From: Chris Metcalf <cmetcalf@mellanox.com>
To: Peter Zijlstra <peterz@infradead.org>,
Thomas Gleixner <tglx@linutronix.de>
Cc: LKML <linux-kernel@vger.kernel.org>,
John Stultz <john.stultz@linaro.org>,
Ingo Molnar <mingo@kernel.org>,
David Gibson <david@gibson.dropbear.id.au>,
Liav Rehana <liavr@mellanox.com>,
Richard Cochran <richardcochran@gmail.com>,
Parit Bhargava <prarit@redhat.com>,
Laurent Vivier <lvivier@redhat.com>,
"Christopher S. Hall" <christopher.s.hall@intel.com>
Subject: Re: [patch 5/6] [RFD] timekeeping: Provide optional 128bit math
Date: Fri, 9 Dec 2016 12:32:07 -0500 [thread overview]
Message-ID: <486a28a2-b7de-67fd-f731-1487b141319b@mellanox.com> (raw)
In-Reply-To: <20161209083011.GD15765@worktop.programming.kicks-ass.net>
On 12/9/2016 3:30 AM, Peter Zijlstra wrote:
> On Fri, Dec 09, 2016 at 07:38:47AM +0100, Peter Zijlstra wrote:
>> On Fri, Dec 09, 2016 at 06:26:38AM +0100, Peter Zijlstra wrote:
>>> Just for giggles, on tilegx the branch is actually slower than doing the
>>> mult unconditionally.
>>>
>>> The problem is that the two multiplies would otherwise completely
>>> pipeline, whereas with the conditional you serialize them.
>> On my Haswell laptop the unconditional version is faster too.
> Only when using x86_64 instructions, once I fixed the i386 variant it
> was slower, probably due to register pressure and the like.
>
>>> (came to light while talking about why the mul_u64_u32_shr() fallback
>>> didn't work right for them, which was a combination of the above issue
>>> and the fact that their compiler 'lost' the fact that these are
>>> 32x32->64 mults and did 64x64 ones instead).
>> Turns out using GCC-6.2.1 we have the same problem on i386, GCC doesn't
>> recognise the 32x32 mults and generates crap.
>>
>> This used to work :/
> Do we want something like so?
>
> ---
> arch/tile/include/asm/Kbuild | 1 -
> arch/tile/include/asm/div64.h | 14 ++++++++++++++
> arch/x86/include/asm/div64.h | 10 ++++++++++
> include/linux/math64.h | 26 ++++++++++++++++++--------
> 4 files changed, 42 insertions(+), 9 deletions(-)
Untested, but I looked at it closely, and it seems like a decent idea.
Acked-by: Chris Metcalf <cmetcalf@mellanox.com> [for tile]
Of course if this is pushed up, it will then probably be too tempting for me not
to add the tilegx-specific mul_u64_u32_shr() to take advantage of pipelining
the two 32x32->64 multiplies :-)
--
Chris Metcalf, Mellanox Technologies
http://www.mellanox.com
next prev parent reply other threads:[~2016-12-09 17:32 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-12-08 20:49 [patch 0/6] timekeeping: Cure the signed/unsigned wreckage Thomas Gleixner
2016-12-08 20:49 ` [patch 1/6] timekeeping: Force unsigned clocksource to nanoseconds conversion Thomas Gleixner
2016-12-08 23:38 ` David Gibson
2016-12-09 11:13 ` [tip:timers/core] timekeeping_Force_unsigned_clocksource_to_nanoseconds_conversion tip-bot for Thomas Gleixner
2016-12-08 20:49 ` [patch 2/6] timekeeping: Make the conversion call chain consistently unsigned Thomas Gleixner
2016-12-08 23:39 ` David Gibson
2016-12-09 11:13 ` [tip:timers/core] " tip-bot for Thomas Gleixner
2016-12-08 20:49 ` [patch 3/6] timekeeping: Get rid of pointless typecasts Thomas Gleixner
2016-12-08 23:40 ` David Gibson
2016-12-09 11:14 ` [tip:timers/core] " tip-bot for Thomas Gleixner
2016-12-08 20:49 ` [patch 4/6] timekeeping: Use mul_u64_u32_shr() instead of open coding it Thomas Gleixner
2016-12-08 23:41 ` David Gibson
2016-12-09 11:14 ` [tip:timers/core] " tip-bot for Thomas Gleixner
2016-12-08 20:49 ` [patch 5/6] [RFD] timekeeping: Provide optional 128bit math Thomas Gleixner
2016-12-09 4:08 ` Ingo Molnar
2016-12-09 4:29 ` Ingo Molnar
2016-12-09 4:39 ` John Stultz
2016-12-09 4:48 ` Peter Zijlstra
2016-12-09 5:22 ` Ingo Molnar
2016-12-09 5:41 ` Peter Zijlstra
2016-12-09 5:11 ` Peter Zijlstra
2016-12-09 6:08 ` Peter Zijlstra
2016-12-09 5:26 ` Peter Zijlstra
2016-12-09 6:38 ` Peter Zijlstra
2016-12-09 8:30 ` Peter Zijlstra
2016-12-09 9:11 ` Peter Zijlstra
2016-12-09 10:01 ` Peter Zijlstra
2016-12-09 17:32 ` Chris Metcalf [this message]
2017-01-14 12:51 ` [tip:timers/core] math64, timers: Fix 32bit mul_u64_u32_shr() and friends tip-bot for Peter Zijlstra
2016-12-09 10:18 ` [patch 5/6] [RFD] timekeeping: Provide optional 128bit math Peter Zijlstra
2016-12-09 17:20 ` Chris Metcalf
2016-12-08 20:49 ` [patch 6/6] [RFD] timekeeping: Get rid of cycle_t Thomas Gleixner
2016-12-08 23:43 ` David Gibson
2016-12-09 4:52 ` [patch 0/6] timekeeping: Cure the signed/unsigned wreckage John Stultz
2016-12-09 5:30 ` Peter Zijlstra
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=486a28a2-b7de-67fd-f731-1487b141319b@mellanox.com \
--to=cmetcalf@mellanox.com \
--cc=christopher.s.hall@intel.com \
--cc=david@gibson.dropbear.id.au \
--cc=john.stultz@linaro.org \
--cc=liavr@mellanox.com \
--cc=linux-kernel@vger.kernel.org \
--cc=lvivier@redhat.com \
--cc=mingo@kernel.org \
--cc=peterz@infradead.org \
--cc=prarit@redhat.com \
--cc=richardcochran@gmail.com \
--cc=tglx@linutronix.de \
/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).