From: Dave Chinner <david@fromorbit.com>
To: "Darrick J. Wong" <djwong@kernel.org>
Cc: chandan.babu@oracle.com, chandanrlinux@gmail.com,
linux-xfs@vger.kernel.org
Subject: Re: [PATCH 11/14] xfs: dynamically allocate cursors based on maxlevels
Date: Tue, 21 Sep 2021 09:36:11 +1000 [thread overview]
Message-ID: <20210920233611.GN1756565@dread.disaster.area> (raw)
In-Reply-To: <20210920230635.GM1756565@dread.disaster.area>
On Tue, Sep 21, 2021 at 09:06:35AM +1000, Dave Chinner wrote:
> FWIW, an example of avoidable runtime calculation overhead of
> constants is xlog_calc_unit_res(). These values are actually
> constant for a given transaction reservation, but at 1.6 million
> transactions a second it shows up at #20 on the flat profile of
> functions using the most CPU:
>
> 0.71% [kernel] [k] xlog_calc_unit_res
>
> 0.71% of 32 CPUs for 1.6 million calculations a second of the same
> constants is a non-trivial amount of CPU time to spend doing
> unnecessary repeated calculations.
>
> Even though the btree cursor constant calculations are simpler than
> the log res calculations, they are more frequent. Hence on general
> principles of efficiency, I don't think we want to be replacing high
> frequency, low overhead slab/zone based allocations with heap
> allocations that require repeated constant calculations and
> size->slab redirection....
FWIW, I have another example that I don't have profiles for right now
because I didn't record them in the patch series that ends up
pre-calculating the AIL push target: xlog_grant_push_threshold().
This threshold is largely a fixed value ahead of the current log
tail (push at >75% of the physical log spacei consumed). We
do that calculation more often than we call xlog_calc_unit_res().
Because xlog_grant_push_threshold() accesses contended atomic
variables, it ends up consume 1-2% of total CPU time when
transactions rates reach the million/s ballpark.
I've currently replaced it with a fixed push threshold calculated at
mount time and let the AIL calculate the LSN of the push target
itself when it needs it. The result is a substantial reduction in
the CPU usage of the hot xfs_log_reserve() path, which also happens
to be the same hot path xlog_calc_unit_res() is called from...
Cheers,
Dave.
--
Dave Chinner
david@fromorbit.com
next prev parent reply other threads:[~2021-09-20 23:38 UTC|newest]
Thread overview: 48+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-18 1:29 [PATCHSET RFC chandan 00/14] xfs: support dynamic btree cursor height Darrick J. Wong
2021-09-18 1:29 ` [PATCH 01/14] xfs: remove xfs_btree_cur_t typedef Darrick J. Wong
2021-09-20 9:53 ` Chandan Babu R
2021-09-21 8:36 ` Christoph Hellwig
2021-09-18 1:29 ` [PATCH 02/14] xfs: don't allocate scrub contexts on the stack Darrick J. Wong
2021-09-20 9:53 ` Chandan Babu R
2021-09-20 17:39 ` Darrick J. Wong
2021-09-21 8:39 ` Christoph Hellwig
2021-09-18 1:29 ` [PATCH 03/14] xfs: dynamically allocate btree scrub context structure Darrick J. Wong
2021-09-20 9:53 ` Chandan Babu R
2021-09-21 8:43 ` Christoph Hellwig
2021-09-22 16:17 ` Darrick J. Wong
2021-09-18 1:29 ` [PATCH 04/14] xfs: stricter btree height checking when looking for errors Darrick J. Wong
2021-09-20 9:54 ` Chandan Babu R
2021-09-18 1:29 ` [PATCH 05/14] xfs: stricter btree height checking when scanning for btree roots Darrick J. Wong
2021-09-20 9:54 ` Chandan Babu R
2021-09-18 1:29 ` [PATCH 06/14] xfs: check that bc_nlevels never overflows Darrick J. Wong
2021-09-20 9:54 ` Chandan Babu R
2021-09-21 8:44 ` Christoph Hellwig
2021-09-18 1:29 ` [PATCH 07/14] xfs: support dynamic btree cursor heights Darrick J. Wong
2021-09-20 9:55 ` Chandan Babu R
2021-09-21 8:49 ` Christoph Hellwig
2021-09-18 1:29 ` [PATCH 08/14] xfs: refactor btree cursor allocation function Darrick J. Wong
2021-09-20 9:55 ` Chandan Babu R
2021-09-21 8:53 ` Christoph Hellwig
2021-09-18 1:29 ` [PATCH 09/14] xfs: fix maxlevels comparisons in the btree staging code Darrick J. Wong
2021-09-20 9:55 ` Chandan Babu R
2021-09-21 8:56 ` Christoph Hellwig
2021-09-22 15:59 ` Darrick J. Wong
2021-09-18 1:30 ` [PATCH 10/14] xfs: encode the max btree height in the cursor Darrick J. Wong
2021-09-20 9:55 ` Chandan Babu R
2021-09-21 8:57 ` Christoph Hellwig
2021-09-18 1:30 ` [PATCH 11/14] xfs: dynamically allocate cursors based on maxlevels Darrick J. Wong
2021-09-20 9:56 ` Chandan Babu R
2021-09-20 23:06 ` Dave Chinner
2021-09-20 23:36 ` Dave Chinner [this message]
2021-09-21 9:03 ` Christoph Hellwig
2021-09-22 18:55 ` Darrick J. Wong
2021-09-22 17:38 ` Darrick J. Wong
2021-09-22 23:10 ` Dave Chinner
2021-09-23 1:58 ` Darrick J. Wong
2021-09-23 5:56 ` Chandan Babu R
2021-09-18 1:30 ` [PATCH 12/14] xfs: compute actual maximum btree height for critical reservation calculation Darrick J. Wong
2021-09-20 9:56 ` Chandan Babu R
2021-09-18 1:30 ` [PATCH 13/14] xfs: compute the maximum height of the rmap btree when reflink enabled Darrick J. Wong
2021-09-20 9:56 ` Chandan Babu R
2021-09-18 1:30 ` [PATCH 14/14] xfs: kill XFS_BTREE_MAXLEVELS Darrick J. Wong
2021-09-20 9:57 ` Chandan Babu R
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=20210920233611.GN1756565@dread.disaster.area \
--to=david@fromorbit.com \
--cc=chandan.babu@oracle.com \
--cc=chandanrlinux@gmail.com \
--cc=djwong@kernel.org \
--cc=linux-xfs@vger.kernel.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).