[cpp-threads] First draft of coherence requirements from n3136

Paul E. McKenney paulmck at linux.vnet.ibm.com
Wed Nov 10 22:38:05 GMT 2010


On Wed, Nov 10, 2010 at 02:22:08PM -0800, Nelson, Clark wrote:
> The write-write coherence requirement should use the words "atomic
> object M", like all the other coherence requirements.

Fixed.

> The phrase "is knows as" should instead read, "is known as" (in four places).

Hey, at least I am consistent!  Fixed.

> I'm tempted to suggest that the names of the different coherence
> requirements actually shouldn't be italicized as definitions, since
> there aren't going to be any (normative) references to them from the
> standard itself.

I de-italicized them.

> In the description of read-write coherence, there should be a comma
> before "then". Also, it currently reads, "shall either ... or shall"
> -- the parallelism is poor (no pun intended). My suggestion would be to
> delete the second "shall".

I added the comma.  Mark Batty beat you to the "shall either ... or shall"
via email, and asked that initialization be treated as an early side effect,
which is what we do elsewhere, so this paragraph now reads:

	If a value computation A of an atomic object M happens before an
	operation B that modifies M then A shall either take its value
	from some side effect X, where X precedes B in the modification
	order of M, or shall take its value from the initial value of
	M. [ Note: This requirement is knows as read-write coherence. —
	end note]

> In the description of write-read coherence, delete the hyphen in
> "happens-before".

Fixed!

> I think the final note needs a little polishing. I think it would be
> better if it started, "The visible sequence of side effects", since
> it no longer appears in the paragraph defining "visible sequence of
> side effects". I also suggest, "The intended reading <del>of these four
> coherence requirements</del> is that....", because we're really talking
> about the reading of almost the whole section.

Fixed!

Update^2 attached.

> Clark
> 
> > -----Original Message-----
> > From: cpp-threads-bounces at decadent.org.uk [mailto:cpp-threads-
> > bounces at decadent.org.uk] On Behalf Of Paul E. McKenney
> > Sent: Wednesday, November 10, 2010 11:34 AM
> > To: michaelw at ca.ibm.com; bkoz at redhat.com; mjb220 at cl.cam.ac.uk
> > Cc: cpp-threads at decadent.org.uk
> > Subject: [cpp-threads] First draft of coherence requirements from n3136
> > 
> > Hello!
> > 
> > Attached is a revision of n3125 and n3136 that has a draft of 1.10p13 and the
> > new coherence paragraphs at the very end of the document.
> > I will be folding in the other comments and extracting wording into working-
> > document order, but wanted to let Mark Batty take a look at this draft of the
> > coherence requirements.
> > 
> > Thoughts?
> > 
> > 							Thanx, Paul
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.decadent.org.uk/pipermail/cpp-threads/attachments/20101110/5f789cef/attachment-0001.html>


More information about the cpp-threads mailing list