From: Michael Veksler <VEKSLER@il.ibm.com>
To: Gabriel Dos Reis <gdr@integrable-solutions.net>
Cc: Dale Johannesen <dalej@apple.com>,
Daniel Berlin <dberlin@dberlin.org>,
gcc@gcc.gnu.org, "D. Hugh Redelmeier" <hugh@mimosa.com>,
joseph@codesourcery.com, Mike Stump <mrs@apple.com>,
Nathan Sidwell <nathan@codesourcery.com>
Subject: Re: volatile semantics
Date: Sun, 17 Jul 2005 05:19:00 -0000 [thread overview]
Message-ID: <OF9B36EC9A.9F0A0C06-ON43257041.001A9E97-43257041.001D3409@il.ibm.com> (raw)
In-Reply-To: <m3mzomnn9q.fsf@uniton.integrable-solutions.net>
Gabriel Dos Reis wrote on 17/07/2005 06:07:29:
> Daniel Berlin <dberlin@dberlin.org> writes:
>
> | Anything it sees anything in a statement with volatile, it marks the
> | statement as volatile, which should stop things from touching it
> | (anything that *does* optimize something marked volatile is buggy).
> great!
>
I can't agree with that as is. I would refine it to:
Anything that *does* optimizes away visible reads or writes of
something marked volatile is buggy.
> | I should note that this will probably annoy the people who reported :
> |
> | http://gcc.gnu.org/bugzilla/show_bug.cgi?id=3506
>
> The rationale is:
>
> GCC doesn't know what constitutes a reference to a volatile memory,
> so it never performs operations on them directly. It will always
> pull the value into a register first.
This can be optimized, without violating the rule for read/write
visibility. It was over a decade since I were on a Pentium validation
team, so I may be wrong:
inc mem - without a LCK prefix is not atomic, generating a
read, and then write - behaving as if it were 3 instructions:
[read mem; inc; write mem].
What I am getting at, is that gcc is *allowed* to optimize any
sequence of the form [read mem; arithmetic; write mem] to
[arithmetic mem], even for a volatile int.
I do *not* say that this *has* to be optimized, only that it could.
If it had to, gcc would have a very difficult time with RISC
load/store architectures. No reasonable interpretation of the
standard may claim that gcc *should* generate a single "inc y".
I claim that in this case gcc *could* generate a single "inc y",
and as such it is a missed optimization (and the PR is valid).
It may be a very difficult optimization to implement, so
it can be postponed indefinitely, but it should not be marked
INVALID. It should be marked a SUSPENDED missed
optimization opportunity (for x86 and probably several
other CISCs - as long as read/write is not atomic).
Michael
next prev parent reply other threads:[~2005-07-17 5:19 UTC|newest]
Thread overview: 118+ messages / expand[flat|nested] mbox.gz Atom feed top
2005-05-03 8:41 Mike Stump
2005-05-03 8:52 ` Paolo Bonzini
2005-05-03 9:53 ` Giovanni Bajo
2005-05-03 14:31 ` Dave Korn
2005-05-03 14:41 ` Nathan Sidwell
2005-05-03 17:04 ` Dale Johannesen
2005-05-03 18:03 ` Nathan Sidwell
2005-05-03 18:32 ` Paul Koning
2005-05-03 18:37 ` Dale Johannesen
2005-05-11 23:01 ` Geoffrey Keating
2005-05-12 14:46 ` Paul Koning
2005-05-03 18:35 ` Dale Johannesen
2005-05-03 18:54 ` Nathan Sidwell
2005-05-03 19:25 ` Dale Johannesen
2005-05-04 8:18 ` Nathan Sidwell
2005-05-04 10:23 ` Andrew Haley
2005-05-04 11:30 ` Gabriel Dos Reis
2005-05-04 17:59 ` Dale Johannesen
2005-05-04 18:01 ` Paul Koning
2005-05-04 19:49 ` Nathan Sidwell
2005-07-16 16:51 ` D. Hugh Redelmeier
2005-07-16 16:56 ` Daniel Berlin
2005-07-16 17:26 ` Nathan Sidwell
2005-07-16 17:32 ` Daniel Berlin
2005-07-16 18:35 ` Nathan Sidwell
2005-07-16 20:35 ` Daniel Berlin
2005-07-16 21:29 ` Gabriel Dos Reis
2005-07-16 21:41 ` Daniel Berlin
2005-07-16 21:59 ` Gabriel Dos Reis
2005-07-16 21:07 ` Gabriel Dos Reis
2005-07-16 19:20 ` D. Hugh Redelmeier
2005-07-16 21:10 ` Gabriel Dos Reis
2005-07-16 20:52 ` Gabriel Dos Reis
2005-07-16 21:07 ` Daniel Berlin
2005-07-16 21:24 ` Gabriel Dos Reis
2005-07-16 21:30 ` Daniel Berlin
2005-07-16 22:19 ` Gabriel Dos Reis
2005-07-16 21:36 ` Daniel Berlin
2005-07-16 22:06 ` Gabriel Dos Reis
2005-07-16 22:17 ` Daniel Berlin
2005-07-16 22:25 ` Gabriel Dos Reis
2005-07-19 7:27 ` Kai Henningsen
2005-07-19 9:25 ` Gabriel Dos Reis
2005-07-16 22:34 ` D. Hugh Redelmeier
2005-07-16 22:57 ` Gabriel Dos Reis
2005-07-17 1:37 ` D. Hugh Redelmeier
2005-07-17 2:24 ` Gabriel Dos Reis
2005-07-17 2:36 ` Daniel Berlin
2005-07-17 3:08 ` Gabriel Dos Reis
2005-07-17 4:32 ` Michael Veksler
2005-07-17 5:19 ` Michael Veksler [this message]
2005-07-17 5:31 ` Gabriel Dos Reis
2005-07-17 7:33 ` Michael Veksler
2005-07-17 14:33 ` Daniel Berlin
2005-07-17 15:30 ` Michael Veksler
2005-07-17 7:53 ` Andrew Pinski
2005-07-17 11:41 ` Gabriel Dos Reis
2005-07-17 7:40 ` D. Hugh Redelmeier
2005-07-17 11:50 ` Gabriel Dos Reis
2005-07-18 19:30 ` Mike Stump
2005-07-18 20:05 ` Gabriel Dos Reis
2005-07-17 15:45 ` Richard Henderson
2005-07-17 16:04 ` Nathan Sidwell
2005-07-17 16:18 ` Richard Henderson
2005-07-17 16:54 ` Gabriel Dos Reis
2005-07-17 16:06 ` Falk Hueffner
2005-07-17 16:18 ` Ian Lance Taylor
2005-07-17 16:44 ` Richard Henderson
2005-07-17 12:49 ` Joseph S. Myers
2005-07-17 2:27 ` Daniel Berlin
2005-07-17 3:14 ` Gabriel Dos Reis
2005-07-17 3:27 ` Daniel Berlin
2005-07-17 20:34 ` Mark Mitchell
2005-07-17 4:38 ` D. Hugh Redelmeier
2005-07-17 5:27 ` Gabriel Dos Reis
2005-07-18 13:13 ` Gerald Pfeifer
2005-07-17 7:54 ` D. Hugh Redelmeier
2005-07-17 10:11 ` Andrew Haley
2005-07-17 12:03 ` Gabriel Dos Reis
2005-07-16 17:33 ` Andrew Haley
2005-07-16 17:53 ` Daniel Berlin
2005-07-17 8:25 ` Ian Lance Taylor
2005-07-22 23:20 ` Geoffrey Keating
2005-07-22 23:33 ` Ian Lance Taylor
2005-07-23 1:28 ` Geoff Keating
2005-07-23 2:59 ` Gabriel Dos Reis
2005-07-23 9:50 ` Geoff Keating
2005-07-23 6:03 ` Ian Lance Taylor
2005-07-23 16:03 ` Mike Stump
2005-07-16 19:05 ` Dale Johannesen
2005-07-16 21:17 ` Gabriel Dos Reis
2005-07-22 23:20 ` Geoffrey Keating
2005-07-25 23:08 ` Olivier Galibert
2005-05-06 0:45 ` Kai Henningsen
2005-05-06 1:42 ` Paul Koning
2005-05-06 2:04 ` Gabriel Dos Reis
2005-05-06 2:57 ` Dale Johannesen
2005-05-03 21:19 ` Thorsten Glaser
2005-05-06 5:06 Paul Schlie
2005-07-17 17:58 Paul Schlie
2005-07-18 1:29 Paul Schlie
2005-07-18 6:36 ` Gabriel Dos Reis
2005-07-18 11:20 ` Paul Schlie
2005-07-18 12:12 ` Paolo Bonzini
2005-07-18 12:17 ` Paul Schlie
2005-07-18 12:27 ` Gabriel Dos Reis
2005-07-18 13:27 ` Paul Schlie
2005-07-18 15:47 ` D. Hugh Redelmeier
2005-07-18 12:24 ` Gabriel Dos Reis
2005-07-18 12:11 ` Jonathan Wakely
2005-07-18 12:31 ` Paul Schlie
2005-07-19 17:56 Paul Schlie
2005-07-19 18:13 ` Gabriel Dos Reis
2005-07-19 18:32 ` Paul Schlie
2005-07-23 2:15 Paul Schlie
2005-07-23 9:50 ` Geoff Keating
2005-07-23 11:39 ` Paul Schlie
2005-07-23 11:44 ` Paul Schlie
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=OF9B36EC9A.9F0A0C06-ON43257041.001A9E97-43257041.001D3409@il.ibm.com \
--to=veksler@il.ibm.com \
--cc=dalej@apple.com \
--cc=dberlin@dberlin.org \
--cc=gcc@gcc.gnu.org \
--cc=gdr@integrable-solutions.net \
--cc=hugh@mimosa.com \
--cc=joseph@codesourcery.com \
--cc=mrs@apple.com \
--cc=nathan@codesourcery.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 read-only IMAP folder(s) and NNTP newsgroup(s).