Re: storing an integer in a double precision

On 5/5/2011 8:29 PM, Tim Prince wrote:

The advantage of Fortran is that only the never-standard features of the
source code in question have to be corrected immediately, and some of
that work ought to be possible under the non-standard compiler currently
in use. The remainder of "porting to modern Fortran" could be done
incrementally. Perhaps this is part of the ground you've gone over.


I have no way of knowing what the competitive pressures Lynn outlined would cause his management to decide was a necessary reaction but I've inferred there would at least be less pressure _if_ the conversion to another popular compiler had been more trivial and, therefore, successful already.

Their code relies on some particular features in the implementation of the Watcom compiler that, while others support the same kinds of things, at least one feature doesn't work exactly the same and correcting this particular issue has proven to be difficult and time-consuming and beyond the resources available to be dedicated to doing so.

One particular implementation detail is not a feature in modern Fortran (although I believe it should be part of the C interop collateral damage in somewhat of the vein of Richard's recent note on "feature tentacle creep") and, while perhaps one might not use it in designing a new code, it is pervasive in the present one. I've looked at alternative ideas on the implementation of the functionality w/o the extension and it is simply butt-ugly to work around it in Standard Fortran whereas C/C++ support the construction natively. Those kinds of things make it difficult to argue to retain Fortran given the other pressures to conform to the competition.

Eventually I'm sure the code could be made to run on the other Fortran compiler that implements the extensions; it may never get the opportunity however.



Relevant Pages