From: Peter Williams <pwil3058@bigpond.net.au>
To: Martin Bligh <mbligh@google.com>
Cc: Con Kolivas <kernel@kolivas.org>, Andrew Morton <akpm@osdl.org>,
linux-kernel@vger.kernel.org, Ingo Molnar <mingo@elte.hu>,
Andy Whitcroft <apw@shadowen.org>
Subject: Re: -mm seems significanty slower than mainline on kernbench
Date: Fri, 13 Jan 2006 09:20:38 +1100 [thread overview]
Message-ID: <43C6D636.8000105@bigpond.net.au> (raw)
In-Reply-To: <43C6B60E.2000003@bigpond.net.au>
Peter Williams wrote:
> Martin Bligh wrote:
>
>>
>>>>
>>>> But I was thinking more about the code that (in the original)
>>>> handled the case where the number of tasks to be moved was less than
>>>> 1 but more than 0 (i.e. the cases where "imbalance" would have been
>>>> reduced to zero when divided by SCHED_LOAD_SCALE). I think that I
>>>> got that part wrong and you can end up with a bias load to be moved
>>>> which is less than any of the bias_prio values for any queued tasks
>>>> (in circumstances where the original code would have rounded up to 1
>>>> and caused a move). I think that the way to handle this problem is
>>>> to replace 1 with "average bias prio" within that logic. This would
>>>> guarantee at least one task with a bias_prio small enough to be moved.
>>>>
>>>> I think that this analysis is a strong argument for my original
>>>> patch being the cause of the problem so I'll go ahead and generate a
>>>> fix. I'll try to have a patch available later this morning.
>>>
>>>
>>>
>>>
>>> Attached is a patch that addresses this problem. Unlike the
>>> description above it does not use "average bias prio" as that
>>> solution would be very complicated. Instead it makes the assumption
>>> that NICE_TO_BIAS_PRIO(0) is a "good enough" for this purpose as this
>>> is highly likely to be the median bias prio and the median is
>>> probably better for this purpose than the average.
>>>
>>> Signed-off-by: Peter Williams <pwil3058@bigpond.com.au>
>>
>>
>>
>> Doesn't fix the perf issue.
>
>
> OK, thanks. I think there's a few more places where SCHED_LOAD_SCALE
> needs to be multiplied by NICE_TO_BIAS_PRIO(0). Basically, anywhere
> that it's added to, subtracted from or compared to a load. In those
> cases it's being used as a scaled version of 1 and we need a scaled
This would have been better said as "the load generated by 1 task"
rather than just "a scaled version of 1". Numerically, they're the same
but one is clearer than the other and makes it more obvious why we need
NICE_TO_BIAS_PRIO(0) * SCHED_LOAD_SCALE and where we need it.
> version of NICE_TO_BIAS_PRIO(0). I'll have another patch later today.
I'm just testing this at the moment.
Peter
--
Peter Williams pwil3058@bigpond.net.au
"Learning, n. The kind of ignorance distinguishing the studious."
-- Ambrose Bierce
next prev parent reply other threads:[~2006-01-12 22:20 UTC|newest]
Thread overview: 59+ messages / expand[flat|nested] mbox.gz Atom feed top
2006-01-11 1:14 -mm seems significanty slower than mainline on kernbench Martin Bligh
2006-01-11 1:31 ` Andrew Morton
2006-01-11 1:41 ` Martin Bligh
2006-01-11 1:48 ` Andrew Morton
2006-01-11 1:49 ` Con Kolivas
2006-01-11 2:38 ` Peter Williams
2006-01-11 3:07 ` Con Kolivas
2006-01-11 3:12 ` Martin Bligh
2006-01-11 3:40 ` Peter Williams
2006-01-11 3:49 ` Con Kolivas
2006-01-11 4:33 ` Peter Williams
2006-01-11 5:14 ` Peter Williams
2006-01-11 6:21 ` Martin J. Bligh
2006-01-11 12:24 ` Peter Williams
2006-01-11 14:29 ` Con Kolivas
2006-01-11 22:05 ` Peter Williams
2006-01-12 0:54 ` Peter Williams
2006-01-12 1:18 ` Con Kolivas
2006-01-12 1:29 ` Peter Williams
2006-01-12 1:36 ` Con Kolivas
2006-01-12 2:23 ` Peter Williams
2006-01-12 2:26 ` Martin Bligh
2006-01-12 6:39 ` Con Kolivas
2006-01-23 19:28 ` Martin Bligh
2006-01-24 1:25 ` Peter Williams
2006-01-24 3:50 ` Peter Williams
2006-01-24 4:41 ` Martin J. Bligh
2006-01-24 6:22 ` Peter Williams
2006-01-24 6:42 ` Martin J. Bligh
2006-01-28 23:20 ` Peter Williams
2006-01-29 0:52 ` Martin J. Bligh
2006-01-12 2:27 ` Con Kolivas
2006-01-12 2:04 ` Martin Bligh
2006-01-12 6:35 ` Martin J. Bligh
2006-01-12 6:41 ` Con Kolivas
2006-01-12 6:54 ` Peter Williams
2006-01-12 18:39 ` Martin Bligh
2006-01-12 20:03 ` Peter Williams
2006-01-12 22:20 ` Peter Williams [this message]
2006-01-13 7:06 ` Peter Williams
2006-01-13 12:00 ` Peter Williams
2006-01-13 16:15 ` Martin J. Bligh
2006-01-13 16:26 ` Andy Whitcroft
2006-01-13 17:54 ` Andy Whitcroft
2006-01-13 20:41 ` Martin Bligh
2006-01-14 0:23 ` Peter Williams
2006-01-14 5:03 ` Nick Piggin
2006-01-14 5:40 ` Con Kolivas
2006-01-14 6:05 ` Nick Piggin
2006-01-14 5:53 ` Peter Williams
2006-01-14 6:13 ` Nick Piggin
2006-01-13 22:59 ` Peter Williams
2006-01-14 18:48 ` Martin J. Bligh
2006-01-15 0:05 ` Peter Williams
2006-01-15 2:04 ` Con Kolivas
2006-01-15 2:09 ` [PATCH] sched - remove unnecessary smpnice ifdefs Con Kolivas
2006-01-15 3:50 ` -mm seems significanty slower than mainline on kernbench Ingo Molnar
2006-01-12 1:25 ` Peter Williams
2006-01-11 1:52 ` Andrew Morton
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=43C6D636.8000105@bigpond.net.au \
--to=pwil3058@bigpond.net.au \
--cc=akpm@osdl.org \
--cc=apw@shadowen.org \
--cc=kernel@kolivas.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mbligh@google.com \
--cc=mingo@elte.hu \
/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).