home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.edu:1503 comp.lang.fortran:3399 comp.lang.misc:2909
- Path: sparky!uunet!wupost!cs.utexas.edu!uwm.edu!ogicse!unmvax!cs.sandia.gov!mccurley
- From: mccurley@cs.sandia.gov (Kevin McCurley)
- Newsgroups: comp.edu,comp.lang.fortran,comp.lang.misc
- Subject: Re: Scientists as Programmers (was Re: Small Language Wanted)
- Message-ID: <1992Sep3.021822.14559@cs.sandia.gov>
- Date: 3 Sep 92 02:18:22 GMT
- Article-I.D.: cs.1992Sep3.021822.14559
- References: <92245.124125KGZXK@ASUACAD.BITNET>> <MEISSNER.92Sep120283 <9224622.17597@mulga.cs.mu.OZ.AU>
- Sender: usenet@cs.sandia.gov (Another name for news)
- Organization: Sandia National Laboratories, Albuquerque, NM
- Lines: 40
-
- In article <9224622.17597@mulga.cs.mu.OZ.AU> fjh@munta.cs.mu.OZ.AU (Fergus James HENDERSON) writes:
- >>C++.
- >
- >Exactly. What's more, with C++ I can have arbitrary precision integers,
- >arbitrary precision rational numbers, fixed point real numbers, etc.
- >(all of these are available using GNU's libg++). If you want
- >quaternions, then it's no big deal to add them too.
-
- I am getting pretty tired of this thread, but there are lessons to be
- learned. I personally do my programming in C because it allows me to
- get close to the silicon and manipulate data structures at high speed
- as is needed for my branch of science. Others in my lab use Fortran
- for several good reasons:
-
- 1. the existing code in their area of science is worth leveraging.
- 2. Fortran manipulates integers, floats, and arrays. Coincidentally
- these are the same things that silicon usually manipulate, and the
- thing that scientists most often need to compute with. This makes it
- easier for the programmer to write efficient code (from a runtime
- viewpoint).
- 3. It provides some degree of portability between suns, sgis, crays,
- and intel and ncube parallel machines.
-
- We have access to the fastest computers (serial, parallel, and
- vector), but the solution is not to just get faster processors to make
- up for sloppy code. One of the groups that consumes the largest
- amount of Cray time here at Sandia decided to switch to C++ to gain an
- advantage in code maintainability, but this change is not for free.
- One cost of C++ is that it hides the operations performed by Silicon
- (a good example is the problem of temporary copies). Another lies in
- the fact that good (read efficient) C++ compilers are hard to come by.
-
- I am always amused by the language wars. It reminds me of the car
- salesman that tries to sell me a two door because that is what he
- sells, rather than what I need. In the end the marketplace will
- decide this issue, and I predict that Fortran, C, C++, and even Cobol
- will find some prosperity in the future.
-
- Kevin McCurley
- Sandia National Laboratories
-