Hi everyone,
I think I can offer to pick up the work again. There were some strange
details about pointers and coarrays which had me thinking that the
approach taken for the native coarrays was doomed to failure (which is why
I didn't continue the work), but I now think that the Fortran standard
took these limitations into account and made, albeit in a somewhat cursed
way, the problematic programmes illegal. I also think I can do it in a
timeframe so it could be merged as an experimental feature for gcc 14.
Regarding funding: I will not be able to meaningfully spend 600 K :)
so maybe a division of work could be done? Andre's company does general
Fortran conformance and bug fixing, and they contract with me for the
native coarrays?
Best regards
Nicolas
On 27/05/2023 18:19, Paul Richard Thomas via Fortran wrote:
Hi Andre,
It's good to hear from you.
I would suggest the following:
(i) Complete F2003 compliance - Now that finalization is within a whisker
of compliance, this mostly leaves putting PDTs right. The framework is all
there, it just needs revamping. I can provide guidance or a statement of
work here. Associate still has some problems that I am working through but
I expect that there will be one or two of the more difficult ones left (eg.
Derived type, sibling function selectors that have not yet been parsed).
(ii) Complete F2008/F2018 compliance - we have owed Ian Chivers and Jane
Sleightholme this information for quite a while. Perhaps we should divide
the forms between us and attempt to fill them out? Or better, perhaps, this
could be the first stage of the work. I am sure that we will find lots of
this like, for example, partial coverage of do concurrent.
(iii) Finishing Nicolas's work on native (did we agree not to call it
that?) co-arrays would also be excellent.
(iv) Finally, a thorough and systematic attack on the PRs would be great,
starting with the meta-bugs.
However, an agreement, as Vladimir Illyich put it, an "What is to be done"
is an important first step.
I can only repeat Thomas's questions about whether or not your company
could provide the administrative framework and, perhaps, some project
management?
Could Sebastian please provide us with information on what is required for
the grant application?
Finally,
Regards
Paul
On Sat, 27 May 2023 at 12:24, Andre Vehreschild via Fortran <
fortran@gcc.gnu.org> wrote:
Hi Thomas,
working full-time on a gfortran engagement would be possible for me. Given
that
my company is Germany based and we have some capacity, that would be
feasible
for us. We also have some knowledge about how to invoice authorities,
which can
be a bit difficult sometimes.
So I regard coarray work as a good starting point. I am also in contact
with
Damian about some of his ideas. What else could we tackle?
Regards,
Andre
On Sat, 27 May 2023 10:08:56 +0200
Thomas Koenig via Fortran wrote:
On 26.05.23 23:22, Jerry D via Fortran wrote:
Sorry about my messages getting chopped.
On 5/25/23 9:34 PM, Jerry DeLisle via Fortran wrote:
Hi all,
I found this message in my spam folder tonight.
Please look. I also sent a note to Damian on this.
Maybe we can get someone to push forward on te native coarrays work?
I think the native coarrays are a good field. General bug fixing would
also be a good idea.
[quoting for the mail]
I would like to discuss with the GFortran developer community whether
there is interest to setup a joint project to pay somebody to work on
GFortran full time. We have funding available for 18 months with 600k
EUR starting mid of June (please do not share this numbers publicly
yet), but we can also ask the fund for more money if needed. What do
you
think? Is it worth to bring this up on the GFortran mailing list or
mattermost server?
It is really good so finally see a source of gfortran funding.
For hiring somebody full-time for a year, I am not sure who would be
available full-time, I think most people who have experience working
on gfortran have other commitments. We would have to see if somebody
has the free time.
What would be great would be a possibility for people to work on
an hourly basis on certain, well-defined projects. This is probably
something that contributors could fit in much better, and would provide
an additional incentive to take up gfortran work again :-)
Do you know if this is, in fact, a possibility?
Best regards
Thomas
--
Andre Vehreschild * Email: vehre ad gmx dot de
