From: Bruce Richardson <bruce.richardson@intel.com>
To: Stephen Hemminger <stephen@networkplumber.org>
Cc: Tyler Retzlaff <roretzla@linux.microsoft.com>,
Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com>,
"dev@dpdk.org" <dev@dpdk.org>, nd <nd@arm.com>
Subject: Re: [RFC] rte_ring: don't use always inline
Date: Fri, 6 May 2022 17:38:10 +0100 [thread overview]
Message-ID: <YnVO8nqi2hIcEBv/@bricha3-MOBL.ger.corp.intel.com> (raw)
In-Reply-To: <20220506084112.5bcc3000@hermes.local>
On Fri, May 06, 2022 at 08:41:12AM -0700, Stephen Hemminger wrote:
> On Fri, 6 May 2022 00:24:34 -0700
> Tyler Retzlaff <roretzla@linux.microsoft.com> wrote:
>
> > On Thu, May 05, 2022 at 10:59:32PM +0000, Honnappa Nagarahalli wrote:
> > > Thanks Stephen. Do you see any performance difference with this change?
> >
> > as a matter of due diligence i think a comparison should be made just
> > to be confident nothing is regressing.
> >
> > i support this change in principal since it is generally accepted best
> > practice to not force inlining since it can remove more valuable
> > optimizations that the compiler may make that the human can't see.
> > the optimizations may vary depending on compiler implementation.
> >
> > force inlining should be used as a targeted measure rather than blanket
> > on every function and when in use probably needs to be periodically
> > reviewed and potentially removed as the code / compiler evolves.
> >
> > also one other consideration is the impact of a particular compiler's
> > force inlining intrinsic/builtin is that it may permit inlining of
> > functions when not declared in a header. i.e. a function from one
> > library may be able to be inlined to another binary as a link time
> > optimization. although everything here is in a header so it's a bit
> > moot.
> >
> > i'd like to see this change go in if possible.
> >
> > thanks
> >
>
>
> Some quick numbers from Gcc 10.3 and 2.7G AMD and ring_perf_autotest
>
> Looks like always inline is faster on second run but just inline is
> slightly faster on first run. Maybe the icache gets loaded for second run,
> but on first pass the smaller code size helps.
>
Interesting, I haven't observed that effect.
The main trouble with the unit tests is that there are so many possible
numbers to compare. We probably need to focus on a few to make sense of it
all. Running on an "Intel(R) Xeon(R) Gold 6330N CPU @ 2.20GHz" with Ubuntu
20.04 (GCC 9.4), I scanned through the numbers looking for signicant
percentage differences. This one caught my eye due to the %age difference:
Basline value:
sudo ./build-baseline/app/test/dpdk-test -- ring_perf_autotest | grep "SP/SC: single:"
...
legacy APIs: SP/SC: single: 9.08
elem APIs: element size 16B: SP/SC: single: 11.13
With patch:
sudo ./build/app/test/dpdk-test -- ring_perf_autotest | grep "SP/SC: single:"
...
legacy APIs: SP/SC: single: 15.81
elem APIs: element size 16B: SP/SC: single: 21.14
So the SP/SC element enqueue cost went from 9-11 cycles to 15-21 cycles.
Percentage-wise, this seems a lot, though in absolute terms it may not be.
Therefore, I think we'll need a decent amount of varied testing before
taking this patch.
/Bruce
prev parent reply other threads:[~2022-05-06 16:38 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-05-05 22:45 Stephen Hemminger
2022-05-05 22:59 ` Honnappa Nagarahalli
2022-05-05 23:10 ` Stephen Hemminger
2022-05-05 23:16 ` Stephen Hemminger
2022-05-06 1:37 ` Honnappa Nagarahalli
2022-05-06 7:24 ` Tyler Retzlaff
2022-05-06 15:12 ` Honnappa Nagarahalli
2022-05-06 15:28 ` Bruce Richardson
2022-05-06 16:33 ` Stephen Hemminger
2022-05-06 16:39 ` Bruce Richardson
2022-05-06 17:48 ` Konstantin Ananyev
2022-05-06 15:41 ` Stephen Hemminger
2022-05-06 16:38 ` Bruce Richardson [this message]
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=YnVO8nqi2hIcEBv/@bricha3-MOBL.ger.corp.intel.com \
--to=bruce.richardson@intel.com \
--cc=Honnappa.Nagarahalli@arm.com \
--cc=dev@dpdk.org \
--cc=nd@arm.com \
--cc=roretzla@linux.microsoft.com \
--cc=stephen@networkplumber.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 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).