From: Al Viro <viro@zeniv.linux.org.uk>
To: David Laight <David.Laight@aculab.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"linux-arch@vger.kernel.org" <linux-arch@vger.kernel.org>
Subject: Re: [PATCH 04/18] csum_and_copy_..._user(): pass 0xffffffff instead of 0 as initial sum
Date: Thu, 23 Jul 2020 16:21:01 +0100 [thread overview]
Message-ID: <20200723152101.GI2786714@ZenIV.linux.org.uk> (raw)
In-Reply-To: <20200723145342.GH2786714@ZenIV.linux.org.uk>
On Thu, Jul 23, 2020 at 03:53:42PM +0100, Al Viro wrote:
> Said that, what you've printed for 1-byte segments (and that's going to be
> seriously affected by the setup costs in csum-copy.S, sensitive to calling
> convention changes) is time to run the 16-iteration loop divided by 1 * 16 / 8;
> IOW, your difference for 16 iterations here is 37*2 = 74 cycles. With
> per-iteration diff being a bit under 5 cycles. Which is not implausible,
> but
> 1) extrapolating to other compiler versions, flags, etc. is not obvious
> 2) the effects of calling convention changes need to be taken into account
> 3) for copying to/from userland the effects of calling convention changes
> are be even larger, and kernel is certainly not going to issue kvec iters of _that_
> sort, TYVM.
To clarify it a bit: the effects of calling conventions change are mostly due
to not passing (and saving) those error pointers, and that could be had with
"pass the initial sum in" - just start these iov_iter.c loops with sum = ~0U
and we get the same warranties re not getting 0 in absence of faults.
The point is, your "~4.5 cycles per vector" is pretty much noise and the
difference between the 3-argument and 4-argument variants could easily be
in the same range. It might be a valid microoptimization, it might be not.
3-argument variant is simpler and IMO in absence of strong data we ought
to go with that.
next prev parent reply other threads:[~2020-07-23 15:21 UTC|newest]
Thread overview: 102+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-07-21 20:24 [RFC][CFT][PATCHSET] saner calling conventions for csum-and-copy primitives Al Viro
2020-07-21 20:25 ` [PATCH 01/18] skb_copy_and_csum_bits(): don't bother with the last argument Al Viro
2020-07-21 20:25 ` [PATCH 02/18] icmp_push_reply(): reorder adding the checksum up Al Viro
2020-07-21 20:25 ` [PATCH 03/18] csum_partial_copy_nocheck(): drop the last argument Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 04/18] csum_and_copy_..._user(): pass 0xffffffff instead of 0 as initial sum Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:55 ` Linus Torvalds
2020-07-21 20:58 ` Linus Torvalds
2020-07-21 21:11 ` Al Viro
2020-07-21 21:16 ` Linus Torvalds
2020-07-21 21:16 ` Linus Torvalds
2020-07-25 17:54 ` Al Viro
2020-07-22 9:45 ` David Laight
2020-07-22 9:27 ` David Laight
2020-07-22 14:42 ` Al Viro
2020-07-22 15:22 ` David Laight
2020-07-22 15:54 ` Al Viro
2020-07-22 16:17 ` David Laight
2020-07-22 17:39 ` Al Viro
2020-07-23 8:29 ` David Laight
2020-07-23 13:54 ` David Laight
2020-07-23 14:30 ` David Laight
2020-07-23 14:53 ` Al Viro
2020-07-23 15:19 ` David Laight
2020-07-23 15:21 ` Al Viro [this message]
2020-07-23 15:36 ` David Laight
2020-07-21 20:25 ` [PATCH 05/18] saner calling conventions for csum_and_copy_..._user() Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 06/18] alpha: propagate the calling convention changes down to csum_partial_copy.c helpers Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 07/18] arm: propagate the calling convention changes down to csum_partial_copy_from_user() Al Viro
2020-07-21 20:25 ` [PATCH 08/18] m68k: get rid of zeroing destination on error in csum_and_copy_from_user() Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 09/18] sh: propage the calling conventions change down to csum_partial_copy_generic() Al Viro
2020-07-21 20:25 ` [PATCH 10/18] i386: propagate " Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 11/18] sparc32: propagate the calling conventions change down to __csum_partial_copy_sparc_generic() Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-22 1:20 ` David Miller
2020-07-21 20:25 ` [PATCH 12/18] mips: csum_and_copy_{to,from}_user() are never called under KERNEL_DS Al Viro
2020-07-21 20:25 ` [PATCH 13/18] mips: __csum_partial_copy_kernel() has no users left Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 14/18] mips: propagate the calling convention change down into __csum_partial_copy_..._user() Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 15/18] xtensa: propagate the calling conventions change down into csum_partial_copy_generic() Al Viro
2020-07-22 8:56 ` Max Filippov
2020-07-21 20:25 ` [PATCH 16/18] sparc64: propagate the calling convention changes down to __csum_partial_copy_...() Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-22 1:21 ` David Miller
2020-07-21 20:25 ` [PATCH 17/18] amd64: switch csum_partial_copy_generic() to new calling conventions Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-21 20:25 ` [PATCH 18/18] ppc: propagate the calling conventions change down to csum_partial_copy_generic() Al Viro
2020-07-21 20:25 ` Al Viro
2020-07-24 1:25 ` [RFC][CFT][PATCHSET v2] saner calling conventions for csum-and-copy primitives Al Viro
2020-07-24 1:25 ` [PATCH v2 01/20] xtensa: fix access check in csum_and_copy_from_user Al Viro
2020-07-24 1:25 ` [PATCH v2 02/20] skb_copy_and_csum_bits(): don't bother with the last argument Al Viro
2020-07-24 1:25 ` [PATCH v2 03/20] icmp_push_reply(): reorder adding the checksum up Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 04/20] unify generic instances of csum_partial_copy_nocheck() Al Viro
2020-07-24 6:41 ` Christoph Hellwig
2020-07-24 12:19 ` Al Viro
2020-07-24 12:23 ` Christoph Hellwig
2020-07-24 12:30 ` Al Viro
2020-07-26 7:11 ` Christoph Hellwig
2020-07-26 7:11 ` Christoph Hellwig
2020-07-27 3:58 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 05/20] csum_partial_copy_nocheck(): drop the last argument Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 06/20] csum_and_copy_..._user(): pass 0xffffffff instead of 0 as initial sum Al Viro
2020-07-24 1:25 ` [PATCH v2 07/20] saner calling conventions for csum_and_copy_..._user() Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 08/20] alpha: propagate the calling convention changes down to csum_partial_copy.c helpers Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 09/20] arm: propagate the calling convention changes down to csum_partial_copy_from_user() Al Viro
2020-07-24 1:25 ` [PATCH v2 10/20] m68k: get rid of zeroing destination on error in csum_and_copy_from_user() Al Viro
2020-07-24 1:25 ` [PATCH v2 11/20] sh: propage the calling conventions change down to csum_partial_copy_generic() Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 12/20] i386: propagate " Al Viro
2020-07-24 1:25 ` [PATCH v2 13/20] sparc32: propagate the calling conventions change down to __csum_partial_copy_sparc_generic() Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 14/20] mips: csum_and_copy_{to,from}_user() are never called under KERNEL_DS Al Viro
2020-07-24 1:25 ` [PATCH v2 15/20] mips: __csum_partial_copy_kernel() has no users left Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 16/20] mips: propagate the calling convention change down into __csum_partial_copy_..._user() Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 17/20] xtensa: propagate the calling conventions change down into csum_partial_copy_generic() Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 18/20] sparc64: propagate the calling convention changes down to __csum_partial_copy_...() Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 19/20] amd64: switch csum_partial_copy_generic() to new calling conventions Al Viro
2020-07-24 1:25 ` Al Viro
2020-07-24 1:25 ` [PATCH v2 20/20] ppc: propagate the calling conventions change down to csum_partial_copy_generic() Al Viro
2020-07-24 1:25 ` Al Viro
2020-10-14 22:26 ` Jason A. Donenfeld
2020-10-14 22:51 ` Linus Torvalds
2020-10-14 22:53 ` Linus Torvalds
2020-10-14 22:54 ` Jason A. Donenfeld
2020-10-14 22:53 ` Jason A. Donenfeld
2020-10-14 23:12 ` Al Viro
2020-10-14 23:02 ` [PATCH] powerpc32: don't adjust unmoved stack pointer in csum_partial_copy_generic() epilogue Jason A. Donenfeld
2020-10-14 23:05 ` Linus Torvalds
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=20200723152101.GI2786714@ZenIV.linux.org.uk \
--to=viro@zeniv.linux.org.uk \
--cc=David.Laight@aculab.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=torvalds@linux-foundation.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).