Re: HPC Considered Harmful
- From: e p chandler <epc8@xxxxxxxx>
- Date: Thu, 29 May 2008 22:25:22 -0700 (PDT)
On May 28, 6:58 pm, Bil Kleb <Bil.Kleb-use...@xxxxxxxx> wrote:
Greg Wilson's interview is fun:
http://itc.conversationsnetwork.org/shows/detail3682.html
Regards,
--http://www.linkedin.com/in/bilkleb
Interesting talk. Not just about HPC but about scientific computing in
general. Points I liked:
1. Don't expect good scientists to be as good at programming or
computing. Helps to explain many nasty Fortran programs I've seen!
2. Productivity often beats performance. Long long ago, I had two
versions of a simulation program. One in Fortran on a Univac, one in
Basic on an Apple II+. Guess which made ME more productive?
2a. Excel was claimed to be used most often for number crunching
followed by things like Matlab and Mathematica. I found that an
engineer's web site which discueesd numerical techniques noted that he
had only used Matlab in school but NEVER Fortran.
3. In science, proving work requires replicating previous results.
Could I possibly reproduce my old simulations? Let's see - decks of
cards, program listings, program output, and one large mainframe -
good luck finding any of these.
4. Important to be able to maintain programs. Again, how much yucky
code do we see posted here? How much time do we spend trying to help
people to interface to existing library routines?
4a. As computers and programming languages evolve, things become
obsolete and the knowledge to use or understand them disappears. Who
will understand an arithmetic IF or an assigned GOTO 100 years from
now? Who will be able to make any sense of a program listing of
"Adventure" written for the PDP-10? Or MS-DOS INT 21H calls, etc.?
4b. IMO commercial and scientific computing do have some things in
common in terms of longevity of code. Again I remember (alas) some
very old and moldy code posted here (obviously from textbooks) about
things like electric fields and internal combustion.
4c. In other forums others often trash older languages like Fortran
and COBOL without realizing how these languages have evolved. At least
the tools are getting better. I really don't want to write upper case,
fixed format, Fortran 77 or earlier code ever again.....
- e
.
- References:
- HPC Considered Harmful
- From: Bil Kleb
- HPC Considered Harmful
- Prev by Date: Re: Fortran 77 alphabet character $
- Next by Date: Re: Is this legal?
- Previous by thread: HPC Considered Harmful
- Next by thread: Slick or Slack
- Index(es):
Relevant Pages
|
