From: Jason Garrett-Glaser <darkshikari@gmail.com>
To: Avi Kivity <avi@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>, Mike Galbraith <efault@gmx.de>,
Andrea Arcangeli <aarcange@redhat.com>,
Linus Torvalds <torvalds@linux-foundation.org>,
Pekka Enberg <penberg@cs.helsinki.fi>,
Andrew Morton <akpm@linux-foundation.org>,
linux-mm@kvack.org, Marcelo Tosatti <mtosatti@redhat.com>,
Adam Litke <agl@us.ibm.com>, Izik Eidus <ieidus@redhat.com>,
Hugh Dickins <hugh.dickins@tiscali.co.uk>,
Nick Piggin <npiggin@suse.de>, Rik van Riel <riel@redhat.com>,
Mel Gorman <mel@csn.ul.ie>, Dave Hansen <dave@linux.vnet.ibm.com>,
Benjamin Herrenschmidt <benh@kernel.crashing.org>,
Mike Travis <travis@sgi.com>,
KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>,
Christoph Lameter <cl@linux-foundation.org>,
Chris Wright <chrisw@sous-sol.org>,
bpicco@redhat.com,
KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>,
Balbir Singh <balbir@linux.vnet.ibm.com>,
Arnd Bergmann <arnd@arndb.de>,
"Michael S. Tsirkin" <mst@redhat.com>,
Peter Zijlstra <peterz@infradead.org>,
Johannes Weiner <hannes@cmpxchg.org>,
Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Subject: Re: [PATCH 00 of 41] Transparent Hugepage Support #17
Date: Sat, 10 Apr 2010 13:58:21 -0700 [thread overview]
Message-ID: <m2u28f2fcbc1004101358q431be476xdd3e39221d6b7b04@mail.gmail.com> (raw)
In-Reply-To: <4BC0E556.30304@redhat.com>
On Sat, Apr 10, 2010 at 1:53 PM, Avi Kivity <avi@redhat.com> wrote:
> On 04/10/2010 11:49 PM, Jason Garrett-Glaser wrote:
>>
>>> 3-5% improvement. I had to tune khugepaged to scan more aggressively
>>> since
>>> the run is so short. The working set is only ~100MB here though.
>>>
>>
>> I'd try some longer runs with larger datasets to do more testing.
>>
>> Some things to try:
>>
>> 1) Pick a 1080p or even 2160p sequence from
>> http://media.xiph.org/video/derf/
>>
>>
>
> Ok, I'm downloading crown_run 2160p, but it will take a while.
You can always cheat by synthesizing a fake sample like this:
ffmpeg -i input.y4m -s 3840x2160 output.y4m
Or something similar.
Do be careful though; extremely fast presets combined with large input
samples will be disk-bottlenecked, so make sure to keep it small
enough to fit in disk cache and "prime" the cache before testing.
>> 2) Use --preset ultrafast or similar to do a ridiculously
>> memory-bandwidth-limited runthrough.
>>
>>
>
> Large pages improve random-access memory bandwidth but don't change
> sequential access. Which of these does --preset ultrafast change?
Hmm, I'm not quite sure. The process is strictly sequential, but
there is clearly enough random access mixed in to cause some sort of
change given your previous test. The main thing faster presets do is
decrease the amount of "work" done at each step, resulting in roughly
the same amount of memory bandwidth being required for each step--but
in a much shorter period of time. Most "work" done at each step stays
well within the L2 cache.
Jason
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
next prev parent reply other threads:[~2010-04-10 20:58 UTC|newest]
Thread overview: 205+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-04-02 0:41 [PATCH 00 of 41] Transparent Hugepage Support #17 Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 01 of 41] define MADV_HUGEPAGE Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 02 of 41] compound_lock Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 03 of 41] alter compound get_page/put_page Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 04 of 41] update futex compound knowledge Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 05 of 41] fix bad_page to show the real reason the page is bad Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 06 of 41] clear compound mapping Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 07 of 41] add native_set_pmd_at Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 08 of 41] add pmd paravirt ops Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 09 of 41] no paravirt version of pmd ops Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 10 of 41] export maybe_mkwrite Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 11 of 41] comment reminder in destroy_compound_page Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 12 of 41] config_transparent_hugepage Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 13 of 41] special pmd_trans_* functions Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 14 of 41] add pmd mangling generic functions Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 15 of 41] add pmd mangling functions to x86 Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 16 of 41] bail out gup_fast on splitting pmd Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 17 of 41] pte alloc trans splitting Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 18 of 41] add pmd mmu_notifier helpers Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 19 of 41] clear page compound Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 20 of 41] add pmd_huge_pte to mm_struct Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 21 of 41] split_huge_page_mm/vma Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 22 of 41] split_huge_page paging Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 23 of 41] clear_copy_huge_page Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 24 of 41] kvm mmu transparent hugepage support Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 25 of 41] _GFP_NO_KSWAPD Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 26 of 41] don't alloc harder for gfp nomemalloc even if nowait Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 27 of 41] transparent hugepage core Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 28 of 41] verify pmd_trans_huge isn't leaking Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 29 of 41] madvise(MADV_HUGEPAGE) Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 30 of 41] pmd_trans_huge migrate bugcheck Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 31 of 41] memcg compound Andrea Arcangeli
2010-04-02 0:41 ` [PATCH 32 of 41] memcg huge memory Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 33 of 41] transparent hugepage vmstat Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 34 of 41] khugepaged Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 35 of 41] skip transhuge pages in ksm for now Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 36 of 41] remove PG_buddy Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 37 of 41] add x86 32bit support Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 38 of 41] mincore transparent hugepage support Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 39 of 41] add pmd_modify Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 40 of 41] mprotect: pass vma down to page table walkers Andrea Arcangeli
2010-04-02 0:42 ` [PATCH 41 of 41] mprotect: transparent huge page support Andrea Arcangeli
2010-04-05 19:09 ` [PATCH 00 of 41] Transparent Hugepage Support #17 Andrew Morton
2010-04-05 19:36 ` Ingo Molnar
2010-04-05 20:26 ` Pekka Enberg
2010-04-05 20:32 ` Linus Torvalds
2010-04-05 20:46 ` Pekka Enberg
2010-04-05 20:58 ` Linus Torvalds
2010-04-05 21:54 ` Ingo Molnar
2010-04-05 23:21 ` Andrea Arcangeli
2010-04-06 0:26 ` Linus Torvalds
2010-04-06 1:08 ` [RFD] " Linus Torvalds
2010-04-06 1:26 ` Andrea Arcangeli
2010-04-06 1:35 ` Linus Torvalds
2010-04-06 1:13 ` Andrea Arcangeli
2010-04-06 1:38 ` Linus Torvalds
2010-04-06 2:23 ` Linus Torvalds
2010-04-06 5:25 ` Nick Piggin
2010-04-06 9:08 ` Ingo Molnar
2010-04-06 9:13 ` Ingo Molnar
2010-04-10 18:47 ` Andrea Arcangeli
2010-04-10 19:02 ` Ingo Molnar
2010-04-10 19:22 ` Avi Kivity
2010-04-10 19:47 ` Ingo Molnar
2010-04-10 20:00 ` Andrea Arcangeli
2010-04-10 20:10 ` Andrea Arcangeli
2010-04-10 20:21 ` Jason Garrett-Glaser
2010-04-10 20:24 ` Avi Kivity
2010-04-10 20:42 ` Avi Kivity
2010-04-10 20:47 ` Andrea Arcangeli
2010-04-10 21:00 ` Avi Kivity
2010-04-10 21:47 ` Andrea Arcangeli
2010-04-11 1:05 ` Andrea Arcangeli
2010-04-11 11:24 ` Ingo Molnar
2010-04-11 11:33 ` Avi Kivity
2010-04-11 12:11 ` Ingo Molnar
2010-04-25 19:27 ` Andrea Arcangeli
2010-04-26 18:01 ` Andrea Arcangeli
2010-04-30 9:55 ` Ingo Molnar
2010-04-30 15:19 ` Andrea Arcangeli
2010-05-02 12:17 ` Ingo Molnar
2010-04-10 20:49 ` Jason Garrett-Glaser
2010-04-10 20:53 ` Avi Kivity
2010-04-10 20:58 ` Jason Garrett-Glaser [this message]
2010-04-11 9:29 ` Avi Kivity
2010-04-11 9:37 ` Jason Garrett-Glaser
2010-04-11 9:40 ` Avi Kivity
2010-04-11 10:22 ` Jason Garrett-Glaser
2010-04-11 11:00 ` Ingo Molnar
2010-04-11 11:19 ` Avi Kivity
2010-04-11 11:30 ` Jason Garrett-Glaser
2010-04-11 11:52 ` hugepages will matter more in the future Ingo Molnar
2010-04-11 12:01 ` Avi Kivity
2010-04-11 12:35 ` Ingo Molnar
2010-04-11 15:22 ` Linus Torvalds
2010-04-11 15:43 ` Avi Kivity
2010-04-11 15:52 ` Linus Torvalds
2010-04-11 16:04 ` Avi Kivity
2010-04-12 7:45 ` Ingo Molnar
2010-04-12 8:14 ` Nick Piggin
2010-04-12 8:22 ` Ingo Molnar
2010-04-12 8:34 ` Nick Piggin
2010-04-12 8:47 ` Avi Kivity
2010-04-12 8:45 ` Andrea Arcangeli
2010-04-11 19:35 ` Andrea Arcangeli
2010-04-12 16:20 ` Rik van Riel
2010-04-12 16:40 ` Linus Torvalds
2010-04-12 16:56 ` Linus Torvalds
2010-04-12 17:06 ` Randy Dunlap
2010-04-12 17:36 ` Andrea Arcangeli
2010-04-12 17:46 ` Rik van Riel
2010-04-11 19:40 ` Andrea Arcangeli
2010-04-12 15:41 ` Linus Torvalds
2010-04-12 11:22 ` Arjan van de Ven
2010-04-12 11:29 ` Avi Kivity
2010-04-17 15:12 ` Arjan van de Ven
2010-04-17 18:18 ` Avi Kivity
2010-04-17 19:05 ` Arjan van de Ven
2010-04-17 19:05 ` Avi Kivity
2010-04-17 19:18 ` Arjan van de Ven
2010-04-17 19:20 ` Avi Kivity
2010-04-12 13:30 ` Andrea Arcangeli
2010-04-12 13:33 ` Avi Kivity
2010-04-12 13:39 ` Andrea Arcangeli
2010-04-12 13:53 ` Avi Kivity
2010-04-13 11:38 ` Ingo Molnar
2010-04-13 13:17 ` Andrea Arcangeli
2010-04-11 10:46 ` [PATCH 00 of 41] Transparent Hugepage Support #17 Ingo Molnar
2010-04-11 10:49 ` Ingo Molnar
2010-04-11 11:30 ` Avi Kivity
2010-04-11 12:08 ` Ingo Molnar
2010-04-11 12:24 ` Avi Kivity
2010-04-11 12:46 ` Ingo Molnar
2010-04-12 6:09 ` Nick Piggin
2010-04-12 6:18 ` Pekka Enberg
2010-04-12 6:48 ` Nick Piggin
2010-04-12 14:29 ` Christoph Lameter
2010-04-12 16:06 ` Nick Piggin
2010-04-12 6:36 ` Avi Kivity
2010-04-12 6:55 ` Ingo Molnar
2010-04-12 7:15 ` Nick Piggin
2010-04-12 7:45 ` Avi Kivity
2010-04-12 8:28 ` Nick Piggin
2010-04-12 9:01 ` Andrea Arcangeli
2010-04-12 9:03 ` Avi Kivity
2010-04-12 9:26 ` Nick Piggin
2010-04-12 9:39 ` Andrea Arcangeli
2010-04-12 10:02 ` Avi Kivity
2010-04-12 10:08 ` Andrea Arcangeli
2010-04-12 10:10 ` Avi Kivity
2010-04-12 10:23 ` Andrea Arcangeli
2010-04-12 10:37 ` Nick Piggin
2010-04-12 10:59 ` Avi Kivity
2010-04-12 12:23 ` Avi Kivity
2010-04-12 13:25 ` Andrea Arcangeli
2010-04-13 0:38 ` Andrew Morton
2010-04-13 6:18 ` Neil Brown
2010-04-13 13:31 ` Andrea Arcangeli
2010-04-13 13:40 ` Mel Gorman
2010-04-13 13:44 ` Andrea Arcangeli
2010-04-13 13:55 ` Mel Gorman
2010-04-13 14:03 ` Andrea Arcangeli
2010-04-12 7:51 ` Ingo Molnar
2010-04-12 7:18 ` Andrea Arcangeli
2010-04-12 6:49 ` Ingo Molnar
2010-04-12 7:35 ` Andrea Arcangeli
2010-04-12 7:08 ` Andrea Arcangeli
2010-04-12 7:21 ` Nick Piggin
2010-04-12 7:50 ` Avi Kivity
2010-04-12 8:07 ` Ingo Molnar
2010-04-12 8:21 ` Andrea Arcangeli
2010-04-12 10:27 ` Mel Gorman
2010-04-12 8:18 ` Andrea Arcangeli
2010-04-12 8:06 ` Andrea Arcangeli
2010-04-12 10:44 ` Mel Gorman
2010-04-12 11:12 ` Avi Kivity
2010-04-12 13:17 ` Andrea Arcangeli
2010-04-12 14:24 ` Christoph Lameter
2010-04-12 14:49 ` Avi Kivity
2010-04-06 9:55 ` Avi Kivity
2010-04-06 9:57 ` Avi Kivity
2010-04-06 11:55 ` Avi Kivity
2010-04-06 13:10 ` Nick Piggin
2010-04-06 13:22 ` Avi Kivity
2010-04-06 13:45 ` Nick Piggin
2010-04-06 13:57 ` Avi Kivity
2010-04-06 16:50 ` Andrea Arcangeli
2010-04-06 17:31 ` Avi Kivity
2010-04-06 18:00 ` Christoph Lameter
2010-04-06 18:04 ` Avi Kivity
2010-04-06 18:47 ` Avi Kivity
2010-04-06 14:44 ` Rik van Riel
2010-04-06 16:43 ` Andrea Arcangeli
2010-04-06 9:30 ` Mel Gorman
2010-04-06 10:32 ` Theodore Tso
2010-04-06 11:16 ` Mel Gorman
2010-04-06 13:13 ` Theodore Tso
2010-04-06 14:55 ` Mel Gorman
2010-04-06 16:46 ` Andrea Arcangeli
2010-04-05 21:01 ` Chris Mason
2010-04-05 21:18 ` Avi Kivity
2010-04-05 21:33 ` Linus Torvalds
2010-04-05 22:33 ` Chris Mason
2010-04-06 8:30 ` Mel Gorman
2010-04-06 11:35 ` Chris Mason
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=m2u28f2fcbc1004101358q431be476xdd3e39221d6b7b04@mail.gmail.com \
--to=darkshikari@gmail.com \
--cc=aarcange@redhat.com \
--cc=agl@us.ibm.com \
--cc=akpm@linux-foundation.org \
--cc=arnd@arndb.de \
--cc=avi@redhat.com \
--cc=balbir@linux.vnet.ibm.com \
--cc=benh@kernel.crashing.org \
--cc=bpicco@redhat.com \
--cc=chrisw@sous-sol.org \
--cc=cl@linux-foundation.org \
--cc=dave@linux.vnet.ibm.com \
--cc=efault@gmx.de \
--cc=hannes@cmpxchg.org \
--cc=hugh.dickins@tiscali.co.uk \
--cc=ieidus@redhat.com \
--cc=kamezawa.hiroyu@jp.fujitsu.com \
--cc=kosaki.motohiro@jp.fujitsu.com \
--cc=linux-mm@kvack.org \
--cc=mel@csn.ul.ie \
--cc=mingo@elte.hu \
--cc=mst@redhat.com \
--cc=mtosatti@redhat.com \
--cc=nishimura@mxp.nes.nec.co.jp \
--cc=npiggin@suse.de \
--cc=penberg@cs.helsinki.fi \
--cc=peterz@infradead.org \
--cc=riel@redhat.com \
--cc=torvalds@linux-foundation.org \
--cc=travis@sgi.com \
/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).