Re: Praise for Gfortran (finally)

Al and Damian,

Many thanks for the praise - every little bit helps:-)

Over the last few years I have developed a 30,000 line Fortran-95
engineering application using simultaneously several compilers (XLF,
LF95, G95, and IVF). During that time I toyed with Gfortran and put
up with the 2 steps forward, 1 step back of each binary build.

We are sharply aware of this problem. It has come about largely
because of the inadequacy, in the past, of our testsuite. Not having
a take-the-standard-and-write-a-testsuite yardstick to go by, ours has
grown as we have fixed bugs and regressions. Thus, our random walk
through bugspace is, at least Markovian; for the main part they do
not get repeated. Also, whilst some of the things that people try to
do are "ïnteresting", the space in which we can wander seems to be
bounded too.

However, I'm now happy to report my "last" issue with it has been
resolved and more importantly performance is almost identical to XLF
and IVF in most cases. Congratulations to all involved with this
project!


Thanks!


I'll second that. I've used roughly half a dozen compilers over the
past five years. Over the past year, gfortran really seems to be
raging ahead of a lot of the competition in features. I really hope
the Gnu Compiler Collection will ultimately assume the leadership role
in the Fortran community that it played in the C++ community a decade
ago. It's great to see volunteers accomplishing what commercial
developers can't or won't accomplish in the same amount of time.

I'm not entirely convinced that were are "raging ahead" but we are
trying. It has been difficult, at times, to get right the balance
between new features, both non-standard and F200x, and ensuring that
gfortran is an effective, standard-compliant F95 compiler.
Astonishingly, in spite of the sometimes anarchic tendencies of the
gfortran maintainers collective, we do seem to go in roughly the right
direction. For example, we have done the ground work for CLASSes and
OOP during 2008, which should be realised in gcc-4.5.

The constraints on commercial compiler developers are very different
to those on a volunteer group. In essence, they cannot afford the one
step back for every two forward; therefore, they take one little step
at a time:-) I suspect that we will arrive at our respective
destinations simultaneously....

One feature of the gfortran maintainers group is that there is a
fairly continuous turn-over, with a characteristic time of a few
years. Thus, we are always on the lookout for new blood. If anybody
out there is interested in joining in, please get in touch.

Thanks again for the kind remarks.

Paul



.

Relevant Pages

  • Re: Praise for Gfortran (finally)
    ... LF95, G95, and IVF). ... During that time I toyed with Gfortran and put ... resolved and more importantly performance is almost identical to XLF ... gfortran is an effective, standard-compliant F95 compiler. ...
    (comp.lang.fortran)
  • Re: gfortran 2007 annual report
    ... On Windows, you probably mean. ... I use gfortran as a native 64 bit ... compiler on GNU/Linux AMD64 now for about 2.5 years. ... but that is pretty common in old fortran source. ...
    (comp.lang.fortran)
  • Re: Gfortran 2006 Year End Status Report
    ... first time gfortran was able to compile the CASTEP ab-initio materials modelling ... Is CASTEP mostly F77 code? ... There's not much use of features in F95 but not F90, ... in this case a compiler I can use. ...
    (comp.lang.fortran)
  • Re: gfortran 2007 annual report
    ... I use gfortran as a native 64 bit ... compiler on GNU/Linux AMD64 now for about 2.5 years. ... I downloaded the latest Cygwin build ... While I'm not trying to disparage an honorable effort, ...
    (comp.lang.fortran)
  • Re: Problem with log function - Intel fortran compiler under Linux
    ... compiler under Linux ... Linux: Red Hat Enterprise Linux, ... Now there are still NaNs appearing, ... I use gfortran to compile the program and that causes ...
    (comp.lang.fortran)