CCL Home Page
Up Directory CCL pentium
From jle@world.std.com Tue Jan 10 21:26:22 1995
Received: from europe.std.com  for jle@world.std.com
	by www.ccl.net (8.6.9/930601.1506) id VAA25126; Tue, 10 Jan 1995 21:26:11 -0500
Received: from world.std.com by europe.std.com (8.6.8.1/Spike-8-1.0)
	id VAA05591; Tue, 10 Jan 1995 21:26:07 -0500
Received: by world.std.com (5.65c/Spike-2.0)
	id AA12045; Tue, 10 Jan 1995 21:26:27 -0500
Date: Tue, 10 Jan 1995 21:26:27 -0500
From: jle@world.std.com (Joe M Leonard)
Message-Id: <199501110226.AA12045@world.std.com>
To: jkl@ccl.net
Subject: PC summary
Status: RO

Alan Wolfson (awolfson@msi.com) writes:

    For the most part, the cooperative multi-tasking takes place
without you having to worry about it.  Programming Microsoft Windows
is similar to programming in the X Window System, but in some ways is
much easier.  The main part of a Window application is the processing
of "Windows messages."  These messages deal with mouse and keyboard
events, exposure events, etc.  All such messages originate from within
the bowels of Windows.  The Windows kernel collects all such events,
and parcels them out to the active applications to handle their own
messages.  During such internal processing is when the cooperation
kicks in.

    If you write a computationally intensive application that doesn't
allow the message handling to occur on a regular basis, then it is not
being "cooperative" in that it doesn't allow other applications to get
their messages for processing.  Of course, if all you want to do as a
single user of a Windows machine is to grind out numbers, then it is
permissible to write a cpu hogging application.  The real problem
comes when you also want to use your spreadsheet or word processor
while the calculations are happening.

    I agree with your assessment of non-cooperative applications and
the response to them by their users.  I have no experience with either
Linux nor OS/2, but have used NT.  People I work with have given very
favorable remarks about both OS/2 and Linux.  All three os's are well
worth looking into as viable options to vanilla Windows.  Another
thing to keep in mind is Windows 95 (96?).  It is supposed to have
preemptive multitasking when (if) it's released.  An additional
advantage of Windows 95 is that it too will use the Win32 API.  You
can therefore write your application now under either Windows or NT,
and it is supposed to run without source modifications under Windows
95 as long as you adhere to the win32 API.

David J. Heisterberg (djh@ccl.net) writes: 

>1) Is there a way of telling whether the chip in a pentium PC is a "good"

No way to tell except by testing -- the external markings haven't
been changed.

>2) Have folks gotten linux running on Pentiums - or are they merely

There's only a few new instructions.  The thing that makes the
difference in performance is proper code ordering.  I know I've
heard about pentium specific optimizations in gcc, but as I don't
have one, I haven't looked too carefully.

>3) Are there graphics cards that should be used and/or avoided (in

Diamond used to be a real hard-ass about letting technical information
out.  Last I heard they've relented.  Everybody has windows drivers,
so that's not a real worry, and about every other issue of Computer
Shopper has a review of graphics boards.  With linux, the issue is
XFree86, and they support most name boards.

>4) Have people used CD-ROM's that are external and swappable between
>PC's and Mac's (obviously a SCSI device)?  Any comments on who's

I "borrow" a Sun CD from work on occasion and use it on my home
machine running windows or linux.  I don't know about Mac CDs.

>Yes, I'm rather used to workstations...

Linux it is, then!

>7) Finally are there PC vendors that I should use and or avoid?

Avoid Packard Bell.  But they're targeted towards the home game-
playing crowd anyway.  Most people seem happy with Micron and
Dell.

Song Ling (sling@euclid.chem.washington.edu) writes:

Only one things I know: There are Fortran compilers for Macs because
I ran some fortran codes on some old Macs years ago, I think the
company name is Absoft, sorry but I don't have either phone number
or e-mail address of it.

By the way, are you sure you would want to switch to a PC?

David C. Doherty (doherty@msc.edu) writes:

>5) Are there fortran compilers for the Mac (There seem to be several
>good ones for the PC)?  If so, who are they, and do they have
>anything resembling the tools on a workstation?  I'll also take >recommendations for C compilers for the Mac...

Language Systems Fortran (MPW) is very good, power mac native in beta.

Absoft will soon be OEMming a F90 compiler from CraySoft (a Cray Research
subsidiary).  Basically the same compiler as on Cray machines, Absoft does
the interface.

Conventional wisdom favors Metrowerks CodeWarrior for C.
>
>6) Are the Windows development environments "good enough" to keep one
>from having to reboot the machine when the code being developed
>crashes (possibly stupid question #106)?  How about on the Mac?
>

MPW is reasonably stable, but don't expect much until these machines
(Macs and PC OSes) sport protected memory.

Thomas Nhan (tnhan@www.chem.washington.edu) writes: 

To preface my comments: I also upgraded my home computer as well.
 I'm running the Linux Slackware Distribution on my Pentium-90
 at home.  Besides the text interface of Un*x, I also have X11R6
 (X windows) running as well.

> 1) Is there a way of telling whether the chip in a pentium PC is a "good"
> or "bad" chip (like the old double-sigma's)?  I've seen posts about
> tests that can be run, but is there any before-purchase things that
> can be done?

I doubt that you can.  If you were to purchase a computer in the store,
 bring that program and check it yourself.  But if you are going to
 purchase from a mail-order company, you might want to wait until
 early summer or late spring, when most of the buggy Pentium are sold :)
 However, Intel is willing to replace any pentium with that FDIV bug
 on demand.

> 2) Have folks gotten linux running on Pentiums - or are they merely
> running 486 code that doesn't know or care the difference?  I assume
> there are different instructions in a Pentium, or am I full of it
> with this?

At the moment, the Linux code _is_ optimized to run on i386 and i486.
 But it _can_ be optimized to i586; since the optimization "-mpentium"
 (within gcc, GNU's C-compiler) is rather new, you would expect some bugs.
 That's why I built my Linux kernel using the i486 optimization.
 Note that the optimization for pentium is rather incomplete because
 Intel is not willing to make available informations that are needed
 because it fears pentium-clone will run faster than the true pentium :)

> 3) Are there graphics cards that should be used and/or avoided (in
> Dos/Windows and in linux)?  Several vendors mention Diamond Stealth
> cards, for example - do these work with Linux? Win3.1? etc...

If you are going to set up Linux, I would suggest that you read the
 FAQs for Linux.  They can be found on "sunsite.unc.edu:/pub/Linux";
 this is a major depository for Linux-software.  There should be
 a documentation that tell you all the monitor and graphics card
 that can run X windows in Linux.

> 4) Have people used CD-ROM's that are external and swappable between
> PC's and Mac's (obviously a SCSI device)?  Any comments on who's
> to use?

I'm not knowledgeable in this area as I only have an ATAPI (IDE)
 NEC CD-ROM.

> 5) Are there fortran compilers for the Mac (There seem to be several
> good ones for the PC)?  If so, who are they, and do they have
> anything resembling the tools on a workstation?  I'll also take >recommendations for C compilers for the Mac...

I have never developed program on a Mac...

> 6) Are the Windows development environments "good enough" to keep one
> from having to reboot the machine when the code being developed
> crashes (possibly stupid question #106)?  How about on the Mac?

I have never developed programs in Windows either.  Thank goodness.

> Yes, I'm rather used to workstations...

Me too :)

> 7) Finally are there PC vendors that I should use and or avoid?

I bought my pentium from Gateway2000.  I am quite satisfied with my
 computer's performance.  But I would not recommend it because:
 1. It always take me at least 1 week to reach the Technical Support people;
    but they are not that knowledgeable (in terms of hardware setup
    that you may need for running Linux) either.
 2. There are very little documentation that comes with the hardware.
 3. I had my power supply and keyboard replaced... this isn't an isolated
    problem as ppl I know also had hardware problems.  (perhaps due to
    shipment?)

> Thanks for your comments (well, the comments I expect to get :-)!
> I'd love to get a big SGI or IBM at home, but I'd rather not be killed
> in my sleep...

You are quite welcome :)  I'd love to have an SGI at home.  But the
 Indy is too slow (for true-3D graphics) and the Indigos are too expensive.
 If you are going to get a Pentium, make sure it have _at_least_ 16 Meg RAM.

Pedro A M Vazquez (vazquez@iqm.unicamp.br) writes:

> 1) Is there a way of telling whether the chip in a pentium PC is a "good"
> or "bad" chip (like the old double-sigma's)?  I've seen posts about
> tests that can be run, but is there any before-purchase things that
> can be done?

No, unless Intel change its mind and label the good chips you'll need
to run a piece of code to test it.

>
> 2) Have folks gotten linux running on Pentiums - or are they merely
> running 486 code that doesn't know or care the difference?  I assume
> there are different instructions in a Pentium, or am I full of it
> with this?
>

There are 3 good Unix like OS for x86 PCs: Linux, FreeBSD and NetBSD,
I'm using FreeBSD at home and run a database on Linux.  All these systems
are distributed in a way they can boot in 386, 486 or Pentium based PCs, 
after you install them you can rebuild the performance critical components 
(kernel, applications,etc) using gcc with the pentium flag.  I think a
kernel rebuild is all you'll need to do, the Xserver is fast enough.

> 3) Are there graphics cards that should be used and/or avoided (in
> Dos/Windows and in linux)?  Several vendors mention Diamond Stealth
> cards, for example - do these work with Linux? Win3.1? etc...
>

Diamond Stealth cards were not supported only under XFree86, the X11
free version for all three systems. This was due Diamond not opening
details of internal clock programming for this card. It seems this has
changed in the last weeks and a Diamond Stealth based Xserver may be
available on future. I can send you a list of supported cards.

> 4) Have people used CD-ROM's that are external and swappable between
> PC's and Mac's (obviously a SCSI device)?  Any comments on who's
> to use?

I've not a Mac with SCSI adapter to test but I'm using all external
SCSI devices from our Suns/RS6k (tape driver, CD, hard discs) on the 
FreeBSD machines.

>
> 5) Are there fortran compilers for the Mac (There seem to be several
> good ones for the PC)?  If so, who are they, and do they have
> anything resembling the tools on a workstation? I'll also take
> recommendations for C compilers for the Mac...

Skipping, I don't know

>
> 6) Are the Windows development environments "good enough" to keep one
> from having to reboot the machine when the code being developed
> crashes (possibly stupid question #106)?  How about on the Mac?
>

Well, under any of the 3 systems above you're under a full X11R6 graphics
environment with OpenLook and Motif (you must buy Motif) and under all
protection you'll find under SunOS/AIX/Irix, etc with access to all
programming tools from a typical Unix machine.

> Yes, I'm rather used to workstations...
Me too.

You'll notice no difference, we have ported a lot of programs from Unix
to FreeBSD for our comp chem teaching lab (mopac, gamess, rasmol, babel,
mm2, etc)

>
> 7) Finally are there PC vendors that I should use and or avoid?
>
I'm out of US, all US PC vendors I know of are Gateway, IBM, DEC, no
problems

> Thanks for your comments (well, the comments I expect to get :-)!
> I'd love to get a big SGI or IBM at home, but I'd rather not be killed
> in my sleep...

If you decide on Pentium try out all 3 systems above, they are all very
good and free. I like FreeBSD because it is more SunOS/OSF style than
Linux but both are great.

Matt Stahl (matt@synthesis.chem.arizona.edu) writes:

> 5) Are there fortran compilers for the Mac (There seem to be several
> good ones for the PC)?  If so, who are they, and do they have
> anything resembling the tools on a workstation?  I'll also take 
> recommendations for C compilers for the Mac...

If you look at the ad's in MacTech magazine you will find quite a few
fortran compilers.  I've never used any.  If you are looking for a good
C/C++ development environment, I think metrowerks codewarrior beats
symantec's Think C hands down.

> 6) Are the Windows development environments "good enough" to keep one
> from having to reboot the machine when the code being developed
> crashes (possibly stupid question #106)?  How about on the Mac?

        If you have MacsBug and Discipline (free and on the net) installed in
your mac system, the chances of avoiding a hard crash are very good.
        BTW, the program called QC from Onyx technology is a very nice mac
tool for stress testing applications.  It makes really nasty bugs much
more reproducible.

Igor Uporov (iuporov@dad.bgsu.edu) writes:

Joe - you can find all information about supported hardware for Linux,
FreeBSD on ftp server freebsd.cdrom.com in directory pub. I'm considering
this question also. What I've learned for sure that:
1. Linux is more flexible then BSD (I mean the circle of supported
hardware much bigger then FreeBSD). You also can run Windows 3.1 appli-
cation directly from Linux.
2. Hardware requirement are softer to run X-Windows under Linux then
under FreeBSD. However most peoples express opinion that 16 Mb RAM (for
PC!!!) would be perfect for running X-Windows.
3. About MS Windows. I believe that all vendors of hardware should sup-
port MS Windows otherwise they have to manufacture the hardware for Mac
no choice.
I can also recommend you to take look at http://www.fintronic.com/linux/catalog.
This company ship PC with installed MS DOS , OS/2 and Linux. They guarantee
that everything will work. Prices little bit high for this service. You
also  support their systems.

      That's all what I know for sure concerning your questions.
                             Good luck. Igor.

Song Ling (sling@euclid.chem.washington.edu) writes:

I am wondering if you can buy a sparc with something like
"internal price", I know that the price of sparc classic 50MHz
microSPARC with 15" color monitor, 16MB RAM, 207MB hard drive
in edu. sources is around $2,500 last year, and I expect it
to continue go down as Apple slashes its prices.
The problem may be that the models are slightly upgraded while
the prices heavily upgraded, I guess the best buys would be
right before the newer models are rolled out.  I'd love to
hear about your opinion.
Song

Randal R. Ketchem (rrk@magnet.fsu.edu) writes:

>5) Are there fortran compilers for the Mac (There seem to be several
>good ones for the PC)?  If so, who are they, and do they have
>anything resembling the tools on a workstation?  I'll also take recommendations
>for C compilers for the Mac...

I just purchased Metrowerks CodeWarrior Gold for the Mac. It includes C,
C++ and Pascal compilers. It also has tools for developing Mac apps, if you
desire. I generally use sgi for my computational work, but CW is great. It
will compile 68K and PPC apps. MW offers an academic discount. Gold is $99
and comes on a CD.

Yes, it only comes on CD. It includes lots of tools and some 3000 pages (95
meg) of documentation. They offer multiple versions at different prices. I
believe that you can reach them at:
support@metrowerks.com
1-800-247-6553 (distributor)
There is a CW newsgroup:
comp.sys.mac.programmer.codewarrior
MW employees contribute constantly to the newsgroup.

Let me know if you need more info.

Paul Verwer (verwer@organon.akzonobel.nl) writes:

> 3) Are there graphics cards that should be used and/or avoided (in
> Dos/Windows and in linux)?  Several vendors mention Diamond Stealth
> cards, for example - do these work with Linux? Win3.1? etc...

As far as linux is concerned: it seems support for graphics cards is
still growing, so it is hard to get an up-to-date answer from just a few
people. Fortunately there is a list of linux-compatible hardware
available, e.g. at sunsite.unc.edu, /pub/Linux/docs/HOWTO/Hardware-HOWTO,
or via Mosaic: http://sunsite.unc.edu/mdw/linux.html.

Per-Ola Norrby (peon@medchem.dfh.dk) writes:

        I can at least answer a few of your questions, but only pertaining
to the Mac.  There ARE good compilers, both FORTRAN and C.  If you get the
Macintosh Programmers Workshop (MPW), you can run several compilers under
this shell, and link mixed codes.  From APDA, you can at least get C,
Pascal, and assembler, from Language Systems you get FORTRAN.  I think LS
has also taken over the Pascal nowadays.  The environment is "Unix-like"
(NOT Unix, but with a similar feeling).  As far as I know, for PowerMac you
can only get native code from the C compiler so far, but LS have
beta-versions of FORTRAN and Pascal and will release full versions "soon"
:-)
        About crashes, you can naturally always make a bad crash that
forces a reboot (I've done a lot of those in C...), but in most cases you
get back to the compiler, sometimes even with a not too cryptic error
message.  There are plenty of tools, like source debuggers and source
checking/formatting.  You can also add your own tools easily, there is a
lot of public domain tools at most Internet Mac sites.
        If you only want C, one of the most popular is THINK C, it's
probably the most user-friendly, but since you are used to workstations,
you should get along well with MPW.
        It seems from your second question you haven't done too much Mac
programming?  Well, you ARE in for a shock.  To get any kind of mileage out
of the Mac (ie, not just a simple terminal window) you have to learn all
the window and menu managers by heart.  Cheer up, it's just about 1000+
pages of fine print (several volumes of "Inside Macintosh").  You
especially need the "event manager" that lets you take care of mouse and
keyboard events.  Incidentally, whenever you ask for one of those events with
a high level command, the system checks if another task wants to take over.
If you have compute-intensive tasks, you should sprinkle some
"waitnextevent" calls through it to avoid locking the Mac.  As a user, I
find it very irritating with programs that don't do that.  I guess you
could design your own "magic" release with the "Time Manager", but when I
have tried, I just got some of my most spectacular system crashes.

Paul Gregory (paul@simulate.chem.vt.edu) writes:

If you are using MS Windoze you will be shocked.  You need to run OS/2 which
is truly multitasking and multithreaded.  There are a couple of OS/2 compilers
for Fortran and C/C++.

Leif Laaksonen (laaksone@csc.fi) writes:

I'm not really an expert in the field but perhaps I could contribute
as "the warning example".

I used to be a Macintosh person myself. I had a Macintosh SE/30 at my
home and a Quadra 650 in my office. First I thew out my home Macintosh
(about a year ago) and replaced it with a PC/486 DX2 66MHz. At that time
I would not have been able to come close to the "goodies" I could buy
for the PC if I would have chosen a Macintosh. Very soon I started to
like my PC running Windows 3.11 so I threw out my Quadra in my office
and replaced it with a DEC Pentium 60MHz.

The reason why I wanted to go in the PC/Windows direction was the
software. There is a lot more software for the Windows machines than
for the Macintosh. One other important aspect was the program development
environments. For Windows there are several (Microsoft, Borland, Symantec...)
development packages available and I wanted to code for this machines.
Now I'm running Slip from my home PC to our network at the computer center
and I'm very happy with the arrangement.

That was the "happy side" of the computer business. I have been for a long
time convinced that the future lies in the client/server business where
the Web is a very good example. Most of us sit daily in front of a PC or
Mac preparing documents (Word processing) or analyzing data (spreadsheets).
I can't understand why we don't already have the software for sending
batch jobs from our PC directly to our computational engines? In fact
we could do a lot of our molecular modelling on our PC and send everything
which needs computation power to an other machine. Then we could just by
clicking cut and paste the output directly into our word processors or
spreadsheet programs. The TCP/IP support on both PCs and Macintoshes is
already so strong that all this could be done. This was a long intro!
So I wanted a PC to my office to see if a PC could replace a graphics
workstation. In fact I would have wanted a 90MHz Pentium but because
the PC support person only had a 60MHz Pentium I could not get a faster ;-)
I also decide I wanted to go for Windows NT 3.5 because it supports
OpenGL and has a good TCP/IP support.

To save money the computer center decided to buy the machine from one
vendor and buy the network card, Sound Card and the CD-ROM from an other
place. The CD-ROM is a Mitsumi and there was no support for Windows NT 3.1
(which I had to start with) so the CD-ROM was out of use for weeks before
I received NT 3.5. At the time when I received NT 3.5 we discovered that
there was no driver to drive the Mitsumi CD-ROM through the Sound Blaster
card. The Sound Blaster vendor did not see NT 3.5 as that important.
The computer center has been using SMC-ethernet cards for a long time
so received one for my Pentium. A few hours later our network people
went bananas because they could see that one machine in our local
network (my Pentium) was sending the message to the network that a
randomly picked local machine was "unreachable", which resulted in that
some machines disconnected all connections to the randomly chosen machine.
So every time I switched my Pentium on some 20 persons was disconnected
from our Convex or Power Onyx or ... This was a nightmare because no one
neither the local hardware supplier, Digital, Microsoft nor SMC wanted
to take any responsibility. I was for example sitting for about 8 hours
in the Microsoft phone support queue without any help. All this is
now very simple but I can tell you that it was not very enjoyable at that
time. Finally the problem was solved by replacing the SMC network card with
a card from Digital.

What am I trying to say here? If you will add things like a network
card, CD-ROM , sound card ... buy it from one vendor so you can
blame them. Digital, DELL, Compaq ... seem to provide complete packages.

I also wanted to have a full 21" screen to be able to use the 1280x1024
pixel mode. Now the experience has shown me that DEC (forgot?) to tell
me at the purchase that the display card does not really support that...

But I'm still very hopeful about the future. I run Mosaic, NetScape,
XVision on my PC and the response is very much faster than with my Quadra.
I still have to figure out how I can use the OpenGL which is supposed
to be here somewhere on my NT 3.5. Perhaps one day I can even use my
Sound Blaster card?

To summarize my summary ;-) I think with the PC you have to be more
careful than with the Mac. If you go for the PC be sure that you know
who to blame if some parts don't work properly.

Paul Sherwood (P.Sherwood@dl.ac.uk) writes:

  First let me say that on a bad day I tend to flame people
who fill my mail box with non-chemical queries to CCL.

   But ..  as I have recently started developing code at home
on a Linux 486DX2, and being quit happy with the setup, I
thought I'd let you know that I reckon it's worth serious
consideration. In terms of stability, the OS is somewhat
easier to crash by  dud code than some commercial Unixes,
but still much better in this respect than I imagine DOS or
Mac to be. Overall the  "package" provided by Slackware is
impressive if you normally use stuff like GNU make, emacs,
Tcl/Tk, and TeX. and the X (X11R6) server is fast.  The f77
kludge (f2c+gcc) works a lot better than you might expect.

 And, you can easily boot up dos/windows when you want to
play DOOM, use MS WORD etc, or code for windows.

Charles Letner (cletner@remcure.bmb.wright.edu) writes:

> 1) Is there a way of telling whether the chip in a pentium PC is a "good"
> or "bad" chip (like the old double-sigma's)?  I've seen posts about
> tests that can be run, but is there any before-purchase things that
> can be done?
         I am a happy owner of a Pentium 90.  And yes it does have the
defective FPU (Intel is replacing).  However, if you decide to go the
Pentium route, insist that you be allowed to run a quick test program on
the actual machine that you will be getting.  You can get a little program
that will do this from:
        http://www.mathworks.com/README.html
with mosaic.  Look for the link to p87test.txt.   This is a asm program
that will run in a few seconds and if the FPU is bad will give you a line
indicating that.  There are also some links to other easy test and lots of
information in that mosaic document.  If you need more help with that let
me know and I'll do what I can.

> 2) Have folks gotten linux running on Pentiums - or are they merely
> running 486 code that doesn't know or care the difference?  I assume
> there are different instructions in a Pentium, or am I full of it
> with this?
        I don't know if linux is up and going on Pentiums but I'd be
willing to bet it is.  Yes there are different instructions for the
Pentium.  The latest borland C++ compiler (4.x) presumable takes advantage
of this, at least there is a switch for it and it does take advantage of 32
bit OS when you ask (of course you must have a 32 bit OS running).  BTW,
windows NT is a thought for a "real" OS on the Pentium (my experience with
OS/2 was verrrry bad).

> 3) Are there graphics cards that should be used and/or avoided (in
> Dos/Windows and in linux)?  Several vendors mention Diamond Stealth
> cards, for example - do these work with Linux? Win3.1? etc...
        I'm not going to say much other than if you want to run windows
and are planning on doing any graphics, a windows compatible accelerated
card with some RAM is a wise idea.  You will probably want to go with a
PCI bus and a PCI graphics card.  I have the 32 bit challenger PCI graphics
card and am very happy with it.  A friend of mine has the Diamond Stealth
card on a 486.  She was very impressed with the apparent increase in speed
with the card (apparent because the cpu didn't speed up, just graphics
updates.  She thought the cpu was going faster...).

> 6) Are the Windows development environments "good enough" to keep one
> from having to reboot the machine when the code being developed
> crashes (possibly stupid question #106)?  How about on the Mac?
        Hey another plug for NT here, it allows you to run app's in their
own memory space so if it does crash it only takes itself out.  You really
should call Microsoft and get some info on NT...  Others that I have
talked to have been happy with NT thought these are system admin. and not
computational geeks (like me).  The one aspect of NT that might be a draw
back is computational software.  I'm not really sure if there is much
available yet.  My plan for the first part of this year is to try and get
some app's going there (BTW rasmol will support NT, so there is a viewing
program available).

        Not to beat a dead horse but,  NT is really the answer here.  If
you are programing in good old windows (16 or 32 bit), your app's run
until it is finished with the work it is currently doing.  This is why
when you print something from a windows app that takes awhile, you end up
sitting and waiting.  NT on the other hand implement preemptive
multi-tasking.  Here the control of the cpu is taken away from the app and
given to the OS (NT in this case).  Who gets to do what is determined by
NT and the priorities.  But none of this is new to you if you are using a
workstation with say UNIX.  The reason I am using NT now is for many of
the issues that it sounds like you are interested in.  It is a true
workstation OS and has many of the features that you find on UNIX OS.  It
is expensive compared to DOS/win though.  I paid ~250 for version 3.5
for workstations (there is a server version for ~$700, if your machine
will be networked this could be an alternative.  You did say home computer
in your first message so this isn't an issue).  But hey,
if you plan an spending >$2000 for a pentium 90 (and more for bells and
whistles) what's another couple hundred to really use the machine.  If you
care for me to expand more on the "virtues" of NT, let me know and I'll be
happy to do so.

David Doherty (doherty@msc.edu) writes:

It's happening, especially since f77 is a subset of f90.  I'm not switching
over yet, because it's not yet on all of the machines on which I work
(although it is available).  You might also be interested to know that
NAG has even announced an f90 compiler for linux.

btw, I believe that Lang. Sys. Fortran is moving to f90 soon too.

I run unix (NetBSD/68K, available for some of the 030 macs) on my SE/30
at home, and how found f2c/gcc to be a very effective f77 compiler.

>Let me ask you all another question - how does one write applications
>for Windows or Macs to use cooperative multitasking?  As somebody

As an example, in Language Systems Fortran, there is just a compiler
switch that specifies how friendly the binary will be. Note that this is
using LS Fortran in its default mode, which means that all of what your
code does is input and output to a big text window.

If you plan to use the toolbox to any great degree (multiple windows,
graphics, etc.) you'll probably be constructing your own event loop, in
which case, I think you'll have some more control over how much of a
hog the code is.

Tom Pierce (rs0thp@rohmhaas.com) writes:

> 1) Is there a way of telling whether the chip in a pentium PC is a "good"
> or "bad" chip (like the old double-sigma's)?  I've seen posts about
> tests that can be run, but is there any before-purchase things that
> can be done?

No - Not that I know of. The Pentium FAQ is
 from: ftp://www.isi.edu/pub/carlton/pentium/FAQ

> 5) Are there fortran compilers for the Mac (There seem to be several
> good ones for the PC)?  If so, who are they, and do they have

I like and use Language Systems Fortran for the Mac.
More info from LANGSYS@aol.com or 703-478-0181

> 7) Finally are there PC vendors that I should use and or avoid?

I like MidWest Micro - a good mail order company with better-than-me
tech support ( and I keep a software development kit in my
office consisting of pliers, screwdrivers and a soldering iron)

Fred Dulles (dulles@prometheus.chem.umn.edu) writes:

> 1) Is there a way of telling whether the chip in a pentium PC is a "good"
> or "bad" chip (like the old double-sigma's)?  I've seen posts about
> tests that can be run, but is there any before-purchase things that
> can be done?
>

        Unless something has changed in the last week or so, there
is not way (i.e no marking) that can tell you whether your pentium
is good or bad.  The best source of information (and there seem to be
a lot of *bad* sources) is the WWW site run by mathworks, inc that is:

        http://www.mathworks.com/

> 2) Have folks gotten linux running on Pentiums - or are they merely
> running 486 code that doesn't know or care the difference?  I assume
> there are different instructions in a Pentium, or am I full of it
> with this?
>

        The pentium is fully 486 compatible and therefore has no
problem in and of itself running linux.  That is far from saying that
all pentium machines will run linux.  I would investigate quite carefully
before buying a machine to run linux, lest you be disappointed.  There
is a lot of good information on the net about this.


> 3) Are there graphics cards that should be used and/or avoided (in
> Dos/Windows and in linux)?  Several vendors mention Diamond Stealth
> cards, for example - do these work with Linux? Win3.1? etc...
>

        I'm sure some cards have linux drives and others don't so
        be careful, as above.

> 4) Have people used CD-ROM's that are external and swappable between
> PC's and Mac's (obviously a SCSI device)?  Any comments on who's
> to use?
>
        I'm told that if the cd-rom comes with drives for both
systems, this is quite doable.

> 5) Are there fortran compilers for the Mac (There seem to be several
> good ones for the PC)?  If so, who are they, and do they have
> anything resembling the tools on a workstation?  I'll also take recommendations
> for C compilers for the Mac...
>
        I know of two commercial compilers, one by Absoft and one
by Language Systems.  Someday (soon? in the next ten years?) the
GNU Fortran compiler is supposed to come out and run on the PowerPC
machines.  (Though some porting might have to happen since the GNU people
don't like Apple.)

> 6) Are the Windows development environments "good enough" to keep one
> from having to reboot the machine when the code being developed
> crashes (possibly stupid question #106)?  How about on the Mac?
>
        I'm not a windows person, but I think for both windows and MacOS
machines, you can't count on this kind of stability.  No doubt the machine
would survive some errors...  You'd do better but by no means perfectly
under OS/2 or Windows NT.  As far as that goes, we've occasionally crashed
out Alpha workstation when playing with graphics code, so its really a
matter of degrees.

> Yes, I'm rather used to workstations...
>
> 7) Finally are there PC vendors that I should use and or avoid?
>
        Oh yes, beyond question.  Exactly which ones depends on what
you want.  (I realize that's not super helpful, its just that it seems
to me to really depend on how much you want to spend vs. how much support
you need vs. how plush a machine you want.)

> Thanks for your comments (well, the comments I expect to get :-)!
> I'd love to get a big SGI or IBM at home, but I'd rather not be killed
> in my sleep...
>

        I assume the threat doesn't (directly) come from the workstation...

        Like you, I am a workstation person (though I am typing this on my Mac
at work.)  Last summer I replaced my IIcx with a PowerMac 6100/60 and
have been very happy, especially one I got a bigger hard-drive.  On the
other hand, if compiling Fortran is a very large part of what you want to
do, that would bias me towards a windows machine.  In my case, I just log
onto my workstation account from home if I get such an urge.  Hope this
helps,

Konrad Hinsen (hinsenk@ERE.UMontreal.CA) writes:

   Is this "magically" done by the compiler/builder or does one have to
   insert "release-cpu-now" function calls throughout compute-intensive code?

In principle the compiler could do this, but probably it would do so at
places where you don't want it. I am not aware of any compiler that
even attempts to handle that problem. In other words, you must insert
these calls yourself.

   I've just realised, that if I want to program at home, I might be in
   for a REAL shock over this issue...

Probably. It is not just a lot of work, but it also makes your programs
inefficient. The system call that checks for other waiting processes is
a lot more expensive that the task switch done by a real multitasking
system.

Personally I have solved this problem by switching to OS/2, which does
real multitasking...

Leif Laaksonen (laaksone@csc.fi) writes:

According to people here at the computing center they say that OS/2
is NOT the way to go. There will not be enough of software for that
platform. Specially when Windows32 software will NOT run on OS/2.

Linux (did you know that the kernel was written by a Finn?) is good
again but the problem is that you will not be able to use all those
1000's of programs written for the Windows market. Supporting a Unix
machine is not easy and it takes a lot of time. If you will run just
a few applications and you have the source code, it's ok.

As much as I might hate Microsoft Windows 95 and NT 3.5 seems to have
the right tools.

As I might have told I'm quite old in this business. If you would have
asked me this question a couple of years ago I would have said the most
rude words about Windows (and any graphical user interface in fact).
Now I'm used to Windows and Macintosh interfaces and I would not like
to go back to a Unix type of interface.

Joe McDaniel (joe@psiint.com) writes:

It really is very easy -- every time you update the window, read or write to
disk, etc., you relinquish control.  The only time you need to do something
explicit is when you are in the middle of a big compute loop with no screen
updates, etc. -- then you could "lock up" the system until that task
completes.  There are two approaches, explicitly relinquish control or do
what almost everyone does -- show a status bar that, when updated, will
relinquish control anyway.

Konrad Hinsen (hinsenk@ERE.UMontreal.CA) writes:

>    Ouch - it strikes me as a potential headache looking at MOPAC (for 
> example)  This was developed w/o consideration of time slices, so to go 
> throughout the code and insert function calls...  *sigh*

No numerical program I know of has been developed with cooperative
multitasking in mind (and for good reasons). Retrofitting such a
support takes a lot of time and requires a detailed knowledge
of the algorithms used - it is not a trivial task to find the
right points for inserting the necessary calls.

Systems with cooperative multitasking are simply unsuitable for
numerical background calculations (and other CPU-intensive things).

> I guess Cache and Hyperchem worked hard at it, since I thought their
> codes run with cooperative multitasking...  Not having seam them,
> (seen them, sorry), I've assumed...

> Have you played with such packages?

No, I don't need them, and I don't use systems with cooperative
multitasking. It is certainly possible that they have invested
a lot of effort to adapt their code; if I were a commercial
developer dependent on the Windows or Mac market, I would
probably do that. But as long as I can choose my operating system,
I will go for a real multitasking system.

Charles Letner (cletner@remcure.bmb.wright.edu) writes:

Glad I could help.  Something to consider with with Chicago/windows
95/daytona etc...  It is not a full 32 bit OS.  Parts of it are
implemented as 32 bit and parts will be 32 bit.  Also, are there plans for
preemptive multitasking?

Paul Gregory (paul@simulate.chem.vt.edu) writes:

I used to use Windoze, I hated it.  It is not a stable computing
environment.  I moved to OS/2 at version 2.0.  It is now at 3.0.
I compile a Fortran program with MS Fortran that I also use on an IBM RS/6000.
Of course, I don't have all the memory limitations of DOS/Windoze.
I don't do any real production on the PC.  I mostly use it to optimize/
modify/test  the Monte Carlo simulation code that runs on the RS/6000.

MS Fortran includes all those graphic functions and threading functions,
although I haven't played with them much.  I think there are more
Fortran compilers for OS/2 that are newer/better.

Walt Reiher (wallyr@netcom.com) writes:

>1) Is there a way of telling whether the chip in a pentium PC is a "good"
>or "bad" chip (like the old double-sigma's)?  I've seen posts about
>tests that can be run, but is there any before-purchase things that
>can be done?
No, there is no way of telling just by looking at it.  (Besides, you'd have
to open the thing up!)  There's a VERY little program I downloaded from
comp.sys.intel which tells you if your CPU is naughty or nice.  It can be
run from a floppy.  You can download it from my anonymous FTP directory:
ftp.netcom.com/pub/wa/wallyr/p87test.zip
     By the way, the fixed Pentiums ARE trickling out of Intel:  I just
installed a fixed CPU in my machine this week.

>2) Have folks gotten linux running on Pentiums - or are they merely
>running 486 code that doesn't know or care the difference?  I assume
>there are different instructions in a Pentium, or am I full of it
>with this?
I've read posts by lots of people with Linux running on a Pentium, but I've
been unable to get it to work (see below).
     My understanding is that the instruction sets are either identical or
very close; 486 code will run on a Pentium.  However, the timings for
instructions are very different and it appears that code well-optimized for
a 486 is NOT very well-optimized for a Pentium.

>3) Are there graphics cards that should be used and/or avoided (in
>Dos/Windows and in linux)?  Several vendors mention Diamond Stealth
>cards, for example - do these work with Linux? Win3.1? etc...
I have a Diamond Viper PCI video card in my Pentium and it runs great with
Win3.1 and WinNT (>> better than Win3.1).  But, Diamond cards are NOT
compatible with X Windows under Linux (and other "free" UNIXes).  If you
only want a VT100-like or a VGA screen (much lower resolution, almost
useless for X Windows), then Diamond's OK.
     The other thing that's making Linux difficult for me is that I got a
great PCI SCSI interface (Adaptec), but it's too new for a Linux driver.
There's supposed to be one in alpha test, though.

>4) Have people used CD-ROM's that are external and swappable between
>PC's and Mac's (obviously a SCSI device)?  Any comments on who's
>to use?
Yes.  It can be done, but I do have a CD-ROM that works with my Mac but not
my PC (the one that came with my SGI; it's NOT supposed to work ANYWHERE
else--the Mac is apparently a fluke).
     I just bought a new Toshiba quad-speed SCSI CD-ROM (3501) that goes
both ways.

>5) Are there fortran compilers for the Mac (There seem to be several
>good ones for the PC)?  If so, who are they, and do they have
>anything resembling the tools on a workstation?  I'll also take recommendations
>for C compilers for the Mac...
I looked into this bit, and it appears the FORTRAN compiler of choice is
the Language Systems compiler.  I don't have it, though.  Unless you get a
Power Mac and get a Power Mac compiler, it seems to me that running
computational programs on a Mac is a waste of time (I love my Mac, but it
really isn't fast enough:  remember, it's a Sun3!).
     What have you heard on the PC side?  I've been looking into this.  I
can tell you that the Watcom compiler looks great on paper but has been
giving me all kinds of trouble; I'm getting a new one.

>6) Are the Windows development environments "good enough" to keep one
>from having to reboot the machine when the code being developed
>crashes (possibly stupid question #106)?  How about on the Mac?
Haven't really developed on the Mac, but Windows 3.1 is inherently unstable
and you can count on it crashing frequently if you are a serious computer
user accustomed to a stable UNIX system--even withOUT crashing
developmental code.
     However, I have found Windows NT 3.5 to be a MUCH superior environment
to work with.  It is MUCH more stable than Windows and not very prone to
going out with developmental code.

>7) Finally are there PC vendors that I should use and or avoid?
See the June 1994 PC World for the results of a user reliability and
service survey.  I bought two Pentiums from a local systems integrator who
emphasizes customer service and free labor on repairs, and I didn't spend
much more than buying by mail order (weigh your local sales tax vs.
shipping costs).

>If the quoted SPECfp92's are to be believed, the P90's are looking a lot
>like the Indigo R4K on my desk...  However, since I'm not into the
>hardware game, the bit about Linux support has me concerned.
My 90MHz Pentium IS quite fast, although it's hard for me to quote a
reasonable benchmark as I can't get MOPAC to run!  In a "stupid little test
case", unoptimized code on the P90 is running 3x the speed of -O2 code on my
4D/25 (don't remember how to convert to a R4K Indigo).
     On the other hand, I was quite disappointed to see the Gaussian
benchmarks John McKelvey quoted on CCL recently.  They indicate the P90 is
much slower than I had hoped.  John's benchmark should be more reliable than
mine...

>Also, two of the Hyperchem developers (Tom Slee and Graham Hurst) have
>got with Watcom - sez they've expanded by 2x over the last year or so...
I don't know if trying to "buy Canadian" has anything to do with it, but one
of my co-workers is also fed up with Watcom for other reasons--their
database product.  I won't go into all the details, but he's experiencing
many of the same kinds of problems I'm having with FORTRAN, which makes me
highly suspicious.  The next place I'm going to look is Microway.

Javier Modrego (modrego@icma0.unizar.es) writes:

I have read your recent questions on CCL about home computer choices and
cooperative multitasking. I am able to answer just a few of them. I don't
know very much about PC but I am also a very happy user of a Mac IIci.
I have been using Fortran in my mac for three years. There are several 
compilers for the mac but the main contenders are the compiler from Language 
Systems and the one from Absoft. Both work inside the MPW (Macintosh 
Programing Workshop) shell. This is a very Unix-like shell which allow the 
integration of many different compilers. In fact you can develop you Mac 
interface in C an your numerical code in Fortran and then link them together 
(you can also try to build the whole interface in fortran also). The compiler 
I have been using is the Language Systems Fortran. I am very happy with it. It 
has many extensions from Vax-Fortran and I have been able to compile and run 
programs from Vax without any change in the code, of course without any fancy 
interface, just calculations. You can ask for information to langsys@aol.com. 
The compiler at the moment runs very well in Mac, I have also purchased the 
upgrade for PowerPC Macs, but I have not been able to test it as yet because 
I have not access to a Power Mac at the moment.

About your questions on C compilers, there are several for the mac. Some run
inside the MPW shell, others are standalone compilers. In the first group are
the compilers from Apple computer and the MPW version of THINK C. On the other
side there is also a version of think C and the compiler from Metrowerk.
There is also a version af GNU C for MPW, but you need the C libraries for
MPW which come with the C compiler from Apple (they are included with the
Language Systems Fortran compiler).

You asked if you need to reboot the computer when the program in development
crashes. The answer is no, if you have Macsbug ( a machine code level 
debugger) or other debugger installed. Most compilers include Macsbug or any 
other debugger.

The question on cooperative multitasking is an easy one. If you want GOOD
cooperative multitasking you must include the code for it in your program at
the appropriate places. I don't know about other compilers. Language Systems
Fortran can do it for you. You can choose several levels of "cooperativity"
at compile time and the compiler will put the necessary system calls in
your code, depending on the level of cooperativity chosen. It can be 0 and
the program will take over the computer while it is running, of course. The
lower the cooperativity the faster the program.

I hope this can be of use to you.  I have included below the abstract of a 
discussion about fortran for the Mac in Infomac which was compiled in 1992. 
Please note that the address of Language Systems shown below is obsolete, now 
is
Language Systems
100, Carpenter Drive
Sterling, VA 20164
The reference to several version numbers can be also outdated.

Alun Carr (ajcarr@ccvax.ucd.ie) writes:

Here is the collection of responses I received in response to my
query about the 'best' (most VAX-compatible, fastest) FORTRAN
compiler for the Mac. It looks like the Language Systems
compiler is the most often recommended.

Many thanks to all those who responded, especially to those who
I didn't manage to reply to individually.

Dr Alun J. Carr
Mechanical Engineering Department
University College Dublin
Belfield
Dublin 4
Ireland

Internet: ajcarr@ccvax.ucd.ie
Phone: +353-1-7061989/2693244
Fax: +353-1-7061756/2830534

====== Report Begins ===========================================

Date: Sat, 31 Oct 92 20:50:09 CST
From: gray@cmgroup.engr.wisc.edu (Gary L. Gray)
Subject: FORTRAN compilers (A)

I have been using Language Systems FORTRAN for almost a year
and I couldn't be happier.  It has a compiler invocation
option called "-vax" which sets up several options that
mimic the VAX environment (it has similar options for Cray
and Microsoft PC FORTRAN).  LS Fortran runs under MPW and is,
as far as I can tell, virtually bug-free.  The compiled code
is VERY fast and the are dozens of compile-time options for
running in the background under System 7.x or MultiFinder,
having a nice Mac interface and many other things.  I also
know that it is compatible with the Daystar PowerCache.  I
believe the educational price is $375 ($595 retail), but that
includes the latest version of MPW and all the manuals for
MPW.

I have also used Absoft's version of FORTRAN (both 020 and the
MPW versions) and find the error diagnostics to be much less
satisfactory and the interface to be less easy to use.  The
compiled code runs at about the same speed in the top-of-the-
line Absoft FORTRAN as with LS FORTRAN, but you must get the
most expensive Absoft product.  The technical support is
excellent with both companies, but Language Systems has help
available via Internet both through AppleLink and America
Online (Absoft may also, but I know LS does).

Language Systems has also been great about keeping their
products current and compatible.  I think it is the way to
go.

Good luck.

Gary L. Gray                *  Engineering Mechanics & Astronautics
gray@cmgroup.engr.wisc.edu  *  University of Wisconsin-Madison

------------------------------

Date: Fri, 30 Oct 1992 22:09:05 EST
From: bouldin@anvil.nrl.navy.mil
Subject: Fortran for the Macintosh (A & C)

The question of what fortran compiler to use on the Mac has come up again, so
I would like to repeat my strong endorsement of the Language Systems Fortran
that runs under MPW. Bug free, fast compiles, compact code, fast execution,
good debugging, full toolbox support, good tech support, full VAX extension
support, I *love* it.

The worst thing I can say about it is that you have to run MPW, which is Unix
in hiding on the Mac. It is a powerful environment, but really confusing and
needlessly complex at times. For large projects, though, which is what old
fortran codes tend to be, it is quite a powerful environment. I maintain and
use 50,000+ lines of fortran code that formerly ran on VAX and PDP/11, and the
conversion wasn't too arduous.

Equal time for opposing views: Absoft also sells a fortran for MPW. It may
be a little faster on execution, though I doubt it for anything but
benchmarks.  They may also be a better choice if you will later move to Unix 
or elsewhere, since they are a multi-platform fortran; LSF is only Mac.

Finally, you might consider using the F2C translator, if you really like C.
I don't and I am also not aware of an implementation of F2C for the Mac.
Anyone know??

------------------------------

Date: Sat, 31 Oct 92 11:00 CST
From: "randy k. hayashi" 
Subject: fortran stuff

Date: Fri, 10 Apr 92 13:04 N
From:  (HEWAT@FRILL.BITNET FAX FRANC
   E[33] 76-48-39...)
Subject: Best FORTRAN compiler for Mac (answer to recent questions)

Distribution-File:
        info-mac@sumex-aim.stanford.edu

>Recently someone asked about existing FORTRAN compilers for the Mac.
>I do not recall seeing any replies written to the digest.  Is there
>such a compiler, or at least a good FORTRAN code editor available.

I have used various versions of 3 different compilers: Language Systems,
Absoft and MicroSoft (its OK, the later was really written by Absoft, and
Absoft still market an updated version). Each has its advantages.

1) MacFortran 2.1 (MS-absoft) has a debugger built in, that allows you to
step through the code, display the contents of variables etc. (cf The
advantages of THINK Pascal over MPW Pascal).  However it has a non-standard
way of accessing the Mac tool box, and it is getting rather old, and not
a high priority with absoft for updates.  Even so, it produces fast code
that even runs under system 7 and the new Macs including the Quadra !
(amazing !) It can compile for non-fpu Macs. And it is less expensive than
the others.

2) MacFortan-II (absoft).  Completely different beast. Produces the fastest
code, especially for the Quadra, many Fortran-90 extensions, compatible with
most Vax and IBM extensions, source compatible with other absoft compilers
for a wide range of workstations, uses Fortran-90/Pascal structures etc.
A professional compiler, but not completely happy with old spaghetti type
code produced by amateurs.  No in-built debugger, runs under MPW only, and
only on fpu equipped Macs. Produces code only for fpu-Macs. Standard Pascal
type toolbox calls, and elementary Mac interface (menus, windows) produced
automatically if required.

3) Language Systems. Code almost as fast as MacFortran-II, not quite as fast
for the Quadra, but maybe faster at compiling.  Fewer options and easier to
use. Produces slightly better automatic Mac interface. Again an MPW compiler
using Fortran-90/Pascal structures, making tool box calling as easy as from
Pascal. No in-built debugger. Produces code for non-fpu Macs as well as for
fpu. Compatible with most Vax and IBM extensions.

CONCLUSION: 
It depends on what you want and who you are. For amateurs developing
programs without a Mac interface, Macfortran 2.1 is fine.  If you want the
fastest code, especially on the Quadra, and you don't care about lower class
Macs without fpu, you will upgrade to MacFortran-II.  If the difference
between 1.7 megaflops and 1.3 on a Quadra is not important, you will go for
Language Systems.  Modern Fortran is still the way to go for numeric
calculation, but you may want to write the Mac interface in THINK Pascal,
(a really friendly engine) avoiding their extensions so that it will also
compile with MPW-Pascal, in which case you can easily link in the Fortran
routines for number crunching.

Alan Hewat, ILL Grenoble, France (email Hewat@FRILL.bitnet)

------------------------------

Date: 12 Apr 92 19:07:00 EDT
From: "Charles E. Bouldin" 
Subject: Fortran Compiler Opinion

Since there have been several requests for info about fortran compilers:
I highly recommend Language Systems Fortran. This excellent compiler is
almost totally VAX compatible, bug free, compiles fast, and produces
code that executes quickly. Using it, I have freed myself from relying
on an ancient PDP-11 for data collection and from a (almost as ancient)
VAX for data analysis. Something like 60,000 lines of PDP and VAX fortran
was entirely moved over to the Mac using the Lang. Sys. fortran.

I concur that the Absoft MPW fortran is slightly better for 040 machines, but
the Lang. Systems is faster (in my tests) on real code on my 030 machines
than Absoft is. Absoft does win the standard benchmarks: Sieve and Whetstone,
but loses on my "real world" test cases. Absoft also takes a *lot* longer to
compile than LSF. I also find the LSF much simpler to use. The only case where
Absoft is substantially faster than LSF is code that has double precision
complex arithmetic. For those of you who are really committed to fortran,
there is an add-on product called AppMaker (similar to Prototyper) that will
let you do point-and-click design of an application shell and then generate
the fortran source for it. Then add the numeric code as subroutines tied to
your menus and you have a 100% fortran program with a Mac interface.

I strongly discourage any use of MacFortran/020; that compiler is LOADED WITH
BUGS!! Some of the bugs produce code that executes INCORRECTLY!!! Avoid it.

LSF is at (703) 478-0781. I *do* beta test for them, but they don't pay me
(too bad!); I just really like the product

------------------------------

Date: 1 November 1992 16:57:41 CST
From: Stephen Kawalko 
Subject: RE: FORTRAN compilers for the Mac (Q)

I don't want to start a flame war but there are a significant number
of scientific programmers using Fortran. Most of the major numerical
libraries LAPACK, IMSL, NAG, etc. are still written in Fortran. More
important is the fact that on most vector-based high performance
computers (e.g. Cray Y-MP) Fortran compilers do a better job of
vectorizing code than a C compiler.

But that is not what you wanted. Language Systems produces a MPW-based
Fortran compiler. It was recommended  to me when I was looking for a
Fortran compiler. However, I never did buy it due to a lack of
sufficient funds.

  Language Systems Corp.
  441 Carlisle Drive
  Herndon, VA 22070

  phone: 703-478-0181
    fax: 703-689-9593

  price: $495 for Fortran compiler and MPW
         $400 for Fortran compiler (without MPW)

Stephen Kawalko (u40857@uicvm.cc.uic.edu)

------------------------------

Date: Mon,  2 Nov 1992 14:25:20 UTC+0100
From: Javier Modrego 
Subject: Fortran for Mac (A)

I have been using Language System Fortran only for short time, but
I think it is a very good choice. Porting Fortran programs from
Vax is very straightforward, you can have your program running almost
immediately with minor modifications. It has an open(n,file=*) option
which brings the standard dialog box to the screen allowing you
to connect any mac file to the n fortran unit. Just modifying the
open sentences in that way you can use almost any vax fortran source
without further changes. The compiler understands most vax-fortran extensions
although it ignores some of them. Of course, if you want to build a complete
mac application things are more complicated.  It works inside MPW, but MPW
is a complex system (very UNIX-like). The good point is that you can
easily link the object files from different languages. This allow you
to write your numerical routines in fortran and you mac-interface routines
in mpw-c or pascal. You can also use LS-Fortran to access the Mac-OS routines
because the compiler comes with all the libraries you need so you can write
a complete mac application using just LS-Fortran
The academic price is $375, including MPW and its set of manuals. The manual
for the compiler itself is very good, form my point of view.
If you have already MPW the price must lower.

        Javier Modrego
        University of Zaragoza
        Spain

------------------------------

Date: Wed, 4 Nov 92 04:52:04 CST
From: gray@cmgroup.engr.wisc.edu (Gary L. Gray)
Subject: f2c for Think C on the Mac

After a little digging I was able to find where I obtained f2c.
Just ftp to:

           elpp1.epfl.ch

and the rest should be pretty clear.

It looks like they have put a newer version than the one I
previously had (the current version appears to be dated May
1992).  It is 797767 bytes in a file called f2c_5_92.cpt_hqx.

Good luck.

Gary L. Gray                *  Engineering Mechanics & Astronautics
gray@cmgroup.engr.wisc.edu  *  University of Wisconsin-Madison

------------------------------

Date: Mon, 2 Nov 92 11:57 MET
From: KEES.RAPPOLDT@IB.AGRO.NL
Subject: Fortran compiler question

Hello Alun Carr,

Some experience from me, a physicist working in soil physics with
physical and biological models from agriculture and biology. During
the last years I have used The Absoft compiler on the Mac. In short:
don't buy it. The Absoft compiler is fast ... and buggy. It works
for perfect FORTRAN programs, but even simple syntactical errors tend
to lead to corrupt code and a hanging machine. So, although I am
good in Fortran, I sometimes needed the VAX compiler or the Microsoft
Fortran compiler on a DOS machine to get my code right.

Recently I decided to finish this story and to buy the MPW Fortran
compiler from Language systems, regularly advertised in MacWorld.
After a little getting used to MPW, I compiled about 10000 lines of
code in standard Fortran-77 without a single problem leading to a
working program. Moreover, there are a lot of options build in,
VAX fortran extensions, CRAY fortran extensions !! So when you get
programs from other people, usually NOT free from language extensions,
you will usually be able to compile with the LS compiler.

The LS compiler generates a window for the application that can be
left open in the background without anything to arrange for it in
the Fortran code. After years with the Absoft thing this was nice to
see.

A final word about 'C'. There is no doubt this is the best choice for
everything that has to do with the machine and the operating system.
Scientist, however, have to survive machines and operating systems
and need a perfectly standardized language which runs flawlessly on
different platforms. Fancy and machine dependent user interfaces
have to be written ABOVE models and not IN models. Large programs
which model natural systems should be 100% (and not 99%) machine
independent. I admit, this could probably be achieved with C as well,
but who will translate and test the huge base of existing code ?

An advise on Fortran on the MAC: many Fortran users work on VAX and
PC with the VAX Fortran and the Microsoft compiler. These compilers
generate code with so-called static memory use. Variables are
initialized zero and retain there value between successive
subroutine calls. This, however is NOT standard Fortran. With the
absoft compiler you need to add a SAVE statement in each routine
in order to achieve static memory use and you have to check
carefully for initialization errors in the program. The MPW compiler
can be instructed to do so without SAVE's and I think can also be
instructed to set everything at 0 initially.

say that the MPW-Language Systems compiler is much better than
Absoft's product.

Good luck with the chrystals,
Kees Rappoldt, Groningen, The Netherlands.
RAPPOLDT@IB.AGRO.NL

------------------------------

Date: Wed, 4 Nov 92 23:37:32 CST
From: GA0095@SIUCVMB.SIU.EDU
Subject: RE: Fortran compilers for Mac

I don't know whether you got any responce by now or not (I am a bit behind
with reading Info-Mac), but here it goes just in case.  The choice is really
limited to Language Systems Fortran and MacFortran II from Absoft.  Both run
under MPW and produce a similar performance on average.  LS Fortran seems to
be a bit more friendly to many people, but Absoft's product can be tweaked
more for faster calcs.  Absoft's product requires FPU, though.  LS Fortran
is, at least was last time I checked, somewhat cheaper and they offer edu
discounts.                     Robert  ga0095@siucvmb.siu.edu



Modified: Wed Jan 11 17:00:00 1995 GMT
Page accessed 11317 times since Sat Apr 17 17:29:27 1999 GMT