(j3.2006) (SC22WG5.5145) [ukfortran] From WG9 minutes

[N.M. Maclaren]

On Dec 10 2013, David Muxworthy wrote:
>
>Some agent acting for the WG5 mail server claims that my ISP
>is a distributor of spam and asks for money to let mail through.
>It will not even let me get to the mail server administrator to
>complain.  Hence this message is sent to a subset of the WG5 list.

That's a known attack.  I can't remember the administrator offhand;
please Email me and I will pass it on - provided that I don't get
caught the same way.

>WG21 (C++) are doing that too.  They have a New Work Item Proposal to
>SC22 (SC22 N4857, voting deadline January 27) entitled "C++ Extensions
>for Parallelism".  The scope is to extend the 2011 standard with
>"types and features related to parallelism", but they are making haste
>slowly.

Dammit!  I have been asleep.  I knew that was mooted, but not that it
had reached the stage of a formal proposal.  The trouble is that most of
the C++ mailing lists are high-volume, and often highly irritant.

>The purpose and justification is:
>
>    The purpose of this Technical Specification is to provide a
>    reference for existing practice to gain experience with the
>    design and pave the way for eventual standardization of this
>    functionality in a future revision of the C++ standard
>    itself. Specific design questions that are still being explored,
>    thus making this not yet suitable for an IS, include (but are not
>    limited to): how to express parallel and vector operations in an
>    extended Standard Template Library style of composable
>    containers, iterators, and algorithms.

The second sentence is precisely why it is currently impossible to do
the task in standard C++ - the STL's design in general and iterators in
particular are fundamentally tied to serial semantics.  When I discussed
that proposal with its proponents, it was along the lines of the first
sentence, and they agreed that any solution might well involve breaking
away from the current design.


Regards,
Nick.