Re: fortran 90 compiler for linux

Pleaes be patient with me as I get on my old soap box.

First, It is not possible to prove that a non-trivial program
is correct. It's undecidable. Therefore there are bugs in every
non-trivial program; and a compiler surely fills the definition
of a non-trivial program.

So all compilers have bugs. Some are known, some are not.
Some could reflect Standard misinterpretations, although
I too am not aware of any.

In the case of commercial compilers like Lahey LF95, PathScale PATHF90
and Intel IFORT,and Silverfrost (formally Salford) FTN95, it has been
my experience that reported (well documented) bugs,
Standards related or not, are taken quite seriously,
acknowledged and fixed, if possible without introducing more
bugs, in reasonably short order (typically next release or
via a "fixed" module that the development team often offers
to send to the bug originator).

This said, at this point in history, again I agree with you.
That is, for the most part most Fotran compilers are reasonably
stable and usable most often without running into bugs, and less
often running into bugs related to Fortran 90 Standards, and probably
even less often related to Fortran 95 Standards.

For the sake of Fortran application integrity, we still need
to work with compiler developers so that when we use their
compilers to generate applicatons (all applicatoins are
"mission critical" in my opinion), those applications
have as much integrity ("soundness") as humanly possible.
(That's why good debugging options like Salford and Lahey's
uninitialized variables, subscript/string ranges, etc.
are so important.

Skip Knoble

On 15 Jul 2005 04:11:58 -0700, beliavsky@xxxxxxx wrote:

-|Walt Brainerd wrote:
-|> Ken Plotkin wrote:
-|>
-|> > On 13 Jul 2005 13:46:33 -0700, "Chris Hulbert"
-|> > <cchgroupmail@xxxxxxxxx> wrote:
-|> >
-|> >
-|> >>Try gfortran or g95
-|> >
-|> >
-|> > Are they complete and standard-conforming?
-|>
-|> There is no complete and standard-conforming
-|> Fortran compiler :-).
-|
-|I don't know of any outstanding bugs in the Lahey/Fujitsu Fortran 95
-|compiler for Windows, version 7.1 in implementing the Fortran 95
-|standard. The few complaints I have seen about LF95 in this newsgroup
-|concern some Technical Reports that followed the F95 standard. I wonder
-|if anyone has a counterexample.

.

Relevant Pages

  • Re: When will this ng come to accept that Fortran needs to go "back to the future" ?
    ... Compilers in particular are _full_ of bugs. ... have ever used a compiler in earniest _without_ submitting a bug report ... I USED the MS V3.31 Fortran compiler for the first time sometime in ...
    (comp.lang.fortran)
  • Re: A Fortran book for returning to Fortran?
    ... features in any modifications you ended up doing to the old code. ... also likely to see better compiler support for current compilers than from ... on modern Fortran when the last time they used Fortran was in the F77 days? ... is lagging considerably behind the standards. ...
    (comp.lang.fortran)
  • Re: detect Inf or NaN
    ... I don't believe it is possible for IEEE 754. ... But the Fortran standard doesn't mandate the ... compilers have existed for systems that did/do not use IEEE. ... distinguishing the two standards. ...
    (comp.lang.fortran)
  • Re: C interop version
    ... interoperability there is also the very recent Technical Specification ... Links to the various standards: http://gcc.gnu.org/wiki/GFortranStandards ... With regards to the compilers: Essentially all recent compilers support the ISO C Bindings feature of Fortran 2003. ...
    (comp.lang.fortran)
  • Re: NLEQ1 and open source Fortran
    ... When I compile it using Intel's Fortran ... Which version of gfortran are you using? ... > fortran compilers and not with open source compilers?? ... Bugs ...
    (comp.lang.fortran)