Sujet : Re: xkcd: CrowdStrike
De : kludge (at) *nospam* panix.com (Scott Dorsey)
Groupes : rec.arts.sf.written rec.arts.comics.stripsDate : 25. Jul 2024, 20:52:03
Autres entêtes
Organisation : Former users of Netcom shell (1989-2000)
Message-ID : <v7uad3$bev$1@panix2.panix.com>
References : 1 2 3 4
Lynn McGuire <
lynnmcguire5@gmail.com> wrote:
>
The real problem is that Fortran changed significantly from F66 / F77 to
F90 and beyond. I have written my own program to do most of the
upgrades for me but I am subject to the old 80 / 20 rule. It is easy to
automate 80% of the work but the last 20% is dadgum hard to automate. I
am also cleaning up some old code from the 1970s that is problematic.
Have you considered gnu fortran? It doesn't produce as fast executables
as the Intel compiler sometimes, but it's pretty good and it has an f77
mode.
f90 brings some very cool stuff for matrix operations, which makes
autoparallelization a lot easier, but on the other hand I don't think
engineers should be allowed to use pointers.
And hollerith fields have to go.
I started off porting my F66 / F77 code to C++ using a very modified
version of F2C. Due to the complexity of input and output between the
two languages (Fortran is record oriented, C is byte oriented), I have
split the project into two parts as my customers need a x64 version of
my software.
The original g77 was just f2c in front of gcc and... it was pretty awful
really, and never did work all that well. Modern gfortran is much better.
--scott
-- "C'est un Nagra. C'est suisse, et tres, tres precis."