Re: Praise for Gfortran (finally)

Dear Tobias and everyone else involved with Gfortran,

I wish to point out that I was not trying to upset anyone. If I upset
you, then I am sorry.

On Wed, 24 Sep 2008, Tobias Burnus wrote:

|------------------------------------------------------------------------|
|"On Sep 24, 12:11 pm, Colin Paul Gloster <Colin_Paul_Glos...@xxxxxxx> |
|wrote: |
|> Thank you for contradicting Steven Bosscher's claim in |
|news:5b3f0a83-bc01-429b-991a-d42a44ed42c7@xxxxxxxxxxxxxxxxxxxxxxxxxxxx |
|> that people who have not studied the principles of compilers develop a|
|> good compiler |
| |
|You mean that people at Intel, Pathscale, NAG, Absoft, Sun, Portland |
|Group, IBM etc. have not "not studied the principles of compilers |
|develop a good compiler"? Or how should I interpret this?" |
|------------------------------------------------------------------------|

I have checked again what I quoted from Steven Bosscher and I concede
that depending on exactly what he posted meant, he might not have
claimed that Gfortran contributors did not bother to read established
literature on established compilation techniques if doing so late in
life is excluded from someone's "background".

|------------------------------------------------------------------------|
|" I would |
|claim that be larger problem is the relatively smaller user base |
|(compared with C), which was also spread over a large number of |
|Fortran compilers; having new features does also not help to reduce |
|the number of bugs bugs. I would also claim that Fortran is complex |
|enough to overlook something when implementing something new." |
|------------------------------------------------------------------------|

I am not promoting any compiler as lacking bugs, but the arguments you
suggested are not valid for FORTRAN 77, which in the 1990's supposedly
was not supported as well as one had expected considering the many
years compiler writers had used. A quotation from Pages xii and xiii
of the second edition of "Numerical Recipes in FORTRAN : The Art of
Scientific Computing":
"[..]
[..] Also used were a variety
of FORTRAN compilers - too numerous (and sometimes too buggy) for individual
acknowledgment. It is a sobering fact that our standard test suite (exercising all the
routines in this book) has uncovered compiler bugs in a large majority of the compil-
ers tried. [..]
[..]"

I do not know whether an Intel; Pathscale; NAG; Absoft; or Portland
Group compiler was one of the buggy compilers mentioned in that book.

The aforementioned pioneer of compiler theory used to work for IBM and
he claimed on the pages I explicitly mntioned in the paper I cited
(which perhaps you did not read) that IBM employees disregarded his
advice on how to make good compilers.

Major criticism of C and C++ is attributed on
WWW.ACM.org/crossroads/xrds14-2/franallen.html
to another expert of compiler theory who used to work for IBM, but a
few weeks ago IBM told me that its licensing options would not permit
me to acquire one of its Ada compilers without also acquiring a C++
compiler. I do not want a C++ compiler.

People at Sun certainly have studied how to make compilers.

|------------------------------------------------------------------------|
|"For gfortran I see the number of Fortran 95 bugs constantly decreasing |
|(judging from the bug reports), but it will of cause never be bugfree |
|and there are enough possibilities for bugs when Fortran 2003/2008 |
|features get implemented." |
|------------------------------------------------------------------------|

Congratulations. Writing a grammar for any practical language is much
harder than it would first seem.

|------------------------------------------------------------------------|
|"> Writing compilers without reading about established techniques on how|
|> to do so is not effective. I am aware of someone who had graduated in |
|> computer science who has tried to write a compiler without consulting |
|> the literature, and he did not do well. |
| |
|I want to point out that there are several GCC developers which have |
|studied computer science and do compiler development for living; I |
|pretty sure that a large number of them knows compiler development |
|literature." |
|------------------------------------------------------------------------|

Yes.

|------------------------------------------------------------------------|
|" That number of people knowing compiler literature inside |
|out might be less in gfortran than in other parts of GCC, but is also |
|not zero. [..]" |
|------------------------------------------------------------------------|

I had not known that.

|------------------------------------------------------------------------|
|"> The aforementioned hero of language formalism and Fortran said that |
|> FORTRAN I was good for the beginning of high-level programming but |
|> that it should not be kept as a basis for better languages. |
| |
|As Fortran is now around for 50 years and still moderately widely |
|used, contrary to some other languages, what does this proof?" |
|------------------------------------------------------------------------|

As proofs are impossible, it does not prove anything. It is however
another case to demonstrate that inertia can result in better things
not significantly replacing inferior things. Other examples of popular
things which are inferior to alternatives are C; French; English;
German; and standard rubber bicycle tubes which can easily be
punctured.

Fortran is not "now around for 50 years". Just because you call
Fortran 2008 Fortran, does not mean that it is Fortran. It has been
claimed that a number of FORTRAN I features were not kept in FORTRAN
77. These omissions resulted in a better and different language.

Yours sincerely,
Colin Paul Gloster

Relevant Pages

  • Re: Maples "fd" crash code, and FORTRAN in Maple
    ... Maybe a bit of Fortran lately. ... suggest that Maple contained FORTRAN. ... the Visual C compiler is eliminable. ... 2-3 bugs can be hit per hour. ...
    (sci.math.symbolic)
  • Re: High Performance Fortran compiler?
    ... looking for a freely available HPF compiler. ... the High-Performace Fortran (HPF) ... Both the g95 and gfortran are Open Source Fortan 95 compilers; ...
    (comp.lang.fortran)
  • Re: "actual argument keyword name"
    ... You'd be surprised at the kind of bugs I've found in IVF, ... I don't deny that IVF, like every other compiler on the planet, has ... User communities for Intel Software Development Products ... Intel Fortran Support ...
    (comp.lang.fortran)
  • Re: Installation and usage of free FORTRAN-77 compiler on Windows XP laptop
    ... Objective C is not that popular on Windows and that Java programmers ... will primarily use Sun's compiler. ... Fortran Library: http://www.fortranlib.com ... Support the GNU GFortran Project: http://gcc.gnu.org/fortran/index.html ...
    (comp.lang.fortran)
  • Re: Gfortran 2006 Year End Status Report
    ... (* Something some compiler developers (I don't mean specifically gfortran ... in this case a compiler I can use. ... This is another area where the gfortran development ... distilling and reporting bugs. ...
    (comp.lang.fortran)