The feature of sub-modules was published as an official ISO-extension to Fortran 2003 back in 2005. Now it is a part of Fortran 2008.
As far as I know only Cray has implemented this feature. I'm not sure if it is low market request, or technical challenges that leads to the low priority among compiler vendors.
I think Metcalf, Reid and Cohen explain very well the needs for this feature for large scale programs in their book "Modern Fortran Exlpained", so I will not repeat them here.
However, based on bad experience with maintainance of very large modules we will not initiate any large scale projects in Fortran before sub-modules are implemented in Intel Visual Fortran. If it takes to long we may be forced to do it in another language.
As a consequence I hope to see sub-modules soon in the Intel compiler!
If nested submodules are tricky to implement you could get things going by releasing an implementation without nested submodules.
What is Intels position and schedule on sub-modules?
Remark:
I do agree that very large modules may indicate a bad design of the actual program. However the present version of the Intel compiler force you to combine interconnected subfeatures of a complex concept (class) into a single module and file. If in additions the related algorithms (methods) are comprehensive the size of the module will be enormous. This is a nightmare from a maintenance point of view. We will not go down that road again.
Kind Regards,
Jens B. Helmers
