[ydl-gen] gfortran under yellowdog

[Derick Centeno]

In order to emulate a conversation my reply follows yours below.

On Mon, 2008-09-15 at 11:22 +0200, furio ercolessi wrote:

> On Sun, Sep 14, 2008 at 03:40:17PM -0400, Derick Centeno wrote:
> > Buon Journo, Furio !
> 
> Hello Derick,
> 
> Thanks for your clarifications about g95 and GPL.  I admit that I
> had not dug into this subject, although I suspected a situation
> of this type.  I do not and did not blame YDL for not including g95, 
> and in fact I suspect that no other distribution includes it.
> 
> Unfortunately, experience (mine to a limited extent, more importantly 
> of other colleagues of mine that develop and maintain large Fortran codes)
> shows g95 to be less problematic than gfortran for serious projects
> (not in the area of performance, but in the area of compiler bugs that 
> burn your time!)
> This surely reflects the brilliance and capability of the single man 
> behind g95, but perhaps also the fact that development resources
> allocated to gfortran are not as healthy as they should be, from
> the point of view of professional Fortran users.
> 

I understand and agree with your assessment.  It is odd that after so
much development throughout the history of modern computing sciences
fortran, not any of the modern languages, remains the workhorse language
for the sciences.  Your conversation presents the impression that you
and your colleagues are involved in a university wide effort.  If this
assessment is correct then I'm sure you can, as a university research
project, define your own GPL variant which would provide you a legal
framework to use g95 quite effectively.  If you proceed along that
direction, it could be helpful for other professionals who require legal
assurances that their work and effort will be protected and unchallenged
in the future.  I'm sure there are more than a few who would be more
secure beyond the claim of the g95 project that "it's free crunch time".

> Of course there are commercial alternatives, for instance the Intel 
> compiler, for who can pay to achieve performance and quality.

Apparently g95 was able to prove themselves superior to commercial
fortran even provided by Intel.  Here's a quote:

FUN3D - NASA aerodynamic and aerothermodynamic analysis and design codes
Bil Kleb, one of the developers, wrote: 

        The g95 compiler is one of only two compilers that have been
        able to compile the FUN3D suite of codes without encountering an
        internal compiler error. Thanks to agile software development
        practices, the FUN3D suite of codes have revealed compiler bugs
        in nearly every Fortran compiler: Intel, Portland Group, Absoft,
        Cray, DEC, SGI, Sun, HP, IBM, PathScale, NAG, NAS, and Salford.
        Lahey-Fujitsu is the only other one we haven't been able to
        break yet.
        

The above was located on this page: http://www.g95.org/g95_status.shtml

It is my hope that "free crunch time" is not a merely smart play on
words.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.terrasoftsolutions.com/pipermail/yellowdog-general/attachments/20080915/0166c9dd/attachment-0001.html 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url : http://lists.terrasoftsolutions.com/pipermail/yellowdog-general/attachments/20080915/0166c9dd/attachment-0001.bin