home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.edu:1409 comp.lang.fortran:3282 comp.lang.misc:2805 sci.math:10699
- Path: sparky!uunet!van-bc!rsoft!agate!ames!pacbell.com!network.ucsd.edu!lyapunov.ucsd.edu!mbk
- From: mbk@lyapunov.ucsd.edu (Matt Kennel)
- Newsgroups: comp.edu,comp.lang.fortran,comp.lang.misc,sci.math
- Subject: Re: Scientists as Programmers (was Re: Small Language Wanted)
- Date: 29 Aug 1992 22:55:05 GMT
- Organization: Institute For Nonlinear Science, UCSD
- Lines: 68
- Message-ID: <17ov89INNdls@network.ucsd.edu>
- References: <17opl4INN5gb@roundup.crhc.uiuc.edu>
- NNTP-Posting-Host: lyapunov.ucsd.edu
- X-Newsreader: Tin 1.1 PL3
-
- hougen@uirvlh.csl.uiuc.edu (Darrell Roy Hougen) writes:
- : mbk@lyapunov.ucsd.edu (Matt Kennel) writes:
- :
- : >I personally would love a "Numerical Eiffel" type of language ...
- :
- : Try "Sather". It is an object-oriented programming language that is
- : competitive with Fortran in speed for complex problems. Of course,
- : the object oriented dispatch makes it slightly slower for some types
- : of problems, but the designers have gone to great lengths to make it
- : fast. The principle designer, Steven Omohundro, is a physics PhD
- : turned computer scientist, so that may explain why he is interested in
- : bridging the gap, as it were.
-
- I've downloaded and compiled Sather already, and I really like what I've
- seen so far. After seeing Sather, I can't understand how C++ could be so
- needlessly complicated. I'm just about ready to switch, but I worry about
- not having my code usable by anybody else.
-
- What I like:
-
- 1) Simple class structure and rules (though I think the CS gurus have some
- problems which I don't completely understand).
-
- 2) Garbage collection.
-
- A couple of things tho:
-
- 1) There is only one basic "loop" construct, which means that you have to
- write i:=i+1 all the time just do do the most basic things. Of course
- it's only "surface syntax", but these things make a real difference
- in real life. In the jihad for simplicity, Oberon discarded the
- FOR i:=1 TO N DO loop, but it was put back into Oberon-2 and I think
- Sather's designers should do the same.
-
- 2) Along the same lines what I really meant by "Numerical Eiffel" was array
- index notation as I've described before, or something to get rid of all
- those silly loops. Perhaps this is taken care of as their parallel
- extension?
-
- 3) The library right now is too bare to be completely useful yet, even though
- the README spells out great plans (better for my purposes than ANYTHING
- I've ever seen. K-D trees!). No complex numbers yet. Can't read binary
- (big blocks of FP numbers) yet.
-
- Implementation issues:
-
- 1) It compiles into "C", and the 2-d arrays are implemented in
- Numerical Recipes style (pointers to arrays of pointers). It's not clear
- this is best for speed, especially with newer compilers I guess.
- Depends on the relative cost of reading memory vs computation.
-
- 2) It doesn't yet output code in such a way that little methods can
- be inlined, say by compiling with GNU CC. If this were to happen, I'd
- think that Sather programs could be quite a bit *faster* than C++, as it
- generates new, specialized, code for classes with type parameters.
- (I.e LIST{INT} gets special treatment vs. LIST{FOOBAR})
-
- : Sather is available from Berkeley via anonymous ftp. However, I can't
- : seem to find the ftp site at the moment. The author's email address
- : is om@icsi.berkeley.edu. Be sure to mention Sather in the subject.
- :
- : Darrell
-
- --
- -Matt Kennel mbk@inls1.ucsd.edu
- -Institute for Nonlinear Science, University of California, San Diego
- -*** AD: Archive for nonlinear dynamics papers & programs: FTP to
- -*** lyapunov.ucsd.edu, username "anonymous".
-