fortran 2003 and openmp

[vweber]

Dear All
I would have expected openmp4.0 to be fully fortran2003 compatible, so it there any reason for not taking that move?
Can fortran users expect openmp to take care of the f2003 features any time soon?
Thanks
Valery

[ejd]

The committee is made up of individuals from each of the member organizations who generally have other responsibilities in their company than just OpenMP. Each topic for a new release is suggested by someone, prioritized, and voted on for possible inclusion. Once a candidate list is drawn up, someone has to chair a sub-committee, lead the discussions, and submit a proposal for addition to the spec. This proposal is then reviewed by the full committee and if it passes a vote of the member organizations, added to the spec. Because of the amount of work required, not a lot of new features can be added for any new version of the spec.

So while it might seem like it should be trivial to add a new version of a base language, it takes a lot of work trying to look at all the new features of a language, trying to figure out how best to support them within OpenMP, and going through the process. Fortran 2003 was looked at and added this time - but not all of the new features are supported. This is a continuing process.

Another thing to be said though, is that a lot of the member organizations are computer manufacturers and so customer feedback is very important. If there is some feature that users say they want, it has a better chance of being higher on the priority list and making it through the process. That said, what is it that you were expecting to see support for?

[vweber]

I would have very much appreciate the classes to be part of openmp4.0 as this is the most or at least one of the most interesting feature in f2003.
I noticed that some compilers seem to support classes allready (eg ibm, intel, cray, pgi) within openmp3.x sections. Wouldnt that help tagging this
feature with the highest priority?
Thanks
Valery

[ejd]

A "reference" implementation does help for any new feature trying to be added, because it gives the members something to look at and "play" with. However, it doesn't guarantee that the priority will be high. For example, individual vendors may not want to push a new feature into a "standard", because it gives their product an advantage over the competition. Another example, is that a vendor might have put out some feature to see if users think it useful and might still have reservations about it (e.g., should it be implemented differently, is the user interface good enough, etc). Remember too, that once the feature is "standardized", then it is easier for a customer to move their code to another platform. From a user perspective useful, but from a vendor's perspective a reduction of revenue.

So it is not as simple as it seems and not necessarily just because of technical issues. In any case, now that you have asked the question, the committee will review it and give a formal response.