Re: OSSODA and (free) Fortran 95/200x

[Scott Robert Ladd <coyote at coyotegulch dot com>]

Toon Moene wrote:
> We are seeing a remarkable decline in the Fortran user community both
>  inside and outside of the general HPC community. Many HPC community
>  codes are turning to C/C++, but Fortran is still very important in
> the NNSA Tri-Laboratory HPC community.
>
> A possible reason behind the decline in the Fortran user base is the
>  lack of a standard-compliant, open-source compiler. There is no
> single cross-platform standard compliant open-source compiler
> solution.

The situation extends beyond the national laboratories into other 
domains. In my experience, the biggest knock against Fortran 95 is 
limited expertise, which is only partially related to the lack of an 
open source compiler. Schools no longer teach Fortran; the knowledge 
base is aging.

I'd love to find some Fortran 95 work -- but I haven't seen any in the 
last couple of years. It's a shame, too; Fortran 95 is a very solid 
programming language with several advantages over C/C++/Java in clarity 
and capability.

> The open-source aspect would get a compiler into the hands of 
> universities at no cost. This would serve to increase the Fortran 
> programmer base, which will, at the very least, provide a base from 
> which to draw Fortran programmers. At best, it will make the language
>  more popular which would in turn increase the likelihood of better 
> quality Fortran compilers and tools."

A free Fortran 95 won't make universities teach courses in Fortran. 
University courses are driven by market and corporate pressures -- and 
there just isn't a strong demand for Fortran 95 expertise.

I don't know of *any* free software projects based in Fortran 95, other 
than some minor efforts of my own. The advantage of having gfortran will 
be that people like me will write free software in Fortran 95. Right 
now, my Fortran code is useless to anyone without a commercial compiler 
-- and that limits the audience significantly.

> As the maintainer of g77 I appreciate this analysis (and I agree with
> it fully), but I am surprised that the document suggests that a free
>  Fortran 95 compiler doesn't exist at all.
>
> The effort to create g95 dates back to 1999.  In July of this year,
> the code base was integrated into the central GCC CVS repository. We
> expect that the code, which is part of an experimental branch at the
> moment, will be ready for release by the end of 2004.

GNU Fortran 95 is not currently finished, or generally available -- and
that is why it is not considered a solution by several
clients/customers/associates of my acquaintance. The "end of 2004" is a 
long way off, in planning terms. If I define a project today, I need to 
define it in terms of the tools I can use today.

> The obvious question, from our side (the developers of free
> software), is: Why would you want to bootstrap a new effort instead
> of supporting an existing one ?

I can't fathom why they'd want to start a new effort; their best option 
would be to accelerate the existing gfortran effort. For example, we 
could really use some technical assistance with implementing OpenMP in 
gfortran -- something very important to certain segments of the HPC 
community.

..Scott

--
Scott Robert Ladd
Coyote Gulch Productions (http://www.coyotegulch.com)
Software Invention for High-Performance Computing
In development: Alex, a database for common folk