This is the mail archive of the fortran@gcc.gnu.org mailing list for the GNU Fortran project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Is GFortran stable enough to actually use?


On Fri, Jul 14, 2006 at 10:56:31AM -0400, Joe Krahn wrote:
> GFortran is able to compile very little of my Fortran90+ code. But, it 
> seem that it is now supposed to be in 'stable' production mode. Should I 
> start sending in a huge number of bug reports, or are there already 
> enough known bugs to stay busy? And, where do I send bug reports. 
> gfortran.org and the Gfortran Wiki seem to no longer exist.

The wiki is temporary down while it is moved to a new server.
You can send reduced testcases to the list.

> Meanwhile, G95 works really well, even better than Intel Fortran.

Check out Polyhedron's benchmarks.  Gfortran out performs g95.

> It's 
> too bad that Gnu is unwilling to let G95 be the real compiler and still 
> let Andy be the controlling author, sort of like Linus and Linux.

This is revisionist history.  Andy is the person who took his ball
and ran home.  After Paul Brook and Steven Bosscher worked to interface
Andy's (very broken) parser to the rest of gcc (middle and back end),
Andy suddenly closed his source.  Then the source he did release was
purposely obfusicated so fixes could not be back ported to gfortran. 
Andy does appear to have Linus envy.

> Right now, G95 is so much better, it's hard to consider wasting time on 
> GFortran.

Does g95 have Cray pointers or OpenMP?  g95 does not compile on FreeBSD
AMD64.  Additionally, read Andy's blog and count the occurrence of
the work "regression".  gfortran has fairly strict requirements with
passing its large testsuite, so development may appear slow.

> But, if the hardcore FSF ideals make it impossible to accept 
> G95, I am willing to at least contribute good bug reports to GFortran.

The 2 code bases have deviated substantially.  For example, gfortran
uses GMP/MPFR for its internal representation of types while Andy
uses BIGNUM (a third party package that would need to be assigned to
FSF if g95 were to be integrated into gcc).

-- 
Steve


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]