future plans?
damian...@nrl.navy.mil
Specifically I'm interested in the following:
(1) Ability to specify intent for pointer arguments
(2) Support for OpenMP
(3) Support for classes and inheritance.
Damian
Andy
damian...@nrl.navy.mil wrote:
> Is there any information on plans for the near future of g95?
> Specifically I'm interested in the following:
>
> (1) Ability to specify intent for pointer arguments
An oversight, I've got this implemented now.
> (2) Support for OpenMP
I was more interested in supporting fortran's array expressions, but
people keep requesting openmp.
> (3) Support for classes and inheritance.
You're only the second one to request this so far. I will eventually
get to it, but not soon.
Andy
Helge Avlesen
On Oct 13, 7:37 pm, "Andy" <a...@firstinter.net> wrote:
> > (2) Support for OpenMP I was more interested in supporting fortran's array expressions, but
> people keep requesting openmp.
I put in my vote for co-arrays.
openmp does not scale and cannot be used across nodes on a cluster. I
also think it looks like compiler diagnostics would be far more
informative for co-array syntax.
Helge
Al Greynolds
OpenMP should be implemented first. It is the de-facto standard for
parallel processing on shared memory hardware (like the new multi-core
CPUs). In fact, of all the Fortran-95 compilers I use (g95, IBM XLF,
Intel VF, Lahey LF95, and gfortran), g95 is the only one that does not
have an implementation of OpenMP
Al Greynolds
www.ruda.com
On Oct 15, 4:18 am, "Helge Avlesen" <hel...@gmail.com> wrote:
> On Oct 13, 7:37 pm, "Andy" <a...@firstinter.net> wrote:
>
> > > (2) Support for OpenMP I was more interested in supporting fortran's array expressions, but
> > people keep requesting openmp.I put in my vote for co-arrays.
Fly Away
On Oct 16, 7:02 am, "Al Greynolds" <awgreyno...@earthlink.net> wrote:
> Co-arrays are definitely more elegant but from a practical standpoint,
> OpenMP should be implemented first. It is the de-facto standard for
> parallel processing on shared memory hardware (like the new multi-core
> CPUs). In fact, of all the Fortran-95 compilers I use (g95, IBM XLF,
> Intel VF, Lahey LF95, and gfortran), g95 is the only one that does not
> have an implementation of OpenMP
>
> Al Greynoldswww.ruda.com
> > Helge
With all my great respect to Andy, implementing OpenMP seems like a big
task for one developer and implementing both OpenMP and co-arrays seems
quite impossible within reasonable timeframe. So I'd rather vote for
the way of the future (co-arrays) so that in a few years we'll have a
good modern elegant tool for parallel processing.
Al Greynolds
thread-safe runtimes) that would be applicable to an eventual Co-array
implementation. Besides I believe a lot of the groundwork has been
already laid in GCC 4.X (OpenMP supports C too). OpenMP also allows
parallelization of existing code that isn't or can't be easily
expressed in any array syntax.
On Oct 16, 8:36 am, "Fly Away" <victor.proso...@gmail.com> wrote:
> > > HelgeWith all my great respect to Andy, implementing OpenMP seems like a big
Garry Willgoose
Al Greynolds wrote:
> I would think there would be a lot of the OpenMP stuff (such as
> thread-safe runtimes) that would be applicable to an eventual Co-array
> implementation. Besides I believe a lot of the groundwork has been
> already laid in GCC 4.X (OpenMP supports C too). OpenMP also allows
> parallelization of existing code that isn't or can't be easily
> expressed in any array syntax.
>
> On Oct 16, 8:36 am, "Fly Away" <victor.proso...@gmail.com> wrote:
> > > > HelgeWith all my great respect to Andy, implementing OpenMP seems like a big
> > task for one developer and implementing both OpenMP and co-arrays seems
> > quite impossible within reasonable timeframe.
In a choice between co-arrays and openMP I would go openMP. I agree
with Al on this. openMP is here now and g95 is the only compiler I use
without openMP (though I note that many compilers only have v1.0, not
2.0, support ... which doesn't support modules). With multi-core CPUs
becoming the norm there is a need to support something. Given the
fairly indifferent takeup of HPF and relatively widespread takeup of
openMP I'm not so sure about arguments about co-arrays being the way of
the future. openMP may not be as elegant but its simple and fast to
develop ... and in any argument between those two Fortran has always
stumped with speed.
Garry Willgoose
Sean
It is implemented in both sun's f95 compiler and intel's fortran
compiler, so I assumed it was part of the standard.
Also as a new member of the discussion group, I would like to say thank
you for the good work you are doing. Oh, and I would vote for co-arrays
over OpenMP.
On Oct 13, 3:18 am, "damian.rou...@nrl.navy.mil"
beli...@aol.com
> Did you have any plans for intrinsic interval arithmetic?
> It is implemented in both sun's f95 compiler and intel's fortran
> compiler, so I assumed it was part of the standard.
I don't think it is. I don't see "interval arithmetic" mentioned in the
table of contents or the index of the book "Fortran 95/2003 explained".
<snip>