Just Wondering

[Stephen Uitti]

In article <12564 at lanl.gov> jlg at lanl.gov (Jim Giles) writes:
> ...case sensitivity discussion...
>Fortunately, for both these cases, C has yet to become a really common
>language in the big computer environment.  Even UNIX based supers mostly
>use Fortran-like languages for production.

One may read this as "common usage is x, therefore x is better".

Supers are becoming UNIX based for the same reason that lots of
new machines are UNIX based - because people can do it.  No one
wants to implement a new OS (other than maybe Big Blue).

People are writting in Fortran, COBOL, or whatever, because that
is what they know.  (Same reason - they can).  This has little to
do with what makes a maintainable system.

I've maintained Fortran code.  It sucks.  Modularity is poor.
Global variables are everywhere.  The common statements sometimes
don't match.  Some programmer wanted to save an assignment by
switching two names.  Pass-By-Name-Only causes all sorts of
strange side effects.  Even the straight forward loops have
labels and goto's.

I've maintained BASIC code.  No parameter passing.  Everything is
global.  One and two character variable names.  Sigh.

I'm not going to say "C is best", even if this is comp.lang.c.
Still, i have seen vectorizing C compilers on Supers.  I've seen
an application get over 150 MFLOPS with it.  I've also seen
scientific types who have programmed in Fortran for the last 30
years use it.  I think the only reason that they switched from
assembler is that the assembler they learned first disappeared
with the machine(s) they first used.  It took years for them to
switch to the "new fangled" Fortran (around 1962?).

If the only well supported scientific language for a
supercomputer was Postscript, you'd find people using it.
You'd even find people who like it.

Stephen.