From: Arnd Bergmann <arnd@arndb.de>
To: Geoff Levand <geoffrey.levand@am.sony.com>
Cc: 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
Subject: Re: [PATCH 21/21]: powerpc/cell spidernet DMA coalescing
Date: Wed, 11 Oct 2006 11:25:09 +0200 [thread overview]
Message-ID: <200610111125.10902.arnd@arndb.de> (raw)
In-Reply-To: <452C4CE0.5010607@am.sony.com>
On Wednesday 11 October 2006 03:46, Geoff Levand wrote:
> >=20
> > The others look good, but this one complicates the code and doesn't hav=
e any benefit. =C2=A020=20
> > for 21 isn't bad.
>=20
> Is the motivation for this change to improve performance by reducing the =
overhead
> of the mapping calls? =C2=A0If so, there may be some benefit for some sys=
tems. =C2=A0Could
> you please elaborate?
>=20
=46rom 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:
=2D the card can do better prefetching of consecutive memory
=2D there are more I/O ptes available for other drivers.
Arnd <><
WARNING: multiple messages have this Message-ID (diff)
From: Arnd Bergmann <arnd@arndb.de>
To: Geoff Levand <geoffrey.levand@am.sony.com>
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
Subject: Re: [PATCH 21/21]: powerpc/cell spidernet DMA coalescing
Date: Wed, 11 Oct 2006 11:25:09 +0200 [thread overview]
Message-ID: <200610111125.10902.arnd@arndb.de> (raw)
In-Reply-To: <452C4CE0.5010607@am.sony.com>
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 <><
WARNING: multiple messages have this Message-ID (diff)
From: Arnd Bergmann <arnd@arndb.de>
To: Geoff Levand <geoffrey.levand@am.sony.com>
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
Subject: Re: [PATCH 21/21]: powerpc/cell spidernet DMA coalescing
Date: Wed, 11 Oct 2006 11:25:09 +0200 [thread overview]
Message-ID: <200610111125.10902.arnd@arndb.de> (raw)
In-Reply-To: <452C4CE0.5010607@am.sony.com>
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 <><
next prev parent reply other threads:[~2006-10-11 9:25 UTC|newest]
Thread overview: 46+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-10-10 20:49 [PATCH 0/21]: powerpc/cell spidernet bugfixes, etc Linas Vepstas
2006-10-10 20:49 ` Linas Vepstas
2006-10-10 20:56 ` [PATCH 1/21]: powerpc/cell spidernet ethtool -i version number info Linas Vepstas
2006-10-10 20:57 ` [PATCH 2/21]: powerpc/cell spidernet burst alignment patch Linas Vepstas
2006-10-10 20:59 ` [PATCH 3/21]: Spidernet module parm permissions Linas Vepstas
2006-10-10 21:00 ` [PATCH 4/21]: powerpc/cell spidernet force-end fix Linas Vepstas
2006-10-10 21:00 ` [PATCH 0/21]: powerpc/cell spidernet bugfixes, etc Andrew Morton
2006-10-10 21:00 ` Andrew Morton
2006-10-10 21:01 ` [PATCH 5/21]: powerpc/cell spidernet zlen min packet length Linas Vepstas
2006-10-10 21:01 ` [PATCH 6/21]: powerpc/cell spidernet add missing netdev watchdog Linas Vepstas
2006-10-10 21:02 ` [PATCH 7/21]: Spidernet fix register field definitions Linas Vepstas
2006-10-10 21:04 ` [PATCH 8/21]: Spidernet stop queue when queue is full Linas Vepstas
2006-10-10 21:05 ` [PATCH 9/21]: powerpc/cell spidernet bogus rx interrupt bit Linas Vepstas
2006-10-10 21:06 ` [PATCH 10/21]: powerpc/cell spidernet fix error interrupt print Linas Vepstas
2006-10-10 21:08 ` [PATCH 11/21]: powerpc/cell spidernet stop error printing patch Linas Vepstas
2006-10-10 21:09 ` [PATCH 12/21]: powerpc/cell spidernet incorrect offset Linas Vepstas
2006-10-10 21:11 ` [PATCH 13/21]: powerpc/cell spidernet low watermark patch Linas Vepstas
2006-10-10 21:13 ` [PATCH 14/21]: powerpc/cell spidernet NAPI polling info Linas Vepstas
2006-10-10 21:14 ` [PATCH 15/21]: powerpc/cell spidernet refine locking Linas Vepstas
2006-10-10 21:15 ` [PATCH 16/21]: powerpc/cell spidernet Linas Vepstas
2006-10-10 21:18 ` [PATCH 17/21]: powerpc/cell spidernet reduce DMA kicking Linas Vepstas
2006-10-10 21:19 ` [PATCH 18/21]: powerpc/cell spidernet variable name change Linas Vepstas
2006-10-10 21:21 ` [PATCH 19/21]: powerpc/cell spidernet DMA direction fix Linas Vepstas
2006-10-10 21:22 ` [PATCH 20/21]: powerpc/cell spidernet release all descrs Linas Vepstas
2006-10-10 21:23 ` [PATCH 21/21]: powerpc/cell spidernet DMA coalescing Linas Vepstas
2006-10-10 23:20 ` jschopp
2006-10-11 1:46 ` Geoff Levand
2006-10-11 1:46 ` Geoff Levand
2006-10-11 9:25 ` Arnd Bergmann [this message]
2006-10-11 9:25 ` Arnd Bergmann
2006-10-11 9:25 ` Arnd Bergmann
2006-10-11 15:20 ` Linas Vepstas
2006-10-11 15:20 ` Linas Vepstas
2006-10-11 15:47 ` Geoff Levand
2006-10-11 15:47 ` Geoff Levand
2006-10-11 22:13 ` Benjamin Herrenschmidt
2006-10-11 22:13 ` Benjamin Herrenschmidt
2006-10-11 7:15 ` Benjamin Herrenschmidt
2006-10-11 7:15 ` Benjamin Herrenschmidt
2006-10-14 2:12 ` Joel Schopp
2006-10-11 16:02 ` [PATCH 0/21]: powerpc/cell spidernet bugfixes, etc Arnd Bergmann
2006-10-11 16:02 ` Arnd Bergmann
2006-10-11 16:42 ` James K Lewis
2006-10-11 16:42 ` James K Lewis
2006-10-11 21:06 ` James K Lewis
2006-10-11 21:06 ` James K Lewis
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200610111125.10902.arnd@arndb.de \
--to=arnd@arndb.de \
--cc=akpm@osdl.org \
--cc=geoffrey.levand@am.sony.com \
--cc=jeff@garzik.org \
--cc=jklewis@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linuxppc-dev@ozlabs.org \
--cc=netdev@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.