linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Andy Lutomirski <luto@amacapital.net>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: George Spelvin <linux@sciencehorizons.net>,
	"Jason A. Donenfeld" <Jason@zx2c4.com>,
	Andi Kleen <ak@linux.intel.com>,
	David Miller <davem@davemloft.net>,
	David Laight <David.Laight@aculab.com>,
	"Daniel J . Bernstein" <djb@cr.yp.to>,
	Eric Biggers <ebiggers3@gmail.com>,
	Eric Dumazet <eric.dumazet@gmail.com>,
	Hannes Frederic Sowa <hannes@stressinduktion.org>,
	Jean-Philippe Aumasson <jeanphilippe.aumasson@gmail.com>,
	"kernel-hardening@lists.openwall.com" 
	<kernel-hardening@lists.openwall.com>,
	Linux Crypto Mailing List <linux-crypto@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Network Development <netdev@vger.kernel.org>,
	Tom Herbert <tom@herbertland.com>,
	"Theodore Ts'o" <tytso@mit.edu>,
	Vegard Nossum <vegard.nossum@gmail.com>
Subject: Re: HalfSipHash Acceptable Usage
Date: Wed, 21 Dec 2016 17:54:26 -0800	[thread overview]
Message-ID: <CALCETrVxFDQGJeQX5k39pM3TvqH4q10SduPY=Os_RiJGEg_0Hg@mail.gmail.com> (raw)
In-Reply-To: <CA+55aFy8fNOxw3bnwkX1S46jKnW6i26mueaiuOsScyN3kFJp+A@mail.gmail.com>

On Wed, Dec 21, 2016 at 9:25 AM, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
> On Wed, Dec 21, 2016 at 7:55 AM, George Spelvin
> <linux@sciencehorizons.net> wrote:
>>
>> How much does kernel_fpu_begin()/kernel_fpu_end() cost?
>
> It's now better than it used to be, but it's absolutely disastrous
> still. We're talking easily many hundreds of cycles. Under some loads,
> thousands.
>
> And I warn you already: it will _benchmark_ a hell of a lot better
> than it will work in reality. In benchmarks, you'll hit all the
> optimizations ("oh, I've already saved away all the FP registers, no
> need to do it again").
>
> In contrast, in reality, especially with things like "do it once or
> twice per incoming packet", you'll easily hit the absolute worst
> cases, where not only does it take a few hundred cycles to save the FP
> state, you'll then return to user space in between packets, which
> triggers the slow-path return code and reloads the FP state, which is
> another few hundred cycles plus.

Hah, you're thinking that the x86 code works the way that Rik and I
want it to work, and you just made my day. :)  What actually happens
is that the state is saved in kernel_fpu_begin() and restored in
kernel_fpu_end(), and it'll take a few hundred cycles best case.  If
you do it a bunch of times in a loop, you *might* trigger a CPU
optimization that notices that the state being saved is the same state
that was just restored, but you're still going to pay the full restore
code each round trip no matter what.

The code is much clearer in 4.10 kernels now that I deleted the unused
"lazy" branches.

>
> Similarly, in benchmarks you'll hit the "modern CPU's power on the AVX
> unit and keep it powered up for a while afterwards", while in real
> life you would quite easily hit the "oh, AVX is powered down because
> we were idle, now it powers up at half speed which is another latency
> hit _and_ the AVX unit won't run full out anyway".

I *think* that was mostly fixed in Broadwell or thereabouts (in terms
of latency -- throughput and power consumption still suffers).

  parent reply	other threads:[~2016-12-22  2:01 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2016-12-19 17:32 Jason A. Donenfeld
     [not found] ` <CAGiyFdduUNSGq24zfsk0ZU=hnOCmewAw8vw6XvDoS-3f+3UPKQ@mail.gmail.com>
2016-12-19 21:00   ` Jason A. Donenfeld
2016-12-20 21:36 ` Theodore Ts'o
2016-12-20 23:07   ` George Spelvin
2016-12-20 23:55   ` Eric Dumazet
2016-12-21  3:28     ` George Spelvin
2016-12-21  5:29       ` Eric Dumazet
2016-12-21  6:34         ` George Spelvin
2016-12-21 14:24           ` Jason A. Donenfeld
2016-12-21 15:55             ` George Spelvin
2016-12-21 16:37               ` Jason A. Donenfeld
2016-12-21 16:41               ` [kernel-hardening] " Rik van Riel
2016-12-21 17:25               ` Linus Torvalds
2016-12-21 18:07                 ` George Spelvin
2016-12-22  1:54                 ` Andy Lutomirski [this message]
2016-12-21 14:42         ` Jason A. Donenfeld
2016-12-21 15:56           ` Eric Dumazet
2016-12-21 16:33             ` Jason A. Donenfeld
2016-12-21 16:39             ` [kernel-hardening] " Rik van Riel
2016-12-21 17:08               ` Eric Dumazet
2016-12-21 18:37             ` George Spelvin
2016-12-21 18:40               ` Jason A. Donenfeld
2016-12-21 22:27               ` Theodore Ts'o
2016-12-22  0:18                 ` George Spelvin
2016-12-22  1:13                 ` George Spelvin

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='CALCETrVxFDQGJeQX5k39pM3TvqH4q10SduPY=Os_RiJGEg_0Hg@mail.gmail.com' \
    --to=luto@amacapital.net \
    --cc=David.Laight@aculab.com \
    --cc=Jason@zx2c4.com \
    --cc=ak@linux.intel.com \
    --cc=davem@davemloft.net \
    --cc=djb@cr.yp.to \
    --cc=ebiggers3@gmail.com \
    --cc=eric.dumazet@gmail.com \
    --cc=hannes@stressinduktion.org \
    --cc=jeanphilippe.aumasson@gmail.com \
    --cc=kernel-hardening@lists.openwall.com \
    --cc=linux-crypto@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux@sciencehorizons.net \
    --cc=netdev@vger.kernel.org \
    --cc=tom@herbertland.com \
    --cc=torvalds@linux-foundation.org \
    --cc=tytso@mit.edu \
    --cc=vegard.nossum@gmail.com \
    --subject='Re: HalfSipHash Acceptable Usage' \
    /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

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).