        linux-kernel@vger.kernel.org, mrubin@google.com
In-Reply-To: <20101206145852.GA24977@basil.fritz.box>
References: <1291421609-14665-1-git-send-email-dhsharp@google.com>
	 <878w03t4sn.fsf@basil.nowhere.org> <20101206135637.GC8135@thunk.org>
	 <20101206145852.GA24977@basil.fritz.box>
Content-Type: text/plain; charset="ISO-8859-15"
Date: Mon, 06 Dec 2010 11:17:35 -0500
Message-ID: <1291652255.16223.29.camel@gandalf.stny.rr.com>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1230
Lines: 30

On Mon, 2010-12-06 at 15:58 +0100, Andi Kleen wrote:
> On Mon, Dec 06, 2010 at 08:56:37AM -0500, Ted Ts'o wrote:
> > On Mon, Dec 06, 2010 at 02:22:00PM +0100, Andi Kleen wrote:
> > > David Sharp <dhsharp@google.com> writes:
> > > >
> > > > These patches are just a start to shrinking the size of trace events. I am
> > > > planning to also make small versions of trace events that are used when
> > > > CONFIG_SMALL_TRACES is enabled. I'm also open to discussion of other ways to
> > > > shrink event sizes.
> > > 
> > > Maybe the simplest would be to lzo them as they get logged?
> > > I assume you already considered that?
> > 
> > The tracing subsystem is supposed to be extremely low-overhead.
> > Compressiong the event log would add considerable CPU overhead.  If we
> 
> lzo as a rule of thumb is about as expensive as 3x memcpy()
> That's not really expensive.

Usually the most expensive thing in the trace is the memcpy(). Which is
why we try to do it only once.

-- Steve


--
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/