On 08/06/2010 09:16 PM, Van Snyder wrote:
> On Thu, 2010-08-05 at 19:14 -0700, Vivek Rao wrote:
>> Especially someone who has proposed many features in Fortran 2008
>> should be aware that several compiler vendors have decided not to
>> produce compilers beyond Fortran 95 (except for some TR's).
Well, the question is why? In any case there is a large number of
Fortran compilers which have implemented some parts of Fortran 2003, add
more Fortran 2003 support and plan to complete Fortran 2003/2008 support
at some point.
With the same argument, one should not have started Fortran 90 or moved
to 95 as a large number of compilers stayed with Fortran 77. And even
today, many users only use Fortran compilers for old Fortran 77 programs
or for Fortran 90/95 programs. The latter also because the developers
restricted themselves to those features all compiler supported. However,
I see a change: More and more Fortran 2003 features are used - some
deliberately and some (like non-integer initialization expressions) more
accidentally. I think that's driven in part by the market concentration
in terms of hardware (thus effectively, fewer compilers need to be
supported) but also by the availability of g95/gfortran which invalidate
excuses such as that for one platform no newer compiler exists or that
is to expensive.
>> [...]
>> Maybe someone can comment about Pathscale.
The latest Pathscale 3.3 beta contains some Fortran 2003 features, cf.
http://www.pathscale.com/docs/3.2.99/release_notes.txt ; I think the
IEEE TR is also supported and some other F2003 features which were added
before 3.3. I think the new PathScale Inc. is now spending more work on
their Fortran compiler - and I saw that they are looking for a Fortran
developer ... (I wrote "new PathScale" as Cray bought Pathscale from
SiCortex (in August 2009, [1]) - and it is now developed by
NetSyncro.com with part of the old staff; parts of the PathScale
compiler became opensource.)
>> The committee should restrict itself to maintenance mode until full
>> F2008 compilers are available.
> Why?
Actually, why not? ;-) [see below]
> With the exception of coarrays, Fortran 2008 was essentially a
> maintenance project.
That actually also matches my impression - I have the feeling that
gfortran will first complete the support for the newer F2008 features
before finishing some of the remaining Fortran 2003 tasks.* The only
larger Fortran 2008 task seems to be coarray support but that's already
supported to some extend by Cray, g95, gfortran (well, only one image
and incomplete), and seemingly by Intel's 11.1. The other tasks seem to
be relatively minor - looking at the list of features and at what
gfortran and other compilers currently support.
(Cf. http://fortranwiki.org/fortran/show/Fortran+2008+status and for
gfortran also http://gcc.gnu.org/wiki/Fortran2008Status )
Thus, I would not be surprised if the first so-called "full Fortran
2008" compiler would be available around the time the draft of the
coarray TR is finished. [So called because it will take years to find
all those cases where minor restrictions has been lifted at various
places in the standard.]
In that sense, I think "restrict itself to maintenance mode" and TR work
until the first Fortran 2008 compiler is available is compatible with
the idea of doing for about a year only maintenance and TR development.
> A significant reason that Fortran fell into bad odor in the computer
> science community between 1966 and 1990 was that it had stagnated.
And one reason that it stagnated after a couple of years after 1990 was
the lack of compilers ...
> If ANSI and ISO committees had taken Vivek's advice in 2004 (and there
> were others offering the same advice), Fortran 2008 wouldn't have
> co-arrays, which are an enormous leap forward compared to PVM or MPI or
> HPC or OpenMP.
While I do like coarrays, I do not see them as being "an enormous leap
forward"; they make Fortran more attractive, they offer some nice
features and integrate nicely into the language, but one has to do the
same effort as with MPI to parallelize the code. (The coarray syntax
seems to be easier, but the structure of the program is the same as with
MPI.)
Bill Long wrote:
> Language standardization takes a long time. By its very nature, it is
> always "ahead" of the average user of the language.
Surprisingly, it is usually also ahead of compiler implementation and
ahead of advanced users ;-)
Tobias
[1]
http://investors.cray.com/phoenix.zhtml?c=98390&p=irol-newsArticle&ID=1324995&highlight=
* GCC: The changes which are in the GCC development version (4.6.0) are
listed at http://gcc.gnu.org/gcc-4.6/changes.html#Fortran While the
Fortran 2008 list is much longer, most of them are small items which
could be implemented quickly - contrary to the on-going Fortran 2003 work.
|