forums.silverfrost.com

Peter Bladon · Joined: 29 Sep 2008 Posts: 4 Location: Glasgow

Does Ftn95 have support for ponter arithmetic?
_________________
Dr. Peter Bladon
Gallowhill House, Larch Avenue
Lenzie, Kirkintilloch
Glasgow G66 4HX

mecej4 · Joined: 31 Oct 2006 Posts: 1899

The F95 standard allows variables to be given the Pointer attribute, and contains provisions for pointer assignment and allocation and status inquiry. FTN95 supports these features.

However, you explicitly wrote "pointer arithmetic". If you mean the sort of calculation that you can do in C with pointer variables, you will probably not find that in Fortran. As you may be aware, one does not use a dedicated dereferencing operator in Fortran (as one does in C, by using an asterisk added as a prefix to a pointer variable).

Being a Fortran 95 compiler, FTN95 does not support the ISO C-Interoperability features of Fortran 2003.

PaulLaidler · Posted: Fri May 15, 2015 6:35 am Post subject:

In other words, the C-Interoperability features of FTN95 do not conform to the 2003 standard. They have the advantage of being much simpler but the disadvantage of lacking strong type checking and portability.

Hopefully it won't be too long before these features are added to FTN95.

JohnCampbell · Joined: 16 Feb 2006 Posts: 2615 Location: Sydney

Paul,

PaulLaidler · Posted: Fri May 15, 2015 3:19 pm Post subject:

It would be to both. The "frontend" is common. But please don't make any suggestions right now.

JohnCampbell · Joined: 16 Feb 2006 Posts: 2615 Location: Sydney

Paul,

I look forward to when you announce that suggestions for preferred features from F2003 and F2008 is open.

I must admit that my list would not be very extensive, as I consider most of the recent additions to be unnecessary.

When I ask the question: "Does the added language complexity make Fortran a more robust language?" I am increasingly thinking No.
My impressions of the recent examples of WHERE and FORALL could be easily replaced by a DO loop structure, with little or no performance loss but better clarity of the implied logic. (Both F95 features!)

I found the recent BBC series on computer languages to be very interesting. Where should Fortran go in the next 20 years ? The added complexity of F2003, F2008 and expected F2015/2020 risks it becoming even less accessible, while no change could see it becoming less relevant. Who will do the numerical coding in the future?

My personal rejection of the approach of F2003 and F2008 has been reinforced by the slow take-up of their features in compilers I am familiar with, most notably Intel ifort. If the new features were so good, it wouldn't have taken all this time for their implementation and for ifort it is still not complete. Is the situation similar for other Fortran environments ?
Where is portability, if most compilers have different subset of F2008 available ?

I think that inter-operability with Cxx could be improved in F95, with C_EXTERNAL being reworked to conform more closely to the Fortran changes of the last 20 years, although probably the best change for me would be to just better understand what is required: Anyone know a good description of what is required ?

John