From: Dave Jones <davej@redhat.com>
To: Neil Horman <nhorman@tuxdriver.com>
Cc: linux-kernel@vger.kernel.org, sebastien.dugue@bull.net,
Thomas Gleixner <tglx@linutronix.de>,
Ingo Molnar <mingo@redhat.com>, "H. Peter Anvin" <hpa@zytor.com>,
x86@kernel.org
Subject: Re: [PATCH v2 2/2] x86: add prefetching to do_csum
Date: Wed, 6 Nov 2013 10:34:29 -0500 [thread overview]
Message-ID: <20131106153429.GA26336@redhat.com> (raw)
In-Reply-To: <1383751399-10298-3-git-send-email-nhorman@tuxdriver.com>
On Wed, Nov 06, 2013 at 10:23:19AM -0500, Neil Horman wrote:
> do_csum was identified via perf recently as a hot spot when doing
> receive on ip over infiniband workloads. After alot of testing and
> ideas, we found the best optimization available to us currently is to
> prefetch the entire data buffer prior to doing the checksum
>
> diff --git a/arch/x86/lib/csum-partial_64.c b/arch/x86/lib/csum-partial_64.c
> index 9845371..9f2d3ee 100644
> --- a/arch/x86/lib/csum-partial_64.c
> +++ b/arch/x86/lib/csum-partial_64.c
> @@ -29,8 +29,15 @@ static inline unsigned short from32to16(unsigned a)
> * Things tried and found to not make it faster:
> * Manual Prefetching
> * Unrolling to an 128 bytes inner loop.
> - * Using interleaving with more registers to break the carry chains.
Did you mean perhaps to remove the "Manual Prefetching" line instead ?
(Curious, what was tried before that made it not worthwhile?)
Dave
next prev parent reply other threads:[~2013-11-06 15:34 UTC|newest]
Thread overview: 105+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-10-11 16:51 [PATCH] x86: Run checksumming in parallel accross multiple alu's Neil Horman
2013-10-12 17:21 ` Ingo Molnar
2013-10-13 12:53 ` Neil Horman
2013-10-14 20:28 ` Neil Horman
2013-10-14 21:19 ` Eric Dumazet
2013-10-14 22:18 ` Eric Dumazet
2013-10-14 22:37 ` Joe Perches
2013-10-14 22:44 ` Eric Dumazet
2013-10-14 22:49 ` Joe Perches
2013-10-15 7:41 ` Ingo Molnar
2013-10-15 10:51 ` Borislav Petkov
2013-10-15 12:04 ` Ingo Molnar
2013-10-15 16:21 ` Joe Perches
2013-10-16 0:34 ` Eric Dumazet
2013-10-16 6:25 ` Ingo Molnar
2013-10-16 16:55 ` Joe Perches
2013-10-17 0:34 ` Neil Horman
2013-10-17 1:42 ` Eric Dumazet
2013-10-18 16:50 ` Neil Horman
2013-10-18 17:20 ` Eric Dumazet
2013-10-18 20:11 ` Neil Horman
2013-10-18 21:15 ` Eric Dumazet
2013-10-20 21:29 ` Neil Horman
2013-10-21 17:31 ` Eric Dumazet
2013-10-21 17:46 ` Neil Horman
2013-10-21 19:21 ` Neil Horman
2013-10-21 19:44 ` Eric Dumazet
2013-10-21 20:19 ` Neil Horman
2013-10-26 12:01 ` Ingo Molnar
2013-10-26 13:58 ` Neil Horman
2013-10-27 7:26 ` Ingo Molnar
2013-10-27 17:05 ` Neil Horman
2013-10-17 8:41 ` Ingo Molnar
2013-10-17 18:19 ` H. Peter Anvin
2013-10-17 18:48 ` Eric Dumazet
2013-10-18 6:43 ` Ingo Molnar
2013-10-28 16:01 ` Neil Horman
2013-10-28 16:20 ` Ingo Molnar
2013-10-28 17:49 ` Neil Horman
2013-10-28 16:24 ` Ingo Molnar
2013-10-28 16:49 ` David Ahern
2013-10-28 17:46 ` Neil Horman
2013-10-28 18:29 ` Neil Horman
2013-10-29 8:25 ` Ingo Molnar
2013-10-29 11:20 ` Neil Horman
2013-10-29 11:30 ` Ingo Molnar
2013-10-29 11:49 ` Neil Horman
2013-10-29 12:52 ` Ingo Molnar
2013-10-29 13:07 ` Neil Horman
2013-10-29 13:11 ` Ingo Molnar
2013-10-29 13:20 ` Neil Horman
2013-10-29 14:17 ` Neil Horman
2013-10-29 14:27 ` Ingo Molnar
2013-10-29 20:26 ` Neil Horman
2013-10-31 10:22 ` Ingo Molnar
2013-10-31 14:33 ` Neil Horman
2013-11-01 9:13 ` Ingo Molnar
2013-11-01 14:06 ` Neil Horman
2013-10-29 14:12 ` David Ahern
2013-10-15 7:32 ` Ingo Molnar
2013-10-15 13:14 ` Neil Horman
2013-10-12 22:29 ` H. Peter Anvin
2013-10-13 12:53 ` Neil Horman
2013-10-18 16:42 ` Neil Horman
2013-10-18 17:09 ` H. Peter Anvin
2013-10-25 13:06 ` Neil Horman
2013-10-14 4:38 ` Andi Kleen
2013-10-14 7:49 ` Ingo Molnar
2013-10-14 21:07 ` Eric Dumazet
2013-10-15 13:17 ` Neil Horman
2013-10-14 20:25 ` Neil Horman
2013-10-15 7:12 ` Sébastien Dugué
2013-10-15 13:33 ` Andi Kleen
2013-10-15 13:56 ` Sébastien Dugué
2013-10-15 14:06 ` Eric Dumazet
2013-10-15 14:15 ` Sébastien Dugué
2013-10-15 14:26 ` Eric Dumazet
2013-10-15 14:52 ` Eric Dumazet
2013-10-15 16:02 ` Andi Kleen
2013-10-16 0:28 ` Eric Dumazet
2013-11-06 15:23 ` x86: Enhance perf checksum profiling and x86 implementation Neil Horman
2013-11-06 15:23 ` [PATCH v2 1/2] perf: Add csum benchmark tests to perf Neil Horman
2013-11-06 15:23 ` [PATCH v2 2/2] x86: add prefetching to do_csum Neil Horman
2013-11-06 15:34 ` Dave Jones [this message]
2013-11-06 15:54 ` Neil Horman
2013-11-06 17:19 ` Joe Perches
2013-11-06 18:11 ` Neil Horman
2013-11-06 20:02 ` Neil Horman
2013-11-06 20:07 ` Joe Perches
2013-11-08 16:25 ` Neil Horman
2013-11-08 16:51 ` Joe Perches
2013-11-08 19:07 ` Neil Horman
2013-11-08 19:17 ` Joe Perches
2013-11-08 20:08 ` Neil Horman
2013-11-08 19:17 ` H. Peter Anvin
2013-11-08 19:01 ` Neil Horman
2013-11-08 19:33 ` Joe Perches
2013-11-08 20:14 ` Neil Horman
2013-11-08 20:29 ` Joe Perches
2013-11-11 19:40 ` Neil Horman
2013-11-11 21:18 ` Ingo Molnar
2013-11-06 18:23 ` Eric Dumazet
2013-11-06 18:59 ` Neil Horman
2013-11-06 20:19 ` Andi Kleen
2013-11-07 21:23 ` Neil Horman
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=20131106153429.GA26336@redhat.com \
--to=davej@redhat.com \
--cc=hpa@zytor.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=nhorman@tuxdriver.com \
--cc=sebastien.dugue@bull.net \
--cc=tglx@linutronix.de \
--cc=x86@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.