Xref: utzoo comp.lang.fortran:3856 comp.sys.super:234 comp.unix.cray:186
Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!ucsd!ucbvax!ucsfcgl!seibel
From: seibel@cgl.ucsf.edu (George Seibel)
Newsgroups: comp.lang.fortran,comp.sys.super,comp.unix.cray
Subject: Re: Wanted: Cray Fortran -> Sun Fortran filter
Message-ID: <15533@cgl.ucsf.EDU>
Date: 27 Sep 90 08:44:00 GMT
References: <415@kaos.MATH.UCLA.EDU>
Sender: daemon@cgl.ucsf.edu
Reply-To: seibel@cgl.ucsf.edu (George Seibel)
Distribution: na
Organization: Computer Graphics Lab, UCSF
Lines: 38

barry@pico.math.ucla.edu (Barry Merriman) writes:
>I'm trying to port a Cray Fortran code to Sun Fortran (f77)
>(the direction of the future? -). Is there a filter available
>to do this?
>There is a filter to convert from Vax VMS Fortran to Sun, so
>a Cray -> VMS filter would be enough. 
>The Cray code runs under the CTSS op sys, on the NERSC Crays
>(formerly MFE) by the way. 

Cray Research has only recently started putting a lot of extensions
in their compiler, so there's a good chance that what you have is
mostly standard fortran.   If you are real lucky, all you'll have
to do is comment out the "call dropfile()" that CTSS insists all
programs contain.   You probably won't be quite that lucky, though.
If the code was tweaked for the cray, there's a good chance you'll
find some calls to library routines, like the CVMGx() series.  You'll
need to provide fortran (or whatever) versions of all those, or
recode the program.   Now that the cray compilers are a lot smarter
than they used to be, some of the old library routines are no longer
required for good performance.

  You might find that the software is relying on ints and reals being
the same size, which is ok on a cray where they are both 64 bits; but
if the software is not numerically stable in 32 bit precision, you
might have some trouble making it double precision.  Or maybe not.  It
all depends on whether or not the original programmer knew what he was
doing, or cared.   At any rate, I kinda doubt that you need a filter,
(or that a robust/complete one exists) rather I suspect a fortran
version of any library routines your code calls is what you'll need, if
anything.   Of course, if the code is really bizarre, all bets are off...
Your VMStran converter would probably clean up some extensions that
might be found in "Cray" fortran.   Wouldn't it be nice if people just
wrote portable code?  If we'd gotten a new ANSI standard around 1983
that just added the simple extensions that have become almost-de facto
standards?   blah blah blah... oh well.  good luck.

George Seibel, UCSF
seibel@cgl.ucsf.edu