home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.edu:1335 comp.lang.fortran:3167
- Newsgroups: comp.edu,comp.lang.fortran
- Path: sparky!uunet!wupost!sdd.hp.com!ux1.cso.uiuc.edu!m.cs.uiuc.edu!vela!pd-dom!schuette
- From: schuette@wl.com (Wade Schuette)
- Subject: Re: scientists as programmers
- Message-ID: <1992Aug26.022458.12622@wl.com>
- Keywords: supercomputer Fortran self-modifying-code AI Japanese Supercomputing
- Organization: Warner Lambert / Parke-Davis
- References: <1992Aug25.034553.2990@linus.mitre.org> <1992Aug25.154501.8654@colorado.edu> <1992Aug25.202307.12365@newshost.lanl.gov>
- Date: Wed, 26 Aug 1992 02:24:58 GMT
- Lines: 54
-
- In article <1992Aug25.202307.12365@newshost.lanl.gov> jlg@cochiti.lanl.gov (Jim Giles) writes:
- >In article <1992Aug25.154501.8654@colorado.edu>, ejh@khonshu.colorado.edu (Edward J. Hartnett) writes:
- >|> [...] No offense to scientists, but I have rarely if ever seen a
- >|> scientist who was a good programmer. [...]
- >
- >You must encounter a different set of scientists than I do. Most of
- >the ones who develop code at all - that I deal with - are among the most
- >talented programmers I've ever met. Possibly this is not common at the
- >*.edu sites.
- >
- >To be sure, it is rare for a language to be designed with an eye toward
- >making the scientist's job easier....
-
-
- I'd pick up on that last point with a tangent on supercomputer and
- parallel computer compilers and languages, including Fortran.
- Japan apparently is far ahead of the US in terms of making complex
- technology friendly to mortals. This may have changed (?), but as
- of a few years ago the percent of Japanese scientists who used
- supercomputers was WAY ahead of the percent in the US, and the
- reasons seemed to be that the Japanese didn't expect the _SCIENTIST_
- to have to learn how to parallelize his/her code, as the compiler
- did that. So the physicist or biologist could work on their work,
- not spend weeks in sessions on learning how to write paralle fortran.
-
-
-
-
- Has that changed?
-
- For that matter, does anyone know if the current supercomputer
- compilers are bright enough to watch their own output actually run
- and learn from it so on their NEXT pass they can stop guessing and
- make better decisions as to how to optimize the data-dependent sections
- of code? Or are humans still supposed to do THAT as well? I never
- could understand why code intended to be run ten million times couldn't
- watch its own work a little and adjust bad decisions made in earlier
- compiles. Sometimes it wouldn't converge, but I'd think a lot of
- programs run almost identical jobs (just as poorly) over and over and
- over on almost identical data sets.
-
- Seems like a small amount of applied AI could have a large impact on
- such code. Also seems that the supercomputer is the only thing around
- smart enough to watch itself and handle this adjustment, and not only
- no scientist outside the field, but almost no human inside the field
- could track the algorithms to do this by hand after the fact.
-
-
-
-
- Comments?
- --
- ==================================================================
- R. Wade Schuette schuette@wl.com Ann Arbor, MI, USA
-