Newsgroups: comp.sys.next
Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!batcomputer!theory.TC.Cornell.EDU!finn
From: finn@theory.TC.Cornell.EDU (Lee Samuel Finn)
Subject: Re: Fortran on the Next platform
Message-ID: <1991May27.150959.18050@batcomputer.tn.cornell.edu>
Sender: news@batcomputer.tn.cornell.edu
Nntp-Posting-Host: theory.tc.cornell.edu
Organization: Cornell Theory Center
References: <SIMMONS.91May24173142@rigel.neep.wisc.edu> <1991May25.003520.20433@math.ucla.edu> <SIMMONS.91May26163744@rigel.neep.wisc.edu>
Distribution: na
Date: Mon, 27 May 1991 15:09:59 GMT

In article <SIMMONS.91May26163744@rigel.neep.wisc.edu> simmons@rigel.neep.wisc.edu (Kim Simmons) writes:
>In article <1991May25.003520.20433@math.ucla.edu> barry@pico.math.ucla.edu (Barry Merriman) writes:
>
>   Path: doug.cae.wisc.edu!zazen!wuarchive!zaphod.mps.ohio-state.edu!usc!ucla-cs!ucla-ma!pico!barry
>   From: barry@pico.math.ucla.edu (Barry Merriman)
>   Newsgroups: comp.sys.next
>   Date: 25 May 91 00:35:20 GMT
>   References: <gzght5+@rpi.edu> <SIMMONS.91May24173142@rigel.neep.wisc.edu>
>   Sender: news@math.ucla.edu
>   Organization: UCLA Dept. of Math, UCLA Inst. for Fusion and Plasma Research
>   Lines: 19
>
>   In article <SIMMONS.91May24173142@rigel.neep.wisc.edu> simmons@rigel.neep.wisc.edu (Kim Simmons) writes:
>   >
>   >Would be nice if NeXT supported FORTRAN like Sun and Dec and IBM and HP does.
>   >This could easily be done via some support for the gnu FORTRAN project.
>   >However, i guess their lack of support for FORTRAN is consistent with their
>   >lack of interest in the technical workstation market.
>
>   Well, Sun has un-bundled their fortran compiler; you have 
>   to buy it now. I doubt many individuals could afford that, either.
>   In general, unbundling compilers seems to be a trend for
>   computer manufactures these days (another e.g.: Connection Machine).
>
>
>
>   --
>   Barry Merriman
>   UCLA Dept. of Math
>   UCLA Inst. for Fusion and Plasma Research
>   barry@math.ucla.edu (Internet)   barry@arnold.math.ucla.edu (NeXTMail)
>
>
>
>True but,
>(1) Sun has a FORTRAN compiler, NeXT does not.
>
>(2) Sun has always and continue to give us their FORTRAN compiler, without us
>having to pay for it (so does DEC by the way).
>
>(3) No students or individuals have Sparcstations around here that i know
>i.e. personal purchase as opposed to say University Departmental purchase.
>
>(4) Because Sun does it does not mean that NeXT should do it, even if they
>had a FORTRAN compiler to do it with.
>--
>===============================================================================
>    Internet:      simmons@rigel.neep.wisc.edu
>    Othernet:      simmons@hoofers.lake.mendota
>--- ----------------------------------------------------------------------- ---
>According to the HitchHikers guide to the galaxy, the one thing we
> *cannot* afford to have is a sense of perspective.
>===============================================================================

Someone at your University is paying for that FORTRAN compiler; Sun
has required payment for it since 1988 (when 4.0 was introduced). 

[Parenthetically, Suns FORTRAN compiler sucks eggs: it's Awful with a
capital A. It generates very inefficient code for math (and that's
optimized), the inline code templates have errors, the i/o routines
are error laden, and the system calls are buggy as well. If you want
to call that a FORTRAN compiler, you're welcome to.]

Incidentally, Sun is unbundling C as well: with 5.0, there will be no
C of any kind unless you pay extra; if you want a C that optimizes,
you will pay more, and if you want an ANSI C, you will pay more yet.

With FORTRAN (and now C), Sun has claimed that by charging for them as
an extra, they can seperately fund a compiler development group to
make the compilers better. IMHO, it hasn't worked with FORTRAN: I've
been using Sun FORTRAN since 84, and it has not changed at all (it's
still got many of the same bugs now as it did then, and every new
release seems to re-introduce bugs that were present in the old
release but fixed by interim patch tapes). 

Now, as for FORTRAN on the NeXT: I've had good experience with the
Absoft FORTRAN compiler. The 2.0 version was very buggy, but they
responded promptly and effectively to my bug reports, and the 3.0
version is a very nice product. The only problem I've run across with
it is that it uses some of the '030 opcodes that are emulated in
software on the '030, and this can slow it down on some operations. 

I've less experience with f2c, except to say that it works. Generally,
since it is really a translator, it will work only as well as your C
compiler, which on the NeXT is pretty good. 

The FSF is working on a FORTRAN compiler (really a frontend that
generates the intermediate code that the engine of gcc turns into
machine language); I've no idea when that will be available.

