From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from down.free-electrons.com ([37.187.137.238]
 helo=mail.free-electrons.com)
 by bombadil.infradead.org with esmtp (Exim 4.85_2 #1 (Red Hat Linux))
 id 1bK0LL-0003cM-0X
 for linux-mtd@lists.infradead.org; Mon, 04 Jul 2016 09:39:07 +0000
Date: Mon, 4 Jul 2016 11:38:35 +0200
From: Boris Brezillon <boris.brezillon@free-electrons.com>
To: Artem Bityutskiy <dedekind1@gmail.com>
Cc: Brian Norris <computersforpeace@gmail.com>, Sascha Hauer
 <s.hauer@pengutronix.de>, linux-mtd@lists.infradead.org, Richard Weinberger
 <richard@nod.at>
Subject: Re: [PATCH v2] UBI: only read necessary size when reading the VID
 header
Message-ID: <20160704113835.7853d77d@bbrezillon>
In-Reply-To: <1467172188.2456.64.camel@gmail.com>
References: <1467114667-30548-1-git-send-email-s.hauer@pengutronix.de>
 <1467118829.2456.40.camel@gmail.com>
 <20160628174304.GA80724@google.com>
 <1467172188.2456.64.camel@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
List-Id: Linux MTD discussion mailing list <linux-mtd.lists.infradead.org>
List-Unsubscribe: <http://lists.infradead.org/mailman/options/linux-mtd>,
 <mailto:linux-mtd-request@lists.infradead.org?subject=unsubscribe>
List-Archive: <http://lists.infradead.org/pipermail/linux-mtd/>
List-Post: <mailto:linux-mtd@lists.infradead.org>
List-Help: <mailto:linux-mtd-request@lists.infradead.org?subject=help>
List-Subscribe: <http://lists.infradead.org/mailman/listinfo/linux-mtd>,
 <mailto:linux-mtd-request@lists.infradead.org?subject=subscribe>

On Wed, 29 Jun 2016 06:49:48 +0300
Artem Bityutskiy <dedekind1@gmail.com> wrote:

> On Tue, 2016-06-28 at 10:43 -0700, Brian Norris wrote:
> > Hi Artem,
> > 
> > I'll comment on the other branches of this thread, but one thing
> > here:
> > 
> > On Tue, Jun 28, 2016 at 04:00:29PM +0300, Artem Bityutskiy wrote:  
> > > Therefore, unless I misunderstood this patch - it introduces a
> > > regression to those old setups at least (forces MTD to use an
> > > intermediate buffer rather than copy data from NAND directly to the
> > > buffer supplied by UBI)  
> > 
> > It's really a balance between speed of the flash and speed of the
> > memcpy().  
> 
> Sure.
> 
> >  I believe Boris may have benchmarked some of this recently,
> > but I'm really inclined to believe that reading several times as much
> > as
> > you need from flash is much worse than doing some extra memcpy().  
> 
> That's probably true.
> 
> >  So
> > even if we introduce an extra memcpy(), it might still be worth it to
> > save the extra wait-for-flash time.  
> 
> Right.
> 
> > Intuitively, I expect that these days, the I/O time is much more
> > significant than any memcpy().  
> 
> All good points. Besides indeed in case of the subpage the memcpy() is
> present anyway for (for unexpected reasons).
> 
> So yeah, I think the concern I rose is a non-issue and we could proceed
> with Sascha's patch. Thanks!

I see I don't have to convince you with real numbers, but, as pointed
by Brian, memcpy() is indeed way faster than NAND I/Os (and ECC
correction steps).