Q4071@pucc.Princeton.EDU (Interface Associates) (07/01/89)
For some weeks, I have been doing some number crunching using the
University's 3081 and FORTRAN. I am looking to offload some of the
tasks onto my 386-based PC, and am looking for a good FORTRAN. I do
not know UNIX, and do not have the time to learn for this project,
so I am looking for a DOS or OS/2-based language (non-microchannel).
I have upgraded to 8MB on the mother board and an 80387 coprocessor.
I understand that OS/2 will use up a lot of my memory, all 8MB of which
I need for arrays. So I am looking for a FORTRAN which will:
1. Run under DOS 3.3, but using storage above the 640K line
with no silly 64K segment restrictions. I do not need
code above the 640K line.
(If the choice is OS/2 or UNIX or nothing, I will have to
use one of them -- not completing the project is not an
option.)
2. Is fast (can use the 80387 AND has a good optimizer).
A good set of scientific subroutines
would be a big plus since I lose the IMSL in the move to PC.
The improvement in VSFORTRAN between opt(0) and opt(3) for
the routines I am using is 3:1 in execution time. A good,
coherent section on 'writing efficient code for XYZ FORTRAN'
which explains the compilers idiosyncrasies will go a long
way toward fulfilling this need.
3. Has usable diagnostics/debug. (I can debug on the 3081 and
download if I must).
4. Has virtual storage/paged array capabilities. When I come
to need this, I will NEED it, but I don't need it for a
while, and can give it if I have to.
5. Is FORTRAN/77 compatible. I suspect that any offerings will
be F'77, but most of my routines were written in HX to get
faster execution, and only recently converted to 77 because of
the IMSL. Reconversion would take less than a day.
6. Has a terrific debugger.
Although price is an object, the price I am willing to pay is related
to how well the above criteria are met. The only suggestion I have
been given is Leahy FORTRAN, but I don't know that much about it.
Please reply by E-mail (or to the net if you can't get through).
I will summarize if there are many responses.
=========================================================================
Robert A. West c/o Interface Associates, Inc. (Q4071@PUCC)
US Mail: 666 Plainsboro Rd. Office Commons, Suite 1A, Plainsboro NJ 08536
Voice : (609) 275-5711davef@lakesys.UUCP (Dave Fenske) (07/04/89)
The LPI compiler is ansi 77 compliant, plus it has numerous VAX extensions. The price isn't too bad either. DF
heathen@blake.acs.washington.edu (Mark Jungck) (07/06/89)
In article <792@lakesys.UUCP> davef@lakesys.UUCP (Dave Fenske) writes: >The LPI compiler is ansi 77 compliant, plus it has numerous VAX extensions. > >The price isn't too bad either. > >DF Who produces the LPI compiler? Where can I get some more info on it? What does it cost? Mark Jungck topexm@max.acs.washington.edu
mattern@ncoast.ORG (Duane L. Mattern) (07/06/89)
> >For some weeks, I have been doing some number crunching using the >University's 3081 and FORTRAN. I am looking to offload some of the >tasks onto my 386-based PC, and am looking for a good FORTRAN. > > 2. Is fast (can use the 80387 AND has a good optimizer). > A good set of scientific subroutines > would be a big plus since I lose the IMSL in the move to PC. >Robert A. West c/o Interface Associates, Inc. (Q4071@PUCC) >Mail: 666 Plainsboro Rd. Office Commons, Suite 1A, Plainsboro NJ 08536 >Voice : (609) 275-5711 ----------------------------------------------------------------------- You don't have to lose IMSL. IMSL makes a PC vers. of the same routines that are available on the mainframe. It runs under DOS, linked to Fortran, but I was using it from C. The routines are reasonable fast on the PC, but there are probably faster routines elsewhere, but you don't have to change any calling statements if you eventually want to port it back to a mainframe. The packages are reasonable priced. Back in 1986 at Purdue, we paid $350 or so for a couple of disks of object module from their IMSL library. Give them a call. --------------------------------------------------------------------- | Duane Mattern (216)433-8186 (mattern@ncoast.uucp) | | Sverdrup Technology, Inc. at NASA Lewis Research Center| | 21000 Brookpark Rd, M/S 77-1 Cleveland, Ohio 44135 | ---------------------------------------------------------------------
andy@mks.UUCP (Andy Toy) (07/08/89)
Check out WATFOR77 and WATFOR87 (with math coprocessor support).
I have used these and various other compilers from WATCOM while
I was at the University of Waterloo and they work quite well.
WATCOM
Order: 1-800-265-4555
USA:
Suite 306-21
1430 Massachusetts Avenue
Cambridge, MA 02138-3810
Canada:
415 Phillip Street
Waterloo, Ontario
N2L 3X2
Telephone: 519-886-3700
Telex: 06-955458
FAX: 519-747-4971
--
Andy Toy, Mortice Kern Systems Inc., Internet: andy%mks@watmath.UWaterloo.ca
35 King Street North, Waterloo, UUCP: uunet!watmath!mks!andy
Ontario, CANADA N2J 2W9 Phone: 519/884-2251 FAX: 519/884-8861