From: Anton Blanchard <anton@samba.org>
To: "Feldman, Scott" <scott.feldman@intel.com>
Cc: David Gibson <dwg@au1.ibm.com>,
linux-kernel@vger.kernel.org,
Nancy Milliner <milliner@us.ibm.com>,
Herman Dierks <hdierks@us.ibm.com>,
Ricardo Gonzalez <ricardoz@us.ibm.com>
Subject: Re: e1000 performance hack for ppc64 (Power4)
Date: Sat, 14 Jun 2003 09:15:01 +1000 [thread overview]
Message-ID: <20030613231501.GC32097@krispykreme> (raw)
In-Reply-To: <C6F5CF431189FA4CBAEC9E7DD5441E010107D932@orsmsx402.jf.intel.com>
> 2-byte alignment is bad for ppc64, so what is minimum alignment that is
> good? (I hope it's not 2K!) What could you do at a higher level to
> present properly aligned buffers to the driver?
Best case 128 bytes - ppc64 cacheline size. Worst case 512 bytes - our
pci-pci bridge likes to prefetch in 512 byte increments. From Herman's
data you can see this in action:
Linux sending one 1514 byte packet (777 Mbits sec rate)
Address PCI command Bytes xfered
420A8 MRM 344
42200 MRM 512
42400 MRM 512
42600 MRL 128
42680 MR 18
With the buffer 512 byte aligned this is completed in 3 transactions.
Yes the hardware could be more intelligent about these unaligned
transactions but we cant do much about that now.
We might be able to do the alignment at a higher level but its not
straightforward (see my previous mail).
> If I'm understanding the patch correctly, you're saying unaligned DMA +
> TCE lookup is more expensive than a data copy? If we copy the data, we
> loss some of the benefits of TSO and Zerocopy and h/w checksum
> offloading! What is more expensive, unaligned DMA or TCE?
Lets ignore TCE lookup overhead for the moment. As Herman pointed all
these DMAs should occur on the same page.
Anton
next prev parent reply other threads:[~2003-06-13 23:01 UTC|newest]
Thread overview: 34+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-06-13 1:16 e1000 performance hack for ppc64 (Power4) Feldman, Scott
2003-06-13 23:15 ` Anton Blanchard [this message]
-- strict thread matches above, loose matches on Subject: below --
2003-06-16 18:56 Feldman, Scott
2003-06-16 18:21 Feldman, Scott
2003-06-16 18:30 ` Dave Hansen
2003-06-15 14:40 Herman Dierks
2003-06-15 14:44 ` David S. Miller
2003-06-16 16:17 ` Nivedita Singhvi
2003-06-15 14:32 Herman Dierks
2003-06-13 23:52 Feldman, Scott
2003-06-13 23:52 ` David S. Miller
2003-06-14 0:55 ` Anton Blanchard
2003-06-14 1:34 ` David S. Miller
2003-06-14 0:03 ` Anton Blanchard
2003-06-13 22:13 Feldman, Scott
2003-06-13 17:03 Herman Dierks
2003-06-13 15:17 Herman Dierks
2003-06-13 16:21 ` Dave Hansen
2003-06-13 22:38 ` Anton Blanchard
2003-06-13 22:46 ` David S. Miller
2003-06-13 23:18 ` Anton Blanchard
2003-06-14 1:52 ` Lincoln Dale
2003-06-14 5:41 ` David S. Miller
2003-06-14 5:52 ` Lincoln Dale
2003-06-14 6:08 ` David S. Miller
2003-06-14 6:14 ` David S. Miller
2003-06-14 6:27 ` William Lee Irwin III
2003-06-14 17:08 ` Greg KH
2003-06-14 17:19 ` Greg KH
2003-06-14 17:21 ` Riley Williams
2003-06-15 3:01 ` David S. Miller
2003-06-14 5:16 ` Nivedita Singhvi
2003-06-14 5:36 ` David S. Miller
2003-06-12 3:32 David Gibson
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=20030613231501.GC32097@krispykreme \
--to=anton@samba.org \
--cc=dwg@au1.ibm.com \
--cc=hdierks@us.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=milliner@us.ibm.com \
--cc=ricardoz@us.ibm.com \
--cc=scott.feldman@intel.com \
/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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).