From: Arnd Bergmann <arnd@arndb.de>
To: Geoff Levand <geoffrey.levand@am.sony.com>
Subject: Re: [PATCH 21/21]: powerpc/cell spidernet DMA coalescing
Date: Wed, 11 Oct 2006 11:25:09 +0200
User-Agent: KMail/1.9.4
Cc: jschopp <jschopp@austin.ibm.com>, Linas Vepstas <linas@austin.ibm.com>,
       akpm@osdl.org, jeff@garzik.org, netdev@vger.kernel.org,
       James K Lewis <jklewis@us.ibm.com>, linux-kernel@vger.kernel.org,
       linuxppc-dev@ozlabs.org
References: <20061010204946.GW4381@austin.ibm.com> <452C2AAA.5070001@austin.ibm.com> <452C4CE0.5010607@am.sony.com>
In-Reply-To: <452C4CE0.5010607@am.sony.com>
MIME-Version: 1.0
Content-Type: text/plain;
  charset="utf-8"
Content-Transfer-Encoding: 8BIT
Content-Disposition: inline
Message-Id: <200610111125.10902.arnd@arndb.de>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1052
Lines: 25

On Wednesday 11 October 2006 03:46, Geoff Levand wrote:
> > 
> > The others look good, but this one complicates the code and doesn't have any benefit.  20 
> > for 21 isn't bad.
> 
> Is the motivation for this change to improve performance by reducing the overhead
> of the mapping calls?  If so, there may be some benefit for some systems.  Could
> you please elaborate?
> 

>From what I understand, this patch drastically reduces the number of
I/O PTEs that are needed in the iommu. With the current static IOMMU
mapping, it should only make a difference during initialization, but
any platform that uses a dynamic mapping of iommu entries will benefit
a lot from it, because:

- the card can do better prefetching of consecutive memory
- there are more I/O ptes available for other drivers.

	Arnd <><
-
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/