I think the crucial part at the moment is whether g77 supports enough F90
features in order to compile ADF. Many routines have been re-written in
Fortran90 since about 2 or 3 years ago, and new developments are carried
out in F90 as well.
I had no trouble compiling ADF2.3 and older versions with g77 on linux
boxes some years ago. I doubt that g77 has implemented enough F90
features in order to compile ADF 99 and higher versions.

Maybe the new Intel F90 compiler would be a good alternative for the PG
compiler. My colleague obtained some promising benchmarks with it. To test
it with ADF, however, means that SCM needs to provide the necessary object
files for this compiler for the routines for which the source code is not
shipped with ADF (these are, by the way, the licence checking routines. it
is not so quite surprising that they are not shipped in source form with
ADF...). Here is the download site for the Intel compiler:
http://developer.intel.com/software/products/eval/

Another option could be the VAST/F90 compiler. I used it some years ago
with reasonable success in order to compile small F90 programs on Linux. I
think it translates F90 to F77 and invokes the g77 afterwards. It's
difficult to predict, though, if it can handle a program as big as ADF.
http://www.psrv.com/lnxf90.html

Cheers,
Jochen

On Sun, 14 Oct 2001, Carlos Pereira wrote:

> >Has anyone attempted to compile ADF with the native linux fortran
> >compiler (g77). It's made large improvements in the last year or so of
> >releases so I've been using it exclusively. I'd prefer to use it for this task
> >as well (as opposed to purchasing the supported portland group compiler)
> >Has anyone else tried this?
>
> Although I didn't have much time to play with ADF in
> the last 18 months, I have it running on a UP2000
> double-Alpha machine and on a double-Intel machine too.
>
> I fully share your opinion, all my computers run Linux
> and in my opinion this ADF dependence of PG compiler
> is the main ADF bug at the moment.
>
> Two years ago, I did compile ADF97 on the Intel platform
> with g77, after some tweaking. It's simple when you know
> it, I wrote an HOWTO that I can send to you if you wish.
>
> Unfortunately with the new version things become more
> complicated and I just gave up, installing the binary
> instead. My interest in ADF decreased considerably since
> then, because science and black boxes are not compatible.
>
> Some parts of ADF code are even secret (closed source)
> because they are [owned|copyrighted|patented] by someone
> else, I was told by Bert te Velde, long time ago.
>
> This is fine if I just want a recipe to get
> results, but if I am really trying to understand
> things, then this kind of procedure is unacceptable.
>
> Carlos Pereira
>
Received on 2001-10-22 16:58:34