Re: gfortran IRC meeting?

[Tobias Burnus <burnus at net-b dot de>]

Dear Paul, dear all,

Paul Richard Thomas wrote:
> Given that we are back in the clear blue skies of stage 1 once more,
> maybe we should have another IRC meeting?

Yes, I am in favour it. Usually we had something like: Saturday 2100 CET 
/ 2000 UTC / 1 p.m. PDT. But I would also be fine with Sunday, e.g. 2100 
CEST / 1900 UTC / 12 noon PDT.  (I hope I did not mix up the different 
daylight-saving/summer-time switch overs. I think in the US it happened 
10 days ago, in the EU it will be upcoming weekend.)

I am not sure whether we could agree to do it tomorrow - or whether that 
would be too soon.

> Jerry's merge of trunk into fortran-dev, reminding me of an
> unfulfilled commitment to do the array descriptor reform.  I would
> like to discuss how to proceed with that, given that I got completely
> stuck with the change to the dimension triplet, whilst other changes
> to the descriptor are more important to fix outstanding PRs.

I think the array-descriptor reform is really needed, but quite invasive 
and tedious (to get correct and fast code).

> We need to sort out how to and who will implement arrays in OOP.  I
> believe that the hooks are there but that some determined effort is
> required to get it going correctly. Unlimited polymorphic entities
> should be implemented too.

It would be useful to understand the requirements of unlimited 
polymorphism/polymorphic arrays before implementing the array descriptor.

> What sort of supprot do you need for co-arrays?

I think one can do with "int corank" - and using (rank + corank) 
dimension triplets rather than just (rank) triplets. Actually, one might 
be even able to do without as the (co)rank is known at compile time, 
though the triplets need to be available (at least for allocatable 
coarrays). The funny think about coarrays is that the coranks do not 
need to match (except for allocatable coarrays).

We should also try to keep the upcoming TR on further interop with C in 
mind. The last draft is at
ftp://ftp.nag.co.uk/sc22wg5/N1801-N1850/N1845.pdf - It is expected to be 
forwarded soon as PDTR to SC 22. As written before, one should consider 
to use it as internal format, to make implementation of the TR easier - 
besides, some smart people have thought about the format. (The format 
allows extra fields.)

Tobias